chore: fix two semantic merge errors in SymM mvcgen

This PR modifies `#eval e` to elaborate `e` with section variables in
scope. While evaluating expressions with free variables is not possible,
this lets `#eval` give a better error message than "unknown identifier."

Example:
```lean
section
variable (n : Nat)
/-- error: Cannot evaluate, contains free variable `n` -/
#guard_msgs in #eval n
end
```

The error is localized to `#eval`. It would be more friendly if the
error were to be placed on uses of free variables.

[Zulip
discussion](https://leanprover.zulipchat.com/#narrow/channel/270676-lean4/topic/Unknown.20identifier.20error.20messages.20for.20.60.23eval.60/near/560864544)

.claude/CLAUDE.md

@@ -1,30 +1,28 @@
-To build Lean you should use `make -j -C build/release`.
+(In the following, use `sysctl -n hw.logicalcpu` instead of `nproc` on macOS)
+
+To build Lean you should use `make -j$(nproc) -C build/release`.
 
 ## Running Tests
 
-See `doc/dev/testing.md` for full documentation. Quick reference:
+See `tests/README.md` for full documentation. Quick reference:
 
 ```bash
 # Full test suite (use after builds to verify correctness)
-make -j -C build/release test ARGS="-j$(nproc)"
+CTEST_PARALLEL_LEVEL="$(nproc)" CTEST_OUTPUT_ON_FAILURE=1 \
+make -C build/release -j "$(nproc)" test
 
 # Specific test by name (supports regex via ctest -R)
-make -j -C build/release test ARGS='-R grind_ematch --output-on-failure'
+CTEST_PARALLEL_LEVEL="$(nproc)" CTEST_OUTPUT_ON_FAILURE=1 \
+make -C build/release -j "$(nproc)" test ARGS='-R grind_ematch'
 
 # Rerun only previously failed tests
-make -j -C build/release test ARGS='--rerun-failed --output-on-failure'
+CTEST_PARALLEL_LEVEL="$(nproc)" CTEST_OUTPUT_ON_FAILURE=1 \
+make -C build/release -j "$(nproc)" test ARGS='--rerun-failed'
 
-# Single test from tests/lean/run/ (quick check during development)
-cd tests/lean/run && ./test_single.sh example_test.lean
-
-# ctest directly (from stage1 build dir)
-cd build/release/stage1 && ctest -j$(nproc) --output-on-failure --timeout 300
+# Single test from tests/foo/bar/ (quick check during development)
+cd tests/foo/bar && ./run_test example_test.lean
 ```
 
-The full test suite includes `tests/lean/`, `tests/lean/run/`, `tests/lean/interactive/`,
-`tests/compiler/`, `tests/pkg/`, Lake tests, and more. Using `make test` or `ctest` runs
-all of them; `test_single.sh` in `tests/lean/run/` only covers that one directory.
-
 ## New features
 
 When asked to implement new features:
@@ -32,8 +30,6 @@ When asked to implement new features:
 * write comprehensive tests first (expecting that these will initially fail)
 * and then iterate on the implementation until the tests pass.
 
-All new tests should go in `tests/lean/run/`. These tests don't have expected output; we just check there are no errors. You should use `#guard_msgs` to check for specific messages.
-
 ## Success Criteria
 
 *Never* report success on a task unless you have verified both a clean build without errors, and that the relevant tests pass.
@@ -41,9 +37,13 @@ All new tests should go in `tests/lean/run/`. These tests don't have expected ou
 ## Build System Safety
 
 **NEVER manually delete build directories** (build/, stage0/, stage1/, etc.) even when builds fail.
-- ONLY use the project's documented build command: `make -j -C build/release`
+- ONLY use the project's documented build command: `make -j$(nproc) -C build/release`
 - If a build is broken, ask the user before attempting any manual cleanup
 
+## stage0 Is a Copy of src
+
+**Never manually edit files under `stage0/`.** The `stage0/` directory is a snapshot of `src/` produced by `make update-stage0`. To change anything in stage0 (CMakeLists.txt, C++ source, etc.), edit the corresponding file in `src/` and let `update-stage0` propagate it.
+
 ## LSP and IDE Diagnostics
 
 After rebuilding, LSP diagnostics may be stale until the user interacts with files. Trust command-line test results over IDE diagnostics.
@@ -59,7 +59,7 @@ Follow the commit convention in `doc/dev/commit_convention.md`.
 **Title format:** `<type>: <subject>` where type is one of: `feat`, `fix`, `doc`, `style`, `refactor`, `test`, `chore`, `perf`.
 Subject should use imperative present tense ("add" not "added"), no capitalization, no trailing period.
 
-**Body format:** The first paragraph must start with "This PR". This paragraph is automatically incorporated into release notes. Use imperative present tense. Include motivation and contrast with previous behavior when relevant.
+**Body format:** The first paragraph must start with "This PR". This paragraph is automatically incorporated into release notes. Use imperative present tense. Include motivation and contrast with previous behavior when relevant. Do NOT use markdown headings (`## Summary`, `## Test plan`, etc.) in PR bodies.
 
 Example:
 ```
@@ -84,6 +84,27 @@ leading quantifiers are stripped when creating a pattern.
 
 If you're unsure which label applies, it's fine to omit the label and let reviewers add it.
 
+## Module System for `src/` Files
+
+Files in `src/Lean/`, `src/Std/`, and `src/lake/Lake/` must have both `module` and `prelude` (CI enforces `^prelude$` on its own line). With `prelude`, nothing is auto-imported — you must explicitly import `Init.*` modules for standard library features. Check existing files in the same directory for the pattern, e.g.:
+
+```lean
+module
+
+prelude
+import Init.While  -- needed for while/repeat
+import Init.Data.String.TakeDrop  -- needed for String.startsWith
+public import Lean.Compiler.NameMangling  -- public if types are used in public signatures
+```
+
+Files outside these directories (e.g. `tests/`, `script/`) use just `module`.
+
 ## CI Log Retrieval
 
 When CI jobs fail, investigate immediately - don't wait for other jobs to complete. Individual job logs are often available even while other jobs are still running. Try `gh run view <run-id> --log` or `gh run view <run-id> --log-failed`, or use `gh run view <run-id> --job=<job-id>` to target the specific failed job. Sleeping is fine when asked to monitor CI and no failures exist yet, but once any job fails, investigate that failure immediately.
+
+## Copyright Headers
+
+New files require a copyright header. To get the year right, always run `date +%Y` rather than relying on memory. The copyright holder should be the author or their current employer — check other recent files by the same author in the repository to determine the correct entity (e.g., "Lean FRO, LLC", "Amazon.com, Inc. or its affiliates").
+
+Test files (in `tests/`) do not need copyright headers.

.claude/commands/release.md

@@ -103,6 +103,15 @@ Every time you run `release_checklist.py`, you MUST:
 This summary should be provided EVERY time you run the checklist, not just after creating new PRs.
 The user needs to see the complete picture of what's waiting for review.
 
+## Checking PR Status When Asked
+
+When the user asks for "status" or you need to report on PRs between checklist runs:
+- **ALWAYS check actual PR state** using `gh pr view <number> --repo <repo> --json state,mergedAt`
+- Do NOT rely on cached CI results or previous checklist output
+- The user may have merged PRs since your last check
+- Report which PRs are MERGED, which are OPEN with CI status, and which are still pending
+- After discovering merged PRs, rerun `release_checklist.py` to advance the release process
+
 ## Nightly Infrastructure
 
 The nightly build system uses branches and tags across two repositories:
@@ -112,6 +121,42 @@ The nightly build system uses branches and tags across two repositories:
 
 When a nightly succeeds with mathlib, all three should point to the same commit. Don't confuse these: branches are in the main lean4 repo, dated tags are in lean4-nightly.
 
+## CI Failures: Investigate Immediately
+
+**CRITICAL: If the checklist reports `❌ CI: X check(s) failing` for any PR, investigate immediately.**
+
+Do NOT:
+- Report it as "CI in progress" or "some checks pending"
+- Wait for the remaining checks to finish before investigating
+- Assume it's a transient failure without checking
+
+DO:
+1. Run `gh pr checks <number> --repo <owner>/<repo>` to see which specific check failed
+2. Run `gh run view <run-id> --repo <owner>/<repo> --log-failed` to see the failure output
+3. Diagnose the failure and report clearly to the user: what failed and why
+4. Propose a fix if one is obvious (e.g., subverso version mismatch, transient elan install error)
+
+The checklist now distinguishes `❌ X check(s) failing, Y still in progress` from `🔄 Y check(s) in progress`.
+Any `❌` in CI status requires immediate investigation — do not move on.
+
+## Waiting for CI or Merges
+
+Use `gh pr checks --watch` to block until a PR's CI checks complete (no polling needed).
+Run these as background bash commands so you get notified when they finish:
+
+```bash
+# Watch CI, then check merge state
+gh pr checks <number> --repo <owner>/<repo> --watch && gh pr view <number> --repo <owner>/<repo> --json state --jq '.state'
+```
+
+For multiple PRs, launch one background command per PR in parallel. When each completes,
+you'll be notified automatically via a task-notification. Do NOT use sleep-based polling
+loops — `--watch` is event-driven and exits as soon as checks finish.
+
+Note: `gh pr checks --watch` exits as soon as ALL checks complete (pass or fail). If some checks
+fail while others are still running, `--watch` will continue until everything settles, then exit
+with a non-zero code. So a background `--watch` finishing = all checks done; check which failed.
+
 ## Error Handling
 
 **CRITICAL**: If something goes wrong or a command fails:

.claude/settings.json

@@ -0,0 +1,13 @@
+{
+  "extraKnownMarketplaces": {
+    "leanprover": {
+      "source": {
+        "source": "github",
+        "repo": "leanprover/skills"
+      }
+    }
+  },
+  "enabledPlugins": {
+    "lean@leanprover": true
+  }
+}

.claude/skills/profiling/SKILL.md

@@ -0,0 +1,26 @@
+---
+name: profiling
+description: Profile Lean programs with demangled names using samply and Firefox Profiler. Use when the user asks to profile a Lean binary or investigate performance.
+allowed-tools: Bash, Read, Glob, Grep
+---
+
+# Profiling Lean Programs
+
+Full documentation: `script/PROFILER_README.md`.
+
+## Quick Start
+
+```bash
+script/lean_profile.sh ./build/release/stage1/bin/lean some_file.lean
+```
+
+Requires `samply` (`cargo install samply`) and `python3`.
+
+## Agent Notes
+
+- The pipeline is interactive (serves to browser at the end). When running non-interactively, run the steps manually instead of using the wrapper script.
+- The three steps are: `samply record --save-only`, `symbolicate_profile.py`, then `serve_profile.py`.
+- `lean_demangle.py` works standalone as a stdin filter (like `c++filt`) for quick name lookups.
+- The `--raw` flag on `lean_demangle.py` gives exact demangled names without postprocessing (keeps `._redArg`, `._lam_0` suffixes as-is).
+- Use `PROFILE_KEEP=1` to keep the temp directory for later inspection.
+- The demangled profile is a standard Firefox Profiler JSON. Function names live in `threads[i].stringArray`, indexed by `threads[i].funcTable.name`.

.claude/skills/zulip-extract/SKILL.md

@@ -0,0 +1,17 @@
+---
+name: zulip-extract
+description: Extract Zulip thread HTML dumps into readable plain text. Use when the user provides a Zulip HTML file or asks to parse/read/convert/summarize a Zulip thread.
+---
+
+# Zulip Thread Extractor
+
+Run the bundled script to convert a Zulip HTML page dump into plain text.
+
+## Usage
+```bash
+python3 .claude/skills/zulip-extract/zulip_thread_extract.py input.html output.txt
+```
+
+The script has zero dependencies beyond Python 3 stdlib.
+It extracts sender, timestamp, message content (with code blocks,
+links, quotes, mentions), and reactions.

.claude/skills/zulip-extract/zulip_thread_extract.py

@@ -0,0 +1,313 @@
+#!/usr/bin/env python3
+"""
+Convert a Zulip HTML page dump to plain text (the visible message thread).
+
+Zero external dependencies — uses only the Python standard library.
+
+Usage:
+    python3 zulip_thread_extract.py input.html [output.txt]
+"""
+
+import sys
+import re
+from html.parser import HTMLParser
+from html import unescape
+
+
+# ---------------------------------------------------------------------------
+# Minimal DOM built from stdlib HTMLParser
+# ---------------------------------------------------------------------------
+
+class Node:
+    """A lightweight DOM node."""
+    __slots__ = ('tag', 'attrs', 'children', 'parent', 'text')
+
+    def __init__(self, tag='', attrs=None):
+        self.tag = tag
+        self.attrs = dict(attrs) if attrs else {}
+        self.children = []
+        self.parent = None
+        self.text = ''  # for text nodes only (tag == '')
+
+    @property
+    def cls(self):
+        return self.attrs.get('class', '')
+
+    def has_class(self, c):
+        return c in self.cls.split()
+
+    def find_all(self, tag=None, class_=None):
+        """Depth-first search for matching descendants."""
+        for child in self.children:
+            if child.tag == '':
+                continue
+            match = True
+            if tag and child.tag != tag:
+                match = False
+            if class_ and not child.has_class(class_):
+                match = False
+            if match:
+                yield child
+            yield from child.find_all(tag, class_)
+
+    def find(self, tag=None, class_=None):
+        return next(self.find_all(tag, class_), None)
+
+    def get_text(self):
+        if self.tag == '':
+            return self.text
+        return ''.join(c.get_text() for c in self.children)
+
+
+class DOMBuilder(HTMLParser):
+    """Build a minimal DOM tree from HTML."""
+
+    VOID_ELEMENTS = frozenset([
+        'area', 'base', 'br', 'col', 'embed', 'hr', 'img', 'input',
+        'link', 'meta', 'param', 'source', 'track', 'wbr',
+    ])
+
+    def __init__(self):
+        super().__init__()
+        self.root = Node('root')
+        self._cur = self.root
+
+    def handle_starttag(self, tag, attrs):
+        node = Node(tag, attrs)
+        node.parent = self._cur
+        self._cur.children.append(node)
+        if tag not in self.VOID_ELEMENTS:
+            self._cur = node
+
+    def handle_endtag(self, tag):
+        # Walk up to find the matching open tag (tolerates misnesting)
+        n = self._cur
+        while n and n.tag != tag and n.parent:
+            n = n.parent
+        if n and n.parent:
+            self._cur = n.parent
+
+    def handle_data(self, data):
+        t = Node()
+        t.text = data
+        t.parent = self._cur
+        self._cur.children.append(t)
+
+    def handle_entityref(self, name):
+        self.handle_data(unescape(f'&{name};'))
+
+    def handle_charref(self, name):
+        self.handle_data(unescape(f'&#{name};'))
+
+
+def parse_html(path):
+    with open(path, 'r', encoding='utf-8') as f:
+        html = f.read()
+    builder = DOMBuilder()
+    builder.feed(html)
+    return builder.root
+
+
+# ---------------------------------------------------------------------------
+# Content extraction
+# ---------------------------------------------------------------------------
+
+SKIP_CLASSES = {
+    'message_controls', 'message_length_controller',
+    'code-buttons-container', 'copy_codeblock', 'code_external_link',
+    'message_edit_notice', 'edit-notifications',
+}
+
+def should_skip(node):
+    return bool(SKIP_CLASSES & set(node.cls.split()))
+
+
+def extract_content(node):
+    """Recursively convert a message_content node into readable text."""
+    parts = []
+    for child in node.children:
+        # Text node
+        if child.tag == '':
+            parts.append(child.text)
+            continue
+
+        if should_skip(child):
+            continue
+
+        cls_set = set(child.cls.split())
+
+        # Code block wrappers  (div.codehilite / div.zulip-code-block)
+        if child.tag == 'div' and ({'codehilite', 'zulip-code-block'} & cls_set):
+            code = child.find('code')
+            lang = child.attrs.get('data-code-language', '')
+            text = code.get_text() if code else child.get_text()
+            parts.append(f'\n```{lang}\n{text}```\n')
+            continue
+
+        # <pre> (bare code blocks without wrapper div)
+        if child.tag == 'pre':
+            code = child.find('code')
+            text = code.get_text() if code else child.get_text()
+            parts.append(f'\n```\n{text}```\n')
+            continue
+
+        # Inline <code>
+        if child.tag == 'code':
+            parts.append(f'`{child.get_text()}`')
+            continue
+
+        # Paragraph
+        if child.tag == 'p':
+            inner = extract_content(child)
+            parts.append(f'\n{inner}\n')
+            continue
+
+        # Line break
+        if child.tag == 'br':
+            parts.append('\n')
+            continue
+
+        # Links
+        if child.tag == 'a':
+            href = child.attrs.get('href', '')
+            text = child.get_text().strip()
+            if href and not href.startswith('#') and text:
+                parts.append(f'[{text}]({href})')
+            else:
+                parts.append(text)
+            continue
+
+        # Block quotes
+        if child.tag == 'blockquote':
+            bq = extract_content(child).strip()
+            parts.append('\n' + '\n'.join(f'> {l}' for l in bq.split('\n')) + '\n')
+            continue
+
+        # Lists
+        if child.tag in ('ul', 'ol'):
+            for i, li in enumerate(c for c in child.children if c.tag == 'li'):
+                pfx = f'{i+1}.' if child.tag == 'ol' else '-'
+                parts.append(f'\n{pfx} {extract_content(li).strip()}')
+            parts.append('\n')
+            continue
+
+        # User mentions
+        if 'user-mention' in cls_set:
+            parts.append(f'@{child.get_text().strip().lstrip("@")}')
+            continue
+
+        # Emoji
+        if 'emoji' in cls_set:
+            alt = child.attrs.get('alt', '') or child.attrs.get('title', '')
+            if alt:
+                parts.append(alt)
+            continue
+
+        # Recurse into everything else
+        parts.append(extract_content(child))
+
+    return ''.join(parts)
+
+
+# ---------------------------------------------------------------------------
+# Thread extraction
+# ---------------------------------------------------------------------------
+
+def extract_thread(html_path, output_path=None):
+    root = parse_html(html_path)
+
+    # Find the message list
+    msg_list = root.find('div', class_='message-list')
+    if not msg_list:
+        print("ERROR: Could not find message list.", file=sys.stderr)
+        sys.exit(1)
+
+    # Topic header
+    header = msg_list.find('div', class_='message_header')
+    stream_name = topic_name = date_str = ''
+    if header:
+        el = header.find('span', class_='message-header-stream-name')
+        if el: stream_name = el.get_text().strip()
+        el = header.find('span', class_='stream-topic-inner')
+        if el: topic_name = el.get_text().strip()
+        el = header.find('span', class_='recipient_row_date')
+        if el:
+            tr = el.find('span', class_='timerender-content')
+            if tr:
+                date_str = tr.attrs.get('data-tippy-content', '') or tr.get_text().strip()
+
+    # Messages
+    messages = []
+    for row in msg_list.find_all('div', class_='message_row'):
+        if not row.has_class('messagebox-includes-sender'):
+            continue
+
+        msg = {}
+
+        sn = row.find('span', class_='sender_name_text')
+        if sn:
+            un = sn.find('span', class_='user-name')
+            msg['sender'] = (un or sn).get_text().strip()
+
+        tm = row.find('a', class_='message-time')
+        if tm:
+            msg['time'] = tm.get_text().strip()
+
+        cd = row.find('div', class_='message_content')
+        if cd:
+            text = extract_content(cd)
+            text = re.sub(r'\n{3,}', '\n\n', text).strip()
+            msg['content'] = text
+
+        # Reactions
+        reactions = []
+        for rx in row.find_all('div', class_='message_reaction'):
+            em = rx.find('div', class_='emoji_alt_code')
+            if em:
+                reactions.append(em.get_text().strip())
+            else:
+                img = rx.find(tag='img')
+                if img:
+                    reactions.append(img.attrs.get('alt', ''))
+            cnt = rx.find('span', class_='message_reaction_count')
+            if cnt and reactions:
+                c = cnt.get_text().strip()
+                if c and c != '1':
+                    reactions[-1] += f' x{c}'
+        if reactions:
+            msg['reactions'] = reactions
+
+        if msg.get('content') or msg.get('sender'):
+            messages.append(msg)
+
+    # Format
+    lines = [
+        '=' * 70,
+        f'# {stream_name} > {topic_name}',
+    ]
+    if date_str:
+        lines.append(f'# Started: {date_str}')
+    lines += [f'# Messages: {len(messages)}', '=' * 70, '']
+
+    for msg in messages:
+        lines.append(f'--- {msg.get("sender","?")}  [{msg.get("time","")}] ---')
+        lines.append(msg.get('content', ''))
+        if msg.get('reactions'):
+            lines.append(f'  Reactions: {", ".join(msg["reactions"])}')
+        lines.append('')
+
+    result = '\n'.join(lines)
+    if output_path:
+        with open(output_path, 'w', encoding='utf-8') as f:
+            f.write(result)
+        print(f"Written {len(messages)} messages to {output_path}")
+    else:
+        print(result)
+
+
+if __name__ == '__main__':
+    if len(sys.argv) < 2:
+        print(f"Usage: {sys.argv[0]} input.html [output.txt]")
+        sys.exit(1)
+    extract_thread(sys.argv[1], sys.argv[2] if len(sys.argv) > 2 else None)
+

.gitattributes

@@ -5,9 +5,3 @@ stage0/** binary linguist-generated
 # The following file is often manually edited, so do show it in diffs
 stage0/src/stdlib_flags.h -binary -linguist-generated
 doc/std/grove/GroveStdlib/Generated/** linguist-generated
-# These files should not have line endings translated on Windows, because
-# it throws off parser tests. Later lines override earlier ones, so the
-# runner code is still treated as ordinary text.
-tests/lean/docparse/* eol=lf
-tests/lean/docparse/*.lean eol=auto
-tests/lean/docparse/*.sh eol=auto

.github/workflows/awaiting-manual.yml

@@ -2,16 +2,19 @@ name: Check awaiting-manual label
 
 on:
   merge_group:
-  pull_request:
+  pull_request_target:
     types: [opened, synchronize, reopened, labeled, unlabeled]
 
+permissions:
+  pull-requests: read
+
 jobs:
   check-awaiting-manual:
     runs-on: ubuntu-latest
     steps:
       - name: Check awaiting-manual label
         id: check-awaiting-manual-label
-        if: github.event_name == 'pull_request'
+        if: github.event_name == 'pull_request_target'
         uses: actions/github-script@v8
         with:
           script: |
@@ -28,7 +31,7 @@ jobs:
             }
       
       - name: Wait for manual compatibility
-        if: github.event_name == 'pull_request' && steps.check-awaiting-manual-label.outputs.awaiting == 'true'
+        if: github.event_name == 'pull_request_target' && steps.check-awaiting-manual-label.outputs.awaiting == 'true'
         run: |
           echo "::notice title=Awaiting manual::PR is marked 'awaiting-manual' but neither 'breaks-manual' nor 'builds-manual' labels are present."
           echo "This check will remain in progress until the PR is updated with appropriate manual compatibility labels."

.github/workflows/awaiting-mathlib.yml

@@ -2,16 +2,19 @@ name: Check awaiting-mathlib label
 
 on:
   merge_group:
-  pull_request:
+  pull_request_target:
     types: [opened, synchronize, reopened, labeled, unlabeled]
 
+permissions:
+  pull-requests: read
+
 jobs:
   check-awaiting-mathlib:
     runs-on: ubuntu-latest
     steps:
       - name: Check awaiting-mathlib label
         id: check-awaiting-mathlib-label
-        if: github.event_name == 'pull_request'
+        if: github.event_name == 'pull_request_target'
         uses: actions/github-script@v8
         with:
           script: |
@@ -28,7 +31,7 @@ jobs:
             }
       
       - name: Wait for mathlib compatibility
-        if: github.event_name == 'pull_request' && steps.check-awaiting-mathlib-label.outputs.awaiting == 'true'
+        if: github.event_name == 'pull_request_target' && steps.check-awaiting-mathlib-label.outputs.awaiting == 'true'
         run: |
           echo "::notice title=Awaiting mathlib::PR is marked 'awaiting-mathlib' but neither 'breaks-mathlib' nor 'builds-mathlib' labels are present."
           echo "This check will remain in progress until the PR is updated with appropriate mathlib compatibility labels."

.github/workflows/build-template.yml

@@ -49,7 +49,7 @@ jobs:
       LSAN_OPTIONS: max_leaks=10
       # somehow MinGW clang64 (or cmake?) defaults to `g++` even though it doesn't exist
       CXX: c++
-      MACOSX_DEPLOYMENT_TARGET: 10.15
+      MACOSX_DEPLOYMENT_TARGET: 11.0
     steps:
       - name: Install Nix
         uses: DeterminateSystems/nix-installer-action@main
@@ -66,16 +66,10 @@ jobs:
           brew install ccache tree zstd coreutils gmp libuv
         if: runner.os == 'macOS'
       - name: Checkout
-        if: (!endsWith(matrix.os, '-with-cache'))
         uses: actions/checkout@v6
         with:
           # the default is to use a virtual merge commit between the PR and master: just use the PR
           ref: ${{ github.event.pull_request.head.sha }}
-      - name: Namespace Checkout
-        if: endsWith(matrix.os, '-with-cache')
-        uses: namespacelabs/nscloud-checkout-action@v8
-        with:
-          ref: ${{ github.event.pull_request.head.sha }}
       - name: Open Nix shell once
         run: true
         if: runner.os == 'Linux'
@@ -85,7 +79,7 @@ jobs:
       - name: CI Merge Checkout
         run: |
           git fetch --depth=1 origin ${{ github.sha }}
-          git checkout FETCH_HEAD flake.nix flake.lock script/prepare-* tests/lean/run/importStructure.lean
+          git checkout FETCH_HEAD flake.nix flake.lock script/prepare-* tests/elab/importStructure.lean
         if: github.event_name == 'pull_request'
       # (needs to be after "Checkout" so files don't get overridden)
       - name: Setup emsdk
@@ -235,25 +229,21 @@ jobs:
         # prefix `if` above with `always` so it's run even if tests failed
         if: always() && steps.test.conclusion != 'skipped'
       - name: Check Test Binary
-        run: ${{ matrix.binary-check }} tests/compiler/534.lean.out
+        run: ${{ matrix.binary-check }} tests/compile/534.lean.out
         if: (!matrix.cross) && steps.test.conclusion != 'skipped'
       - name: Build Stage 2
         run: |
           make -C build -j$NPROC stage2
-        if: matrix.test-speedcenter
+        if: matrix.test-bench
       - name: Check Stage 3
         run: |
           make -C build -j$NPROC check-stage3
         if: matrix.check-stage3
-      - name: Test Speedcenter Benchmarks
+      - name: Test Benchmarks
         run: |
-          # Necessary for some timing metrics but does not work on Namespace runners
-          # and we just want to test that the benchmarks run at all here
-          #echo -1 | sudo tee /proc/sys/kernel/perf_event_paranoid
-          export BUILD=$PWD/build PATH=$PWD/build/stage1/bin:$PATH
-          cd tests/bench
-          nix shell .#temci -c temci exec --config speedcenter.yaml --included_blocks fast --runs 1
-        if: matrix.test-speedcenter
+          cd tests
+          nix develop -c make -C ../build -j$NPROC bench
+        if: matrix.test-bench
       - name: Check rebootstrap
         run: |
           set -e

.github/workflows/check-stdlib-flags.yml

@@ -1,9 +1,12 @@
 name: Check stdlib_flags.h modifications
 
 on:
-  pull_request:
+  pull_request_target:
     types: [opened, synchronize, reopened, labeled, unlabeled]
 
+permissions:
+  pull-requests: read
+
 jobs:
   check-stdlib-flags:
     runs-on: ubuntu-latest

.github/workflows/ci.yml

@@ -258,8 +258,8 @@ jobs:
                 "check-rebootstrap": level >= 1,
                 "check-stage3": level >= 2,
                 "test": true,
-                // NOTE: `test-speedcenter` currently seems to be broken on `ubuntu-latest`
-                "test-speedcenter": large && level >= 2,
+                // NOTE: `test-bench` currently seems to be broken on `ubuntu-latest`
+                "test-bench": large && level >= 2,
                 // We are not warning-free yet on all platforms, start here
                 "CMAKE_OPTIONS": "-DLEAN_EXTRA_CXX_FLAGS=-Werror",
               },
@@ -269,6 +269,8 @@ jobs:
                 "enabled": level >= 2,
                 "test": true,
                 "CMAKE_PRESET": "reldebug",
+                // * `elab_bench/big_do` crashes with exit code 134
+                "CTEST_OPTIONS": "-E 'elab_bench/big_do'",
               },
               {
                 "name": "Linux fsanitize",

.github/workflows/pr-body.yml

@@ -2,17 +2,23 @@ name: Check PR body for changelog convention
 
 on:
   merge_group:
-  pull_request:
+  pull_request_target:
     types: [opened, synchronize, reopened, edited, labeled, converted_to_draft, ready_for_review]
 
+permissions:
+  pull-requests: read
+
 jobs:
   check-pr-body:
     runs-on: ubuntu-latest
     steps:
       - name: Check PR body
-        if: github.event_name == 'pull_request'
+        if: github.event_name == 'pull_request_target'
         uses: actions/github-script@v8
         with:
+          # Safety note: this uses pull_request_target, so the workflow has elevated privileges.
+          # The PR title and body are only used in regex tests (read-only string matching),
+          # never interpolated into shell commands, eval'd, or written to GITHUB_ENV/GITHUB_OUTPUT.
           script: |
             const { title, body, labels, draft } = context.payload.pull_request;
             if (!draft && /^(feat|fix):/.test(title) && !labels.some(label => label.name == "changelog-no")) {

.gitignore

@@ -1,7 +1,6 @@
 *~
 \#*
 .#*
-*.lock
 .lake
 lake-manifest.json
 /build
@@ -18,8 +17,12 @@ compile_commands.json
 *.idea
 tasks.json
 settings.json
+!.claude/settings.json
 .gdb_history
 .vscode/*
+!.vscode/settings.json
+!.vscode/tasks.json
+!.vscode/extensions.json
 script/__pycache__
 *.produced.out
 CMakeSettings.json

.vscode/extensions.json

@@ -0,0 +1,5 @@
+{
+	"recommendations": [
+		"leanprover.lean4"
+	]
+}

.vscode/settings.json

@@ -0,0 +1,12 @@
+{
+	"files.insertFinalNewline": true,
+	"files.trimTrailingWhitespace": true,
+	// These require the CMake Tools extension (ms-vscode.cmake-tools).
+	"cmake.buildDirectory": "${workspaceFolder}/build/release",
+	"cmake.generator": "Unix Makefiles",
+	"[lean4]": {
+		"editor.rulers": [
+			100
+		]
+	}
+}

.vscode/tasks.json

@@ -0,0 +1,34 @@
+{
+	"version": "2.0.0",
+	"tasks": [
+		{
+			"label": "build",
+			"type": "shell",
+			"command": "make -C build/release -j$(nproc 2>/dev/null || sysctl -n hw.logicalcpu 2>/dev/null || echo 4)",
+			"problemMatcher": [],
+			"group": {
+				"kind": "build",
+				"isDefault": true
+			}
+		},
+		{
+			"label": "build-old",
+			"type": "shell",
+			"command": "make -C build/release -j$(nproc 2>/dev/null || sysctl -n hw.logicalcpu 2>/dev/null || echo 4) LAKE_EXTRA_ARGS=--old",
+			"problemMatcher": [],
+			"group": {
+				"kind": "build"
+			}
+		},
+		{
+			"label": "test",
+			"type": "shell",
+			"command": "NPROC=$(nproc 2>/dev/null || sysctl -n hw.logicalcpu 2>/dev/null || echo 4); CTEST_OUTPUT_ON_FAILURE=1 make -C build/release test -j$NPROC ARGS=\"-j$NPROC\"",
+			"problemMatcher": [],
+			"group": {
+				"kind": "test",
+				"isDefault": true
+			}
+		}
+	]
+}

CMakeLists.txt

@@ -1,4 +1,8 @@
-cmake_minimum_required(VERSION 3.11)
+cmake_minimum_required(VERSION 3.21)
+
+if(NOT CMAKE_GENERATOR MATCHES "Makefiles")
+  message(FATAL_ERROR "Only makefile generators are supported")
+endif()
 
 option(USE_MIMALLOC "use mimalloc" ON)
 
@@ -70,13 +74,7 @@ if(NOT CMAKE_SYSTEM_NAME MATCHES "Emscripten")
       BUILD_IN_SOURCE ON
       INSTALL_COMMAND ""
     )
-    set(
-      CADICAL
-      ${CMAKE_BINARY_DIR}/cadical/cadical${CMAKE_EXECUTABLE_SUFFIX}
-      CACHE FILEPATH
-      "path to cadical binary"
-      FORCE
-    )
+    set(CADICAL ${CMAKE_BINARY_DIR}/cadical/cadical${CMAKE_EXECUTABLE_SUFFIX})
     list(APPEND EXTRA_DEPENDS cadical)
   endif()
   list(APPEND CL_ARGS -DCADICAL=${CADICAL})
@@ -153,6 +151,7 @@ ExternalProject_Add(
   INSTALL_COMMAND ""
   DEPENDS stage2
   EXCLUDE_FROM_ALL ON
+  STEP_TARGETS configure
 )
 
 # targets forwarded to appropriate stages
@@ -163,6 +162,25 @@ add_custom_target(update-stage0-commit COMMAND $(MAKE) -C stage1 update-stage0-c
 
 add_custom_target(test COMMAND $(MAKE) -C stage1 test DEPENDS stage1)
 
+add_custom_target(
+  bench
+  COMMAND $(MAKE) -C stage2
+  COMMAND $(MAKE) -C stage2 -j1 bench
+  DEPENDS stage2
+)
+add_custom_target(
+  bench-part1
+  COMMAND $(MAKE) -C stage2
+  COMMAND $(MAKE) -C stage2 -j1 bench-part1
+  DEPENDS stage2
+)
+add_custom_target(
+  bench-part2
+  COMMAND $(MAKE) -C stage2
+  COMMAND $(MAKE) -C stage2 -j1 bench-part2
+  DEPENDS stage2
+)
+
 add_custom_target(clean-stdlib COMMAND $(MAKE) -C stage1 clean-stdlib DEPENDS stage1)
 
 install(CODE "execute_process(COMMAND make -C stage1 install)")

CMakePresets.json

@@ -41,7 +41,7 @@
         "SMALL_ALLOCATOR": "OFF",
         "USE_MIMALLOC": "OFF",
         "BSYMBOLIC": "OFF",
-        "LEAN_TEST_VARS": "MAIN_STACK_SIZE=16000 LSAN_OPTIONS=max_leaks=10"
+        "LEAN_TEST_VARS": "MAIN_STACK_SIZE=16000 TEST_STACK_SIZE=16000 LSAN_OPTIONS=max_leaks=10"
       },
       "generator": "Unix Makefiles",
       "binaryDir": "${sourceDir}/build/sanitize"

doc/.gitignore

@@ -1 +0,0 @@
-out

doc/dev/index.md

@@ -61,10 +61,10 @@ you can then put `my_name/lean4:my-tag` in your `lean-toolchain` file in a proje
 
 ### VS Code
 
-There is a `lean.code-workspace` file that correctly sets up VS Code with workspace roots for the stage0/stage1 setup described above as well as with other settings.
-You should always load it when working on Lean, such as by invoking
+There is a `.vscode/` directory that correctly sets up VS Code with settings, tasks, and recommended extensions.
+Simply open the repository folder in VS Code, such as by invoking
 ```
-code lean.code-workspace
+code .
 ```
 on the command line.
 

doc/dev/release_checklist.md

@@ -65,7 +65,14 @@ We'll use `v4.6.0` as the intended release version as a running example.
       - The `lakefile.toml` should always refer to dependencies via their `main` or `master` branch,
         not a toolchain tag
         (with the exception of `ProofWidgets4`, which *must* use a sequential version tag).
+      - **Important:** After creating and pushing the ProofWidgets4 tag (see above),
+        the mathlib4 lakefile must be updated to reference the new tag (e.g. `v0.0.87`).
+        The `release_steps.py` script handles this automatically by looking up the latest
+        ProofWidgets4 tag compatible with the target toolchain.
       - Push the PR branch to the main Mathlib repository rather than a fork, or CI may not work reliably
+      - The "Verify Transient and Automated Commits" CI check on toolchain bump PRs can be ignored —
+        it often fails on automated commits (`x:` prefixed) from the nightly-testing history that can't be
+        reproduced in CI. This does not block merging.
     - `repl`:
       There are two copies of `lean-toolchain`/`lakefile.lean`:
       in the root, and in `test/Mathlib/`. Edit both, and run `lake update` in both directories.
@@ -146,6 +153,9 @@ We'll use `v4.7.0-rc1` as the intended release version in this example.
     * The repository does not need any changes to move to the new version.
     * Note that sometimes there are *unreviewed* but necessary changes on the `nightly-testing` branch of the repository.
       If so, you will need to merge these into the `bump_to_v4.7.0-rc1` branch manually.
+    * The `nightly-testing` branch may also contain temporary fix scripts (e.g. `fix_backward_defeq.py`,
+      `fix_deprecations.py`) that were used to adapt to breaking changes during the nightly cycle.
+      These should be reviewed and removed if no longer needed, as they can interfere with CI checks.
   - For each of the repositories listed in `script/release_repos.yml`,
     - Run `script/release_steps.py v4.7.0-rc1 <repo>` (e.g. replacing `<repo>` with `batteries`), which will walk you through the following steps:
       - Create a new branch off `master`/`main` (as specified in the `branch` field), called `bump_to_v4.7.0-rc1`.

doc/dev/testing.md

@@ -1,5 +1,9 @@
 # Test Suite
 
+**Warning:** This document is partially outdated.
+It describes the old test suite, which is currently in the process of being replaced.
+The new test suite's documentation can be found at [`tests/README.md`](../../tests/README.md).
+
 After [building Lean](../make/index.md) you can run all the tests using
 ```
 cd build/release

doc/examples/.gitignore

@@ -0,0 +1,2 @@
+*.out.produced
+*.exit.produced

doc/examples/bintree.lean.out.expected

@@ -0,0 +1,2 @@
+Tree.node (Tree.node (Tree.leaf) 1 "one" (Tree.leaf)) 2 "two" (Tree.node (Tree.leaf) 3 "three" (Tree.leaf))
+[(1, "one"), (2, "two"), (3, "three")]

doc/examples/compiler/run_test

@@ -0,0 +1,11 @@
+#!/usr/bin/env bash
+source ../../../tests/env_test.sh
+source "$TEST_DIR/util.sh"
+
+leanmake --always-make bin
+
+exec_capture test.lean \
+  ./build/bin/test hello world
+
+check_exit test.lean
+check_out test.lean

doc/examples/compiler/test.lean.out.expected

@@ -0,0 +1 @@
+[hello, world]

doc/examples/interp.lean.out.expected

@@ -0,0 +1,3 @@
+30
+interp.lean:146:4: warning: declaration uses `sorry`
+3628800

doc/examples/palindromes.lean.out.expected

@@ -0,0 +1,2 @@
+true
+false

doc/examples/phoas.lean.out.expected

@@ -0,0 +1,2 @@
+"(((fun x_1 => (fun x_2 => (x_1 + x_2))) 1) 2)"
+"((((fun x_1 => (fun x_2 => (x_1 + x_2))) 1) 2) + 5)"

doc/examples/run_test

@@ -0,0 +1,9 @@
+#!/usr/bin/env bash
+source ../../tests/env_test.sh
+source "$TEST_DIR/util.sh"
+
+exec_capture "$1" \
+  lean -Dlinter.all=false "$1"
+
+check_exit "$1"
+check_out "$1"

doc/examples/test_single.sh

@@ -1,4 +0,0 @@
-#!/usr/bin/env bash
-source ../../tests/common.sh
-
-exec_check_raw lean -Dlinter.all=false "$f"

doc/std/grove/lean-toolchain

@@ -1 +1 @@
-lean4
+../../../build/release/stage1

lean-toolchain

@@ -1 +1 @@
-lean4
+build/release/stage1

lean.code-workspace

@@ -1,72 +0,0 @@
-{
-	"folders": [
-		{
-			"path": "."
-		},
-		{
-			"path": "src"
-		},
-		{
-			"path": "tests"
-		},
-		{
-			"path": "script"
-		}
-	],
-	"settings": {
-		// Open terminal at root, not current workspace folder
-		// (there is not way to directly refer to the root folder included as `.` above)
-		"terminal.integrated.cwd": "${workspaceFolder:src}/..",
-		"files.insertFinalNewline": true,
-		"files.trimTrailingWhitespace": true,
-		"cmake.buildDirectory": "${workspaceFolder}/build/release",
-		"cmake.generator": "Unix Makefiles",
-		"[markdown]": {
-			"rewrap.wrappingColumn": 70
-		},
-		"[lean4]": {
-			"editor.rulers": [
-				100
-			]
-		}
-	},
-	"tasks": {
-		"version": "2.0.0",
-		"tasks": [
-			{
-				"label": "build",
-				"type": "shell",
-				"command": "make -C build/release -j$(nproc 2>/dev/null || sysctl -n hw.logicalcpu 2>/dev/null || echo 4)",
-				"problemMatcher": [],
-				"group": {
-					"kind": "build",
-					"isDefault": true
-				}
-			},
-			{
-				"label": "build-old",
-				"type": "shell",
-				"command": "make -C build/release -j$(nproc 2>/dev/null || sysctl -n hw.logicalcpu 2>/dev/null || echo 4) LAKE_EXTRA_ARGS=--old",
-				"problemMatcher": [],
-				"group": {
-					"kind": "build"
-				}
-			},
-			{
-				"label": "test",
-				"type": "shell",
-				"command": "NPROC=$(nproc 2>/dev/null || sysctl -n hw.logicalcpu 2>/dev/null || echo 4); CTEST_OUTPUT_ON_FAILURE=1 make -C build/release test -j$NPROC ARGS=\"-j$NPROC\"",
-				"problemMatcher": [],
-				"group": {
-					"kind": "test",
-					"isDefault": true
-				}
-			}
-		]
-	},
-	"extensions": {
-		"recommendations": [
-			"leanprover.lean4"
-		]
-	}
-}

releases_drafts/environment.md

@@ -1,6 +0,0 @@
-**Breaking Changes**
-
-* The functions `Lean.Environment.importModules` and `Lean.Environment.finalizeImport` have been extended with a new parameter `loadExts : Bool := false` that enables environment extension state loading.
-  Their previous behavior corresponds to setting the flag to `true` but is only safe to do in combination with `enableInitializersExecution`; see also the `importModules` docstring.
-  The new default value `false` ensures the functions can be used correctly multiple times within the same process when environment extension access is not needed.
-  The wrapper function `Lean.Environment.withImportModules` now always calls `importModules` with `loadExts := false` as it is incompatible with extension loading.

releases_drafts/module-system.md

@@ -1,54 +0,0 @@
-This release introduces the Lean module system, which allows files to
-control the visibility of their contents for other files. In previous
-releases, this feature was available as a preview when the option
-`experimental.module` was set to `true`; it is now a fully supported
-feature of Lean.
-
-# Benefits
-
-Because modules reduce the amount of information exposed to other
-code, they speed up rebuilds because irrelevant changes can be
-ignored, they make it possible to be deliberate about API evolution by
-hiding details that may change from clients, they help proofs be
-checked faster by avoiding accidentally unfolding definitions, and
-they lead to smaller executable files through improved dead code
-elimination.
-
-# Visibility
-
-A source file is a module if it begins with the `module` keyword.  By
-default, declarations in a module are private; the `public` modifier
-exports them. Proofs of theorems and bodies of definitions are private
-by default even when their signatures are public; the bodies of
-definitions can be made public by adding the `@[expose]`
-attribute. Theorems and opaque constants never expose their bodies.
-
-`public section` and `@[expose] section` change the default visibility
-of declarations in the section.
-
-# Imports
-
-Modules may only import other modules. By default, `import` adds the
-public information of the imported module to the private scope of the
-current module. Adding the `public` modifier to an import places the
-imported modules's public information in the public scope of the
-current module, exposing it in turn to the current module's clients.
-
-Within a package, `import all` can be used to import another module's
-private scope into the current module; this can be used to separate
-lemmas or tests from definition modules without exposing details to
-downstream clients.
-
-# Meta Code
-
-Code used in metaprograms must be marked `meta`. This ensures that the
-code is compiled and available for execution when it is needed during
-elaboration. Meta code may only reference other meta code. A whole
-module can be made available in the meta phase using `meta import`;
-this allows code to be shared across phases by importing the module in
-each phase. Code that is reachable from public metaprograms must be
-imported via `public meta import`, while local metaprograms can use
-plain `meta import` for their dependencies.
-
-
-The module system is described in detail in [the Lean language reference](https://lean-reference-manual-review.netlify.app/find/?domain=Verso.Genre.Manual.section&name=files).

script/PROFILER_README.md

@@ -0,0 +1,178 @@
+# Lean Profiler
+
+Profile Lean programs with demangled names using
+[samply](https://github.com/mstange/samply) and
+[Firefox Profiler](https://profiler.firefox.com).
+
+Python 3, no external dependencies.
+
+## Quick start
+
+```bash
+# One command: record, symbolicate, demangle, and open in Firefox Profiler
+script/lean_profile.sh ./my_lean_binary [args...]
+
+# See all options
+script/lean_profile.sh --help
+```
+
+Requirements: `samply` (`cargo install samply`), `python3`.
+
+## Reading demangled names
+
+The demangler transforms low-level C symbol names into readable Lean names
+and annotates them with compact modifiers.
+
+### Basic names
+
+| Raw symbol | Demangled |
+|---|---|
+| `l_Lean_Meta_Sym_main` | `Lean.Meta.Sym.main` |
+| `lp_std_List_map` | `List.map (std)` |
+| `_init_l_Foo_bar` | `[init] Foo.bar` |
+| `initialize_Init_Data` | `[module_init] Init.Data` |
+| `_lean_main` | `[lean] main` |
+
+### Modifier flags `[...]`
+
+Compiler-generated suffixes are folded into a bracket annotation after the
+name. These indicate *how* the function was derived from the original source
+definition.
+
+| Flag | Meaning | Compiler suffix |
+|---|---|---|
+| `arity`↓ | Reduced-arity specialization | `_redArg` |
+| `boxed` | Boxed calling-convention wrapper | `_boxed` |
+| `impl` | Implementation detail | `_impl` |
+| λ | Lambda-lifted closure | `_lam_N`, `_lambda_N`, `_elam_N` |
+| `jp` | Join point | `_jp_N` |
+| `closed` | Extracted closed subterm | `_closed_N` |
+| `private` | Private (module-scoped) definition | `_private.Module.0.` prefix |
+
+Examples:
+
+```
+Lean.Meta.Simp.simpLambda [boxed, λ]     -- boxed wrapper of a lambda-lifted closure
+Lean.Meta.foo [arity↓, private]           -- reduced-arity version of a private def
+```
+
+Multiple flags are comma-separated. Order reflects how they were collected
+(innermost suffix first).
+
+### Specializations `spec at ...`
+
+When the compiler specializes a function at a particular call site, the
+demangled name shows `spec at <context>` after the base name and its flags.
+The context names the function whose body triggered the specialization, and
+may carry its own modifier flags:
+
+```
+<base-name> [<base-flags>] spec at <context>[<context-flags>]
+```
+
+Examples:
+
+```
+-- foo specialized at call site in bar
+Lean.Meta.foo spec at Lean.Meta.bar
+
+-- foo (with a lambda closure) specialized at bar (with reduced arity and a lambda)
+Lean.Meta.foo [λ] spec at Lean.Meta.bar[λ, arity↓]
+
+-- chained specialization: foo specialized at bar, then at baz
+Lean.Meta.foo spec at Lean.Meta.bar spec at Lean.Meta.baz[arity↓]
+```
+
+Context flags use the same symbols as base flags. When a context has no
+flags, the brackets are omitted.
+
+### Other annotations
+
+| Pattern | Meaning |
+|---|---|
+| `<apply/N>` | Lean runtime apply function (N arguments) |
+| `.cold.N` suffix | LLVM cold-path clone (infrequently executed) |
+| `(pkg)` suffix | Function from package `pkg` |
+
+## Tools
+
+### `script/lean_profile.sh` -- Full profiling pipeline
+
+Records a profile, symbolicates it via samply's API, demangles Lean names,
+and opens the result in Firefox Profiler. This is the recommended workflow.
+
+```bash
+script/lean_profile.sh ./build/release/stage1/bin/lean src/Lean/Elab/Term.lean
+```
+
+Environment variables:
+
+| Variable | Default | Description |
+|---|---|---|
+| `SAMPLY_RATE` | 1000 | Sampling rate in Hz |
+| `SAMPLY_PORT` | 3756 | Port for samply symbolication server |
+| `SERVE_PORT` | 3757 | Port for serving the demangled profile |
+| `PROFILE_KEEP` | 0 | Set to 1 to keep the temp directory |
+
+### `script/profiler/lean_demangle.py` -- Name demangler
+
+Demangles individual symbol names. Works as a stdin filter (like `c++filt`)
+or with arguments.
+
+```bash
+echo "l_Lean_Meta_Sym_main" | python3 script/profiler/lean_demangle.py
+# Lean.Meta.Sym.main
+
+python3 script/profiler/lean_demangle.py --raw l_foo___redArg
+# foo._redArg  (exact name, no postprocessing)
+```
+
+As a Python module:
+
+```python
+from lean_demangle import demangle_lean_name, demangle_lean_name_raw
+
+demangle_lean_name("l_foo___redArg")       # "foo [arity↓]"
+demangle_lean_name_raw("l_foo___redArg")   # "foo._redArg"
+```
+
+### `script/profiler/symbolicate_profile.py` -- Profile symbolicator
+
+Calls samply's symbolication API to resolve raw addresses into symbol names,
+then demangles them. Used internally by `lean_profile.sh`.
+
+### `script/profiler/serve_profile.py` -- Profile server
+
+Serves a profile JSON file to Firefox Profiler without re-symbolication
+(which would overwrite demangled names). Used internally by `lean_profile.sh`.
+
+### `script/profiler/lean_demangle_profile.py` -- Standalone profile rewriter
+
+Demangles names in an already-symbolicated profile file (if you have one
+from another source).
+
+```bash
+python3 script/profiler/lean_demangle_profile.py profile.json.gz -o demangled.json.gz
+```
+
+## Tests
+
+```bash
+cd script/profiler && python3 -m unittest test_demangle -v
+```
+
+## How it works
+
+The demangler is a faithful port of Lean 4's `Name.demangleAux` from
+`src/Lean/Compiler/NameMangling.lean`. It reverses the encoding used by
+`Name.mangle` / `Name.mangleAux` which turns hierarchical Lean names into
+valid C identifiers:
+
+- `_` separates name components (`Lean.Meta` -> `Lean_Meta`)
+- `__` encodes a literal underscore in a component name
+- `_xHH`, `_uHHHH`, `_UHHHHHHHH` encode special characters
+- `_N_` encodes numeric name components
+- `_00` is a disambiguation prefix for ambiguous patterns
+
+After demangling, a postprocessing pass folds compiler-generated suffixes
+into human-readable annotations (see [Reading demangled names](#reading-demangled-names)).

script/benchReelabRss.lean

@@ -83,7 +83,7 @@ def main (args : List String) : IO Unit := do
       lastRSS? := some rss
 
     let avgRSSDelta := totalRSSDelta / (n - 2)
-    IO.println s!"avg-reelab-rss-delta: {avgRSSDelta}"
+    IO.println s!"measurement: avg-reelab-rss-delta {avgRSSDelta*1024} b"
 
     let _ ← Ipc.collectDiagnostics requestNo uri versionNo
     (← Ipc.stdin).writeLspMessage (Message.notification "exit" none)

script/benchReelabWatchdogRss.lean

@@ -82,7 +82,7 @@ def main (args : List String) : IO Unit := do
       lastRSS? := some rss
 
     let avgRSSDelta := totalRSSDelta / (n - 2)
-    IO.println s!"avg-reelab-rss-delta: {avgRSSDelta}"
+    IO.println s!"measurement: avg-reelab-rss-delta {avgRSSDelta*1024} b"
 
     let _ ← Ipc.collectDiagnostics requestNo uri versionNo
     Ipc.shutdown requestNo

script/fmt

@@ -9,5 +9,5 @@ find -regex '.*/CMakeLists\.txt\(\.in\)?\|.*\.cmake\(\.in\)?' \
   ! -path "./stage0/*" \
   -exec \
     uvx gersemi --in-place --line-length 120 --indent 2 \
-    --definitions src/cmake/Modules/ src/CMakeLists.txt \
+    --definitions src/cmake/Modules/ src/CMakeLists.txt tests/CMakeLists.txt \
     -- {} +

script/gen_constants_cpp.py

@@ -1,4 +1,4 @@
-#!/usr/bin/env python
+#!/usr/bin/env python3
 # -*- coding: utf-8 -*-
 #
 # Copyright (c) 2015 Microsoft Corporation. All rights reserved.

script/gen_tokens_cpp.py

@@ -1,4 +1,4 @@
-#!/usr/bin/env python
+#!/usr/bin/env python3
 # -*- coding: utf-8 -*-
 #
 # Copyright (c) 2015 Microsoft Corporation. All rights reserved.

script/lean-toolchain

@@ -1 +1 @@
-lean4
+../build/release/stage1

script/lean_profile.sh

@@ -0,0 +1,133 @@
+#!/usr/bin/env bash
+# Profile a Lean binary with demangled names.
+#
+# Usage:
+#   script/lean_profile.sh ./my_lean_binary [args...]
+#
+# Records a profile with samply, symbolicates via samply's API,
+# demangles Lean symbol names, and opens the result in Firefox Profiler.
+#
+# Requirements: samply (cargo install samply), python3
+#
+# Options (via environment variables):
+#   SAMPLY_RATE    — sampling rate in Hz (default: 1000)
+#   SAMPLY_PORT    — port for samply symbolication server (default: 3756)
+#   SERVE_PORT     — port for serving the demangled profile (default: 3757)
+#   PROFILE_KEEP   — set to 1 to keep the raw profile after demangling
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+PROFILER_DIR="$SCRIPT_DIR/profiler"
+SYMBOLICATE="$PROFILER_DIR/symbolicate_profile.py"
+SERVE_PROFILE="$PROFILER_DIR/serve_profile.py"
+
+usage() {
+    cat >&2 <<EOF
+Usage: $0 [options] <lean-binary> [args...]
+
+Profile a Lean binary and view the results in Firefox Profiler
+with demangled Lean names.
+
+Requirements:
+  samply    cargo install samply
+  python3   (included with macOS / most Linux distros)
+
+Environment variables:
+  SAMPLY_RATE    sampling rate in Hz (default: 1000)
+  SAMPLY_PORT    port for samply symbolication server (default: 3756)
+  SERVE_PORT     port for serving the demangled profile (default: 3757)
+  PROFILE_KEEP   set to 1 to keep the temp directory after profiling
+
+Reading demangled names:
+  Compiler suffixes are shown as modifier flags after the name:
+    [arity↓]   reduced-arity specialization     (_redArg)
+    [boxed]    boxed calling-convention wrapper  (_boxed)
+    [λ]        lambda-lifted closure             (_lam_N, _lambda_N, _elam_N)
+    [jp]       join point                        (_jp_N)
+    [closed]   extracted closed subterm          (_closed_N)
+    [private]  private (module-scoped) def       (_private.Module.0. prefix)
+    [impl]     implementation detail             (_impl)
+
+  Specializations appear after the flags:
+    Lean.Meta.foo [λ] spec at Lean.Meta.bar[λ, arity↓]
+      = foo (with lambda closure), specialized at bar (lambda, reduced arity)
+
+  Multiple "spec at" entries indicate chained specializations.
+  See script/PROFILER_README.md for full documentation.
+EOF
+    exit "${1:-0}"
+}
+
+if [ $# -eq 0 ]; then
+    usage 1
+fi
+
+case "${1:-}" in
+    -h|--help) usage 0 ;;
+esac
+
+if ! command -v samply &>/dev/null; then
+    echo "error: samply not found. Install with: cargo install samply" >&2
+    exit 1
+fi
+
+RATE="${SAMPLY_RATE:-1000}"
+PORT="${SAMPLY_PORT:-3756}"
+SERVE="${SERVE_PORT:-3757}"
+TMPDIR=$(mktemp -d /tmp/lean-profile-XXXXXX)
+TMPFILE="$TMPDIR/profile.json.gz"
+DEMANGLED="$TMPDIR/profile-demangled.json.gz"
+SAMPLY_LOG="$TMPDIR/samply.log"
+SAMPLY_PID=""
+
+cleanup() {
+    if [ -n "$SAMPLY_PID" ]; then
+        kill "$SAMPLY_PID" 2>/dev/null || true
+        wait "$SAMPLY_PID" 2>/dev/null || true
+    fi
+    # Safety net: kill anything still on the symbolication port
+    lsof -ti :"$PORT" 2>/dev/null | xargs kill 2>/dev/null || true
+    [ "${PROFILE_KEEP:-0}" = "1" ] || rm -rf "$TMPDIR"
+}
+trap cleanup EXIT
+
+# Step 1: Record
+echo "Recording profile (rate=${RATE} Hz)..." >&2
+samply record --save-only -o "$TMPFILE" -r "$RATE" "$@"
+
+# Step 2: Start samply server for symbolication
+echo "Starting symbolication server..." >&2
+samply load --no-open -P "$PORT" "$TMPFILE" > "$SAMPLY_LOG" 2>&1 &
+SAMPLY_PID=$!
+
+# Wait for server to be ready
+for i in $(seq 1 30); do
+    if grep -q "Local server listening" "$SAMPLY_LOG" 2>/dev/null; then
+        break
+    fi
+    sleep 0.2
+done
+
+# Extract the token from samply's output
+TOKEN=$(grep -oE '[a-z0-9]{30,}' "$SAMPLY_LOG" | head -1)
+
+if [ -z "$TOKEN" ]; then
+    echo "error: could not get samply server token" >&2
+    exit 1
+fi
+
+SERVER_URL="http://127.0.0.1:${PORT}/${TOKEN}"
+
+# Step 3: Symbolicate + demangle
+echo "Symbolicating and demangling..." >&2
+python3 "$SYMBOLICATE" --server "$SERVER_URL" "$TMPFILE" -o "$DEMANGLED"
+
+# Step 4: Kill symbolication server
+kill "$SAMPLY_PID" 2>/dev/null || true
+wait "$SAMPLY_PID" 2>/dev/null || true
+SAMPLY_PID=""
+
+# Step 5: Serve the demangled profile directly (without samply's re-symbolication)
+echo "Opening in Firefox Profiler..." >&2
+python3 "$SERVE_PROFILE" "$DEMANGLED" -P "$SERVE"

script/lib/update-stage0

@@ -1,7 +1,7 @@
 #!/usr/bin/env bash
 set -euo pipefail
 
-rm -r stage0 || true
+rm -rf stage0 || true
 # don't copy untracked files
 # `:!` is git glob flavor for exclude patterns
 for f in $(git ls-files src ':!:src/lake/*' ':!:src/Leanc.lean'); do

script/profiler/lean_demangle.py

@@ -0,0 +1,82 @@
+#!/usr/bin/env python3
+"""
+Lean name demangler — thin wrapper around the Lean CLI tool.
+
+Spawns ``lean --run lean_demangle_cli.lean`` as a persistent subprocess
+and communicates via stdin/stdout pipes. This ensures a single source
+of truth for demangling logic (the Lean implementation in
+``Lean.Compiler.NameDemangling``).
+
+Usage as a filter (like c++filt):
+    echo "l_Lean_Meta_Sym_main" | python lean_demangle.py
+
+Usage as a module:
+    from lean_demangle import demangle_lean_name
+    print(demangle_lean_name("l_Lean_Meta_Sym_main"))
+"""
+
+import atexit
+import os
+import subprocess
+import sys
+
+_process = None
+_script_dir = os.path.dirname(os.path.abspath(__file__))
+_cli_script = os.path.join(_script_dir, "lean_demangle_cli.lean")
+
+
+def _get_process():
+    """Get or create the persistent Lean demangler subprocess."""
+    global _process
+    if _process is not None and _process.poll() is None:
+        return _process
+
+    lean = os.environ.get("LEAN", "lean")
+    _process = subprocess.Popen(
+        [lean, "--run", _cli_script],
+        stdin=subprocess.PIPE,
+        stdout=subprocess.PIPE,
+        stderr=subprocess.DEVNULL,
+        text=True,
+        bufsize=1,  # line buffered
+    )
+    atexit.register(_cleanup)
+    return _process
+
+
+def _cleanup():
+    global _process
+    if _process is not None:
+        try:
+            _process.stdin.close()
+            _process.wait(timeout=5)
+        except Exception:
+            _process.kill()
+        _process = None
+
+
+def demangle_lean_name(mangled):
+    """
+    Demangle a C symbol name produced by the Lean 4 compiler.
+
+    Returns a human-friendly demangled name, or the original string
+    if it is not a Lean symbol.
+    """
+    try:
+        proc = _get_process()
+        proc.stdin.write(mangled + "\n")
+        proc.stdin.flush()
+        result = proc.stdout.readline().rstrip("\n")
+        return result if result else mangled
+    except Exception:
+        return mangled
+
+
+def main():
+    """Filter stdin, demangling Lean names."""
+    for line in sys.stdin:
+        print(demangle_lean_name(line.rstrip("\n")))
+
+
+if __name__ == "__main__":
+    main()

script/profiler/lean_demangle_cli.lean

@@ -0,0 +1,32 @@
+/-
+Copyright (c) 2026 Lean FRO, LLC. All rights reserved.
+Released under Apache 2.0 license as described in the file LICENSE.
+Authors: Kim Morrison
+-/
+module
+
+import Lean.Compiler.NameDemangling
+
+/-!
+Lean name demangler CLI tool. Reads mangled symbol names from stdin (one per
+line) and writes demangled names to stdout. Non-Lean symbols pass through
+unchanged. Like `c++filt` but for Lean names.
+
+Usage:
+    echo "l_Lean_Meta_foo" | lean --run lean_demangle_cli.lean
+    cat symbols.txt | lean --run lean_demangle_cli.lean
+-/
+
+open Lean.Name.Demangle
+
+def main : IO Unit := do
+  let stdin ← IO.getStdin
+  let stdout ← IO.getStdout
+  repeat do
+    let line ← stdin.getLine
+    if line.isEmpty then break
+    let sym := line.trimRight
+    match demangleSymbol sym with
+    | some s => stdout.putStrLn s
+    | none => stdout.putStrLn sym
+    stdout.flush

script/profiler/lean_demangle_profile.py

@@ -0,0 +1,117 @@
+#!/usr/bin/env python3
+"""
+Lean name demangler for samply / Firefox Profiler profiles.
+
+Reads a profile JSON (plain or gzipped), demangles Lean function names
+in the string table, and writes the result back.
+
+Usage:
+    python lean_demangle_profile.py profile.json -o profile-demangled.json
+    python lean_demangle_profile.py profile.json.gz -o profile-demangled.json.gz
+"""
+
+import argparse
+import gzip
+import json
+import sys
+
+from lean_demangle import demangle_lean_name
+
+
+def _demangle_string_array(string_array):
+    """Demangle Lean names in a string array in-place. Returns count."""
+    count = 0
+    for i, s in enumerate(string_array):
+        if not isinstance(s, str):
+            continue
+        demangled = demangle_lean_name(s)
+        if demangled != s:
+            string_array[i] = demangled
+            count += 1
+    return count
+
+
+def rewrite_profile(profile):
+    """
+    Demangle Lean names in a Firefox Profiler profile dict (in-place).
+
+    Handles two profile formats:
+    - Newer: shared.stringArray (single shared string table)
+    - Older/samply: per-thread stringArray (each thread has its own)
+    """
+    count = 0
+
+    # Shared string table (newer Firefox Profiler format)
+    shared = profile.get("shared")
+    if shared is not None:
+        sa = shared.get("stringArray")
+        if sa is not None:
+            count += _demangle_string_array(sa)
+
+    # Per-thread string tables (samply format)
+    for thread in profile.get("threads", []):
+        sa = thread.get("stringArray")
+        if sa is not None:
+            count += _demangle_string_array(sa)
+
+    return count
+
+
+def process_profile_file(input_path, output_path):
+    """Read a profile, demangle names, write it back."""
+    is_gzip = input_path.endswith('.gz')
+
+    if is_gzip:
+        with gzip.open(input_path, 'rt', encoding='utf-8') as f:
+            profile = json.load(f)
+    else:
+        with open(input_path, 'r', encoding='utf-8') as f:
+            profile = json.load(f)
+
+    count = rewrite_profile(profile)
+
+    out_gzip = output_path.endswith('.gz') if output_path else is_gzip
+
+    if output_path:
+        if out_gzip:
+            with gzip.open(output_path, 'wt', encoding='utf-8') as f:
+                json.dump(profile, f, ensure_ascii=False)
+        else:
+            with open(output_path, 'w', encoding='utf-8') as f:
+                json.dump(profile, f, ensure_ascii=False)
+    else:
+        json.dump(profile, sys.stdout, ensure_ascii=False)
+        sys.stdout.write('\n')
+
+    return count
+
+
+def main():
+    parser = argparse.ArgumentParser(
+        description="Demangle Lean names in samply/Firefox Profiler profiles")
+    parser.add_argument('input', help='Input profile (JSON or .json.gz)')
+    parser.add_argument('-o', '--output',
+                        help='Output path (default: stdout for JSON, '
+                             'or input with -demangled suffix)')
+    args = parser.parse_args()
+
+    output = args.output
+    if output is None and not sys.stdout.isatty():
+        output = None  # write to stdout
+    elif output is None:
+        # Generate output filename
+        inp = args.input
+        if inp.endswith('.json.gz'):
+            output = inp[:-8] + '-demangled.json.gz'
+        elif inp.endswith('.json'):
+            output = inp[:-5] + '-demangled.json'
+        else:
+            output = inp + '-demangled'
+
+    count = process_profile_file(args.input, output)
+    if output:
+        print(f"Demangled {count} names, wrote {output}", file=sys.stderr)
+
+
+if __name__ == '__main__':
+    main()

script/profiler/serve_profile.py

@@ -0,0 +1,94 @@
+#!/usr/bin/env python3
+"""
+Serve a Firefox Profiler JSON file and open it in the browser.
+
+Unlike `samply load`, this does NOT provide a symbolication API,
+so Firefox Profiler will use the names already in the profile as-is.
+"""
+
+import argparse
+import gzip
+import http.server
+import io
+import sys
+import threading
+import webbrowser
+import urllib.parse
+
+
+class ProfileHandler(http.server.BaseHTTPRequestHandler):
+    """Serve the profile JSON and handle CORS for Firefox Profiler."""
+
+    profile_data = None  # set by main()
+
+    def do_GET(self):
+        if self.path == "/profile.json":
+            self.send_response(200)
+            self.send_header("Content-Type", "application/json")
+            self.send_header("Content-Encoding", "gzip")
+            self.send_header("Access-Control-Allow-Origin", "*")
+            self.end_headers()
+            self.wfile.write(self.profile_data)
+        else:
+            self.send_response(404)
+            self.end_headers()
+
+    def do_OPTIONS(self):
+        # CORS preflight
+        self.send_response(200)
+        self.send_header("Access-Control-Allow-Origin", "*")
+        self.send_header("Access-Control-Allow-Methods", "GET")
+        self.send_header("Access-Control-Allow-Headers", "Content-Type")
+        self.end_headers()
+
+    def log_message(self, format, *args):
+        pass  # suppress request logs
+
+
+def main():
+    parser = argparse.ArgumentParser(
+        description="Serve a profile JSON for Firefox Profiler")
+    parser.add_argument("profile", help="Profile file (.json or .json.gz)")
+    parser.add_argument("-P", "--port", type=int, default=3457,
+                        help="Port to serve on (default: 3457)")
+    parser.add_argument("-n", "--no-open", action="store_true",
+                        help="Do not open the browser")
+    args = parser.parse_args()
+
+    # Read the profile data (keep it gzipped for efficient serving)
+    if args.profile.endswith(".gz"):
+        with open(args.profile, "rb") as f:
+            ProfileHandler.profile_data = f.read()
+    else:
+        with open(args.profile, "rb") as f:
+            raw = f.read()
+        buf = io.BytesIO()
+        with gzip.GzipFile(fileobj=buf, mode="wb") as gz:
+            gz.write(raw)
+        ProfileHandler.profile_data = buf.getvalue()
+
+    http.server.HTTPServer.allow_reuse_address = True
+    server = http.server.HTTPServer(("127.0.0.1", args.port), ProfileHandler)
+    profile_url = f"http://127.0.0.1:{args.port}/profile.json"
+    encoded = urllib.parse.quote(profile_url, safe="")
+    viewer_url = f"https://profiler.firefox.com/from-url/{encoded}"
+
+    if not args.no_open:
+        # Open browser after a short delay to let server start
+        def open_browser():
+            webbrowser.open(viewer_url)
+        threading.Timer(0.5, open_browser).start()
+
+    print(f"Serving profile at {profile_url}", file=sys.stderr)
+    print(f"Firefox Profiler: {viewer_url}", file=sys.stderr)
+    print("Press Ctrl+C to stop.", file=sys.stderr)
+
+    try:
+        server.serve_forever()
+    except KeyboardInterrupt:
+        print("\nStopped.", file=sys.stderr)
+        server.server_close()
+
+
+if __name__ == "__main__":
+    main()

script/profiler/symbolicate_profile.py

@@ -0,0 +1,198 @@
+#!/usr/bin/env python3
+"""
+Symbolicate a raw samply profile using samply's symbolication API,
+then demangle Lean names.
+
+Usage:
+    python symbolicate_profile.py --server http://127.0.0.1:3000/TOKEN \
+        raw-profile.json.gz -o symbolicated-demangled.json.gz
+"""
+
+import argparse
+import gzip
+import json
+import sys
+import urllib.request
+
+from lean_demangle import demangle_lean_name
+
+
+def symbolicate_and_demangle(profile, server_url):
+    """
+    Symbolicate a raw samply profile via the symbolication API,
+    then demangle Lean names. Modifies the profile in-place.
+    Returns the number of names resolved.
+    """
+    libs = profile.get("libs", [])
+    memory_map = [[lib["debugName"], lib["breakpadId"]] for lib in libs]
+
+    count = 0
+    for thread in profile.get("threads", []):
+        count += _process_thread(thread, libs, memory_map, server_url)
+
+    return count
+
+
+def _process_thread(thread, libs, memory_map, server_url):
+    """Symbolicate and demangle one thread. Returns count of resolved names."""
+    sa = thread.get("stringArray")
+    ft = thread.get("frameTable")
+    func_t = thread.get("funcTable")
+    rt = thread.get("resourceTable")
+
+    if not all([sa, ft, func_t, rt]):
+        return 0
+
+    # Build mapping: func_index -> (lib_index, address)
+    # A function may be referenced by multiple frames; pick any address.
+    func_info = {}  # func_idx -> (lib_idx, address)
+    for i in range(ft.get("length", 0)):
+        addr = ft["address"][i]
+        func_idx = ft["func"][i]
+        if func_idx in func_info:
+            continue
+        res_idx = func_t["resource"][func_idx]
+        if res_idx < 0 or res_idx >= rt.get("length", 0):
+            continue
+        lib_idx = rt["lib"][res_idx]
+        if lib_idx < 0 or lib_idx >= len(libs):
+            continue
+        func_info[func_idx] = (lib_idx, addr)
+
+    if not func_info:
+        return 0
+
+    # Batch symbolication: group by lib, send all addresses at once
+    frames_to_symbolicate = []
+    func_order = []  # track which func each frame corresponds to
+    for func_idx, (lib_idx, addr) in func_info.items():
+        frames_to_symbolicate.append([lib_idx, addr])
+        func_order.append(func_idx)
+
+    # Call the symbolication API
+    symbols = _call_symbolication_api(
+        server_url, memory_map, frames_to_symbolicate)
+
+    if not symbols:
+        return 0
+
+    # Update stringArray with demangled names
+    count = 0
+    for func_idx, symbol_name in zip(func_order, symbols):
+        if symbol_name is None:
+            continue
+        demangled = demangle_lean_name(symbol_name)
+        name_idx = func_t["name"][func_idx]
+        if name_idx < len(sa):
+            sa[name_idx] = demangled
+            count += 1
+
+    return count
+
+
+def _call_symbolication_api(server_url, memory_map, frames):
+    """
+    Call the Firefox Profiler symbolication API v5.
+    frames: list of [lib_index, address]
+    Returns: list of symbol names (or None for unresolved frames).
+    """
+    url = server_url.rstrip("/") + "/symbolicate/v5"
+
+    # Send all frames as one "stack" in one job
+    req_body = json.dumps({
+        "memoryMap": memory_map,
+        "stacks": [frames],
+    }).encode()
+
+    req = urllib.request.Request(
+        url,
+        data=req_body,
+        headers={"Content-Type": "application/json"},
+    )
+
+    try:
+        with urllib.request.urlopen(req, timeout=60) as resp:
+            result = json.loads(resp.read())
+    except Exception as e:
+        print(f"Symbolication API error: {e}", file=sys.stderr)
+        return None
+
+    if "error" in result:
+        print(f"Symbolication API error: {result['error']}", file=sys.stderr)
+        return None
+
+    # Extract symbol names from result
+    results = result.get("results", [])
+    if not results:
+        return None
+
+    stacks = results[0].get("stacks", [[]])
+    if not stacks:
+        return None
+
+    symbols = []
+    for frame_result in stacks[0]:
+        if isinstance(frame_result, dict):
+            symbols.append(frame_result.get("function"))
+        elif isinstance(frame_result, str):
+            symbols.append(frame_result)
+        else:
+            symbols.append(None)
+
+    return symbols
+
+
+def process_file(input_path, output_path, server_url):
+    """Read a raw profile, symbolicate + demangle, write it back."""
+    is_gzip = input_path.endswith('.gz')
+
+    if is_gzip:
+        with gzip.open(input_path, 'rt', encoding='utf-8') as f:
+            profile = json.load(f)
+    else:
+        with open(input_path, 'r', encoding='utf-8') as f:
+            profile = json.load(f)
+
+    count = symbolicate_and_demangle(profile, server_url)
+
+    out_gzip = output_path.endswith('.gz') if output_path else is_gzip
+    if output_path:
+        if out_gzip:
+            with gzip.open(output_path, 'wt', encoding='utf-8') as f:
+                json.dump(profile, f, ensure_ascii=False)
+        else:
+            with open(output_path, 'w', encoding='utf-8') as f:
+                json.dump(profile, f, ensure_ascii=False)
+    else:
+        json.dump(profile, sys.stdout, ensure_ascii=False)
+        sys.stdout.write('\n')
+
+    return count
+
+
+def main():
+    parser = argparse.ArgumentParser(
+        description="Symbolicate a raw samply profile and demangle Lean names")
+    parser.add_argument('input', help='Raw profile (JSON or .json.gz)')
+    parser.add_argument('-o', '--output', help='Output path')
+    parser.add_argument('--server', required=True,
+                        help='Samply server URL (e.g., http://127.0.0.1:3000/TOKEN)')
+    args = parser.parse_args()
+
+    output = args.output
+    if output is None:
+        inp = args.input
+        if inp.endswith('.json.gz'):
+            output = inp[:-8] + '-demangled.json.gz'
+        elif inp.endswith('.json'):
+            output = inp[:-5] + '-demangled.json'
+        else:
+            output = inp + '-demangled'
+
+    count = process_file(args.input, output, args.server)
+    print(f"Symbolicated and demangled {count} names, wrote {output}",
+          file=sys.stderr)
+
+
+if __name__ == '__main__':
+    main()

script/release_checklist.py

@@ -11,7 +11,7 @@ IMPORTANT: Keep this documentation up-to-date when modifying the script's behavi
 What this script does:
 1. Validates preliminary Lean4 release infrastructure:
    - Checks that the release branch (releases/vX.Y.0) exists
-   - Verifies CMake version settings are correct
+   - Verifies CMake version settings are correct (both src/ and stage0/)
    - Confirms the release tag exists
    - Validates the release page exists on GitHub (created automatically by CI after tag push)
    - Checks the release notes page on lean-lang.org (updated while bumping the `reference-manual` repository)
@@ -326,6 +326,42 @@ def check_cmake_version(repo_url, branch, version_major, version_minor, github_t
     print(f"  ✅ CMake version settings are correct in {cmake_file_path}")
     return True
 
+def check_stage0_version(repo_url, branch, version_major, version_minor, github_token):
+    """Verify that stage0/src/CMakeLists.txt has the same version as src/CMakeLists.txt.
+
+    The stage0 pre-built binaries stamp .olean headers with their baked-in version.
+    If stage0 has a different version (e.g. from a 'begin development cycle' bump),
+    the release tarball will contain .olean files with the wrong version.
+    """
+    stage0_cmake = "stage0/src/CMakeLists.txt"
+    content = get_branch_content(repo_url, branch, stage0_cmake, github_token)
+    if content is None:
+        print(f"  ❌ Could not retrieve {stage0_cmake} from {branch}")
+        return False
+
+    errors = []
+    for line in content.splitlines():
+        stripped = line.strip()
+        if stripped.startswith("set(LEAN_VERSION_MAJOR "):
+            actual = stripped.split()[-1].rstrip(")")
+            if actual != str(version_major):
+                errors.append(f"LEAN_VERSION_MAJOR: expected {version_major}, found {actual}")
+        elif stripped.startswith("set(LEAN_VERSION_MINOR "):
+            actual = stripped.split()[-1].rstrip(")")
+            if actual != str(version_minor):
+                errors.append(f"LEAN_VERSION_MINOR: expected {version_minor}, found {actual}")
+
+    if errors:
+        print(f"  ❌ stage0 version mismatch in {stage0_cmake}:")
+        for error in errors:
+            print(f"     {error}")
+        print(f"     The stage0 compiler stamps .olean headers with its baked-in version.")
+        print(f"     Run `make update-stage0` to rebuild stage0 with the correct version.")
+        return False
+
+    print(f"  ✅ stage0 version matches in {stage0_cmake}")
+    return True
+
 def extract_org_repo_from_url(repo_url):
     """Extract the 'org/repo' part from a GitHub URL."""
     if repo_url.startswith("https://github.com/"):
@@ -441,7 +477,10 @@ def get_pr_ci_status(repo_url, pr_number, github_token):
     conclusions = [run['conclusion'] for run in check_runs if run.get('status') == 'completed']
     in_progress = [run for run in check_runs if run.get('status') in ['queued', 'in_progress']]
 
+    failed = sum(1 for c in conclusions if c in ['failure', 'timed_out', 'action_required'])
     if in_progress:
+        if failed > 0:
+            return "failure", f"{failed} check(s) failing, {len(in_progress)} still in progress"
         return "pending", f"{len(in_progress)} check(s) in progress"
 
     if not conclusions:
@@ -450,7 +489,6 @@ def get_pr_ci_status(repo_url, pr_number, github_token):
     if all(c == 'success' for c in conclusions):
         return "success", f"All {len(conclusions)} checks passed"
 
-    failed = sum(1 for c in conclusions if c in ['failure', 'timed_out', 'action_required'])
     if failed > 0:
         return "failure", f"{failed} check(s) failed"
 
@@ -680,6 +718,9 @@ def main():
         # Check CMake version settings
         if not check_cmake_version(lean_repo_url, branch_name, version_major, version_minor, github_token):
             lean4_success = False
+        # Check that stage0 version matches (stage0 stamps .olean headers with its version)
+        if not check_stage0_version(lean_repo_url, branch_name, version_major, version_minor, github_token):
+            lean4_success = False
 
     # Check for tag and release page
     if not tag_exists(lean_repo_url, toolchain, github_token):
@@ -836,6 +877,14 @@ def main():
             continue
         print(f"  ✅ On compatible toolchain (>= {toolchain})")
 
+        # For reference-manual, check that the release notes title is correct BEFORE tagging.
+        # This catches the case where the toolchain bump PR was merged without updating
+        # the release notes title (e.g., still showing "-rc1" for a stable release).
+        if name == "reference-manual":
+            if not check_reference_manual_release_title(url, toolchain, branch, github_token):
+                repo_status[name] = False
+                continue
+
         # Special handling for ProofWidgets4
         if name == "ProofWidgets4":
             if not check_proofwidgets4_release(url, toolchain, github_token):
@@ -916,8 +965,8 @@ def main():
             
             print(f"  ✅ Bump branch {bump_branch} exists")
             
-            # For batteries and mathlib4, update the lean-toolchain to the latest nightly
-            if branch_created and name in ["batteries", "mathlib4"]:
+            # Update the lean-toolchain to the latest nightly for newly created bump branches
+            if branch_created:
                 latest_nightly = get_latest_nightly_tag(github_token)
                 if latest_nightly:
                     nightly_toolchain = f"leanprover/lean4:{latest_nightly}"
@@ -957,14 +1006,15 @@ def main():
         # Find the actual minor version in CMakeLists.txt
         for line in cmake_lines:
             if line.strip().startswith("set(LEAN_VERSION_MINOR "):
-                actual_minor = int(line.split()[-1].rstrip(")"))
+                m = re.search(r'set\(LEAN_VERSION_MINOR\s+(\d+)', line)
+                actual_minor = int(m.group(1)) if m else 0
                 version_minor_correct = actual_minor >= next_minor
                 break
         else:
             version_minor_correct = False
             
         is_release_correct = any(
-            l.strip().startswith("set(LEAN_VERSION_IS_RELEASE 0)") 
+            re.match(r'set\(LEAN_VERSION_IS_RELEASE\s+0[\s)]', l.strip())
             for l in cmake_lines
         )
         

script/release_repos.yml

@@ -65,13 +65,6 @@ repositories:
     branch: master
     dependencies: [lean4-unicode-basic]
 
-  - name: doc-gen4
-    url: https://github.com/leanprover/doc-gen4
-    toolchain-tag: true
-    stable-branch: false
-    branch: main
-    dependencies: [lean4-cli, BibtexQuery]
-
   - name: reference-manual
     url: https://github.com/leanprover/reference-manual
     toolchain-tag: true
@@ -84,8 +77,7 @@ repositories:
     toolchain-tag: false
     stable-branch: false
     branch: main
-    dependencies:
-      - batteries
+    dependencies: []
 
   - name: aesop
     url: https://github.com/leanprover-community/aesop
@@ -107,10 +99,16 @@ repositories:
       - lean4checker
       - batteries
       - lean4-cli
-      - doc-gen4
       - import-graph
       - plausible
 
+  - name: doc-gen4
+    url: https://github.com/leanprover/doc-gen4
+    toolchain-tag: true
+    stable-branch: false
+    branch: main
+    dependencies: [lean4-cli, BibtexQuery, mathlib4]
+
   - name: cslib
     url: https://github.com/leanprover/cslib
     toolchain-tag: true

script/release_steps.py

@@ -24,6 +24,7 @@ What this script does:
    - Safety checks for repositories using bump branches
    - Custom build and test procedures
    - lean-fro.org: runs scripts/update.sh to regenerate site content
+   - mathlib4: updates ProofWidgets4 pin (v0.0.X sequential tags, not v4.X.Y)
 
 6. Commits the changes with message "chore: bump toolchain to {version}"
 
@@ -59,6 +60,8 @@ import re
 import subprocess
 import shutil
 import json
+import requests
+import base64
 from pathlib import Path
 
 # Color functions for terminal output
@@ -115,6 +118,60 @@ def find_repo(repo_name, config):
         sys.exit(1)
     return matching_repos[0]
 
+def get_github_token():
+    try:
+        result = subprocess.run(['gh', 'auth', 'token'], capture_output=True, text=True)
+        if result.returncode == 0:
+            return result.stdout.strip()
+    except FileNotFoundError:
+        pass
+    return None
+
+def find_proofwidgets_tag(version):
+    """Find the latest ProofWidgets4 tag that uses the given toolchain version.
+
+    ProofWidgets4 uses sequential version tags (v0.0.X) rather than toolchain-based tags.
+    This function finds the most recent tag whose lean-toolchain matches the target version
+    exactly, checking the 20 most recent tags.
+    """
+    github_token = get_github_token()
+    api_base = "https://api.github.com/repos/leanprover-community/ProofWidgets4"
+    headers = {'Authorization': f'token {github_token}'} if github_token else {}
+
+    response = requests.get(f"{api_base}/git/matching-refs/tags/v0.0.", headers=headers, timeout=30)
+    if response.status_code != 200:
+        return None
+
+    tags = response.json()
+    tag_names = []
+    for tag in tags:
+        ref = tag['ref']
+        if ref.startswith('refs/tags/v0.0.'):
+            tag_name = ref.replace('refs/tags/', '')
+            try:
+                version_num = int(tag_name.split('.')[-1])
+                tag_names.append((version_num, tag_name))
+            except (ValueError, IndexError):
+                continue
+
+    if not tag_names:
+        return None
+
+    # Sort by version number (descending) and check recent tags
+    tag_names.sort(reverse=True)
+    target = f"leanprover/lean4:{version}"
+    for _, tag_name in tag_names[:20]:
+        # Fetch lean-toolchain for this tag
+        api_url = f"{api_base}/contents/lean-toolchain?ref={tag_name}"
+        resp = requests.get(api_url, headers=headers, timeout=30)
+        if resp.status_code != 200:
+            continue
+        content = base64.b64decode(resp.json().get("content", "").replace("\n", "")).decode('utf-8').strip()
+        if content == target:
+            return tag_name
+
+    return None
+
 def setup_downstream_releases_dir():
     """Create the downstream_releases directory if it doesn't exist."""
     downstream_dir = Path("downstream_releases")
@@ -422,10 +479,84 @@ def execute_release_steps(repo, version, config):
         print(blue("Updating lakefile.toml..."))
         run_command(f'perl -pi -e \'s/"v4\\.[0-9]+(\\.[0-9]+)?(-rc[0-9]+)?"/"' + version + '"/g\' lakefile.*', cwd=repo_path)
         run_command("lake update", cwd=repo_path, stream_output=True)
+    elif repo_name == "verso":
+        # verso has nested Lake projects in test-projects/ that each have their own
+        # lake-manifest.json with a subverso pin. After updating the root manifest via
+        # `lake update`, sync the de-modulized subverso rev into all sub-manifests.
+        # The sub-projects use an old toolchain (v4.21.0) that doesn't support module/prelude
+        # syntax, so they need the de-modulized version (tagged no-modules/<root-rev>).
+        # The "SubVerso version consistency" CI check accepts either the root or de-modulized rev.
+        run_command("lake update", cwd=repo_path, stream_output=True)
+        print(blue("Syncing de-modulized subverso rev to test-project sub-manifests..."))
+        sync_script = (
+            'ROOT_REV=$(jq -r \'.packages[] | select(.name == "subverso") | .rev\' lake-manifest.json); '
+            'SUBVERSO_URL=$(jq -r \'.packages[] | select(.name == "subverso") | .url\' lake-manifest.json); '
+            'DEMOD_REV=$(git ls-remote "$SUBVERSO_URL" "refs/tags/no-modules/$ROOT_REV" | awk \'{print $1}\'); '
+            'find test-projects -name lake-manifest.json -print0 | '
+            'xargs -0 -I{} sh -c \'jq --arg rev "$DEMOD_REV" \'.packages |= map(if .name == "subverso" then .rev = $rev else . end)\' "{}" > /tmp/lm_tmp.json && mv /tmp/lm_tmp.json "{}"\''
+        )
+        run_command(sync_script, cwd=repo_path)
+        print(green("Synced de-modulized subverso rev to all test-project sub-manifests"))
     elif dependencies:
         run_command(f'perl -pi -e \'s/"v4\\.[0-9]+(\\.[0-9]+)?(-rc[0-9]+)?"/"' + version + '"/g\' lakefile.*', cwd=repo_path)
         run_command("lake update", cwd=repo_path, stream_output=True)
 
+    # For reference-manual, update the release notes title to match the target version.
+    # e.g., for a stable release, change "Lean 4.28.0-rc1 (date)" to "Lean 4.28.0 (date)"
+    # e.g., for rc2, change "Lean 4.28.0-rc1 (date)" to "Lean 4.28.0-rc2 (date)"
+    if repo_name == "reference-manual":
+        base_version = version.lstrip('v').split('-')[0]  # "4.28.0"
+        file_name = f"v{base_version.replace('.', '_')}.lean"
+        release_notes_file = repo_path / "Manual" / "Releases" / file_name
+
+        if release_notes_file.exists():
+            is_rc = "-rc" in version
+            if is_rc:
+                # For RC releases, update to the exact RC version
+                display_version = version.lstrip('v')  # "4.28.0-rc2"
+            else:
+                # For stable releases, strip any RC suffix
+                display_version = base_version  # "4.28.0"
+
+            print(blue(f"Updating release notes title in {file_name}..."))
+            content = release_notes_file.read_text()
+            # Match the #doc line title: "Lean X.Y.Z-rcN (date)" or "Lean X.Y.Z (date)"
+            new_content = re.sub(
+                r'(#doc\s+\(Manual\)\s+"Lean\s+)\d+\.\d+\.\d+(-rc\d+)?(\s+\([^)]*\)"\s*=>)',
+                rf'\g<1>{display_version}\3',
+                content
+            )
+            if new_content != content:
+                release_notes_file.write_text(new_content)
+                print(green(f"Updated release notes title to Lean {display_version}"))
+            else:
+                print(green("Release notes title already correct"))
+        else:
+            print(yellow(f"Release notes file {file_name} not found, skipping title update"))
+
+    # For mathlib4, update ProofWidgets4 pin (it uses sequential v0.0.X tags, not v4.X.Y)
+    if repo_name == "mathlib4":
+        print(blue("Checking ProofWidgets4 version pin..."))
+        pw_tag = find_proofwidgets_tag(version)
+        if pw_tag:
+            print(blue(f"Updating ProofWidgets4 pin to {pw_tag}..."))
+            for lakefile in repo_path.glob("lakefile.*"):
+                content = lakefile.read_text()
+                # Only update the ProofWidgets4 dependency line, not other v0.0.X pins
+                new_content = re.sub(
+                    r'(require\s+"leanprover-community"\s*/\s*"proofwidgets"\s*@\s*git\s+"v)0\.0\.\d+(")',
+                    rf'\g<1>{pw_tag.removeprefix("v")}\2',
+                    content
+                )
+                if new_content != content:
+                    lakefile.write_text(new_content)
+                    print(green(f"Updated ProofWidgets4 pin in {lakefile.name}"))
+            run_command("lake update proofwidgets", cwd=repo_path, stream_output=True)
+            print(green(f"Updated ProofWidgets4 to {pw_tag}"))
+        else:
+            print(yellow(f"Could not find a ProofWidgets4 tag for toolchain {version}"))
+            print(yellow("You may need to update the ProofWidgets4 pin manually"))
+
     # Commit changes (only if there are changes)
     print(blue("Checking for changes to commit..."))
     try:

src/CMakeLists.txt

@@ -1,6 +1,4 @@
-cmake_minimum_required(VERSION 3.10)
-cmake_policy(SET CMP0054 NEW)
-cmake_policy(SET CMP0110 NEW)
+cmake_minimum_required(VERSION 3.21)
 if(NOT CMAKE_GENERATOR MATCHES "Unix Makefiles")
   message(FATAL_ERROR "The only supported CMake generator at the moment is 'Unix Makefiles'")
 endif()
@@ -9,11 +7,17 @@ if(NOT DEFINED STAGE)
 endif()
 include(ExternalProject)
 project(LEAN CXX C)
-set(LEAN_VERSION_MAJOR 4)
-set(LEAN_VERSION_MINOR 29)
-set(LEAN_VERSION_PATCH 0)
-set(LEAN_VERSION_IS_RELEASE 0) # This number is 1 in the release revision, and 0 otherwise.
+set(LEAN_VERSION_MAJOR 4 CACHE STRING "")
+set(LEAN_VERSION_MINOR 30 CACHE STRING "")
+set(LEAN_VERSION_PATCH 0 CACHE STRING "")
+set(LEAN_VERSION_IS_RELEASE 0 CACHE STRING "") # This number is 1 in the release revision, and 0 otherwise.
 set(LEAN_SPECIAL_VERSION_DESC "" CACHE STRING "Additional version description like 'nightly-2018-03-11'")
+# project(LEAN) above implicitly creates empty LEAN_VERSION_{MAJOR,MINOR,PATCH}
+# normal variables (CMake sets <PROJECT>_VERSION_* for the project name). These
+# shadow the cache values. Remove them so ${VAR} falls through to the cache.
+unset(LEAN_VERSION_MAJOR)
+unset(LEAN_VERSION_MINOR)
+unset(LEAN_VERSION_PATCH)
 set(LEAN_VERSION_STRING "${LEAN_VERSION_MAJOR}.${LEAN_VERSION_MINOR}.${LEAN_VERSION_PATCH}")
 if(LEAN_SPECIAL_VERSION_DESC)
   string(APPEND LEAN_VERSION_STRING "-${LEAN_SPECIAL_VERSION_DESC}")

src/Init/Classical.lean

@@ -69,9 +69,11 @@ theorem em (p : Prop) : p ∨ ¬p :=
 theorem exists_true_of_nonempty {α : Sort u} : Nonempty α → ∃ _ : α, True
   | ⟨x⟩ => ⟨x, trivial⟩
 
+@[implicit_reducible]
 noncomputable def inhabited_of_nonempty {α : Sort u} (h : Nonempty α) : Inhabited α :=
   ⟨choice h⟩
 
+@[implicit_reducible]
 noncomputable def inhabited_of_exists {α : Sort u} {p : α → Prop} (h : ∃ x, p x) : Inhabited α :=
   inhabited_of_nonempty (Exists.elim h (fun w _ => ⟨w⟩))
 
@@ -81,6 +83,7 @@ noncomputable scoped instance (priority := low) propDecidable (a : Prop) : Decid
     | Or.inl h => ⟨isTrue h⟩
     | Or.inr h => ⟨isFalse h⟩
 
+@[implicit_reducible]
 noncomputable def decidableInhabited (a : Prop) : Inhabited (Decidable a) where
   default := inferInstance
 
@@ -142,7 +145,7 @@ is classically true but not constructively. -/
 
 /-- Transfer decidability of `¬ p` to decidability of `p`. -/
 -- This can not be an instance as it would be tried everywhere.
-@[instance_reducible]
+@[implicit_reducible]
 def decidable_of_decidable_not (p : Prop) [h : Decidable (¬ p)] : Decidable p :=
   match h with
   | isFalse h => isTrue (Classical.not_not.mp h)

src/Init/Control/Do.lean

@@ -0,0 +1,63 @@
+/-
+Copyright (c) 2025 Lean FRO LLC. All rights reserved.
+Released under Apache 2.0 license as described in the file LICENSE.
+Authors: Sebastian Graf
+-/
+module
+
+prelude
+public import Init.Control.Except
+public import Init.Control.Option
+
+public section
+
+/-!
+This module provides specialized wrappers around `ExceptT` to support the `do` elaborator.
+
+Specifically, the types here are used to tunnel early `return`, `break` and `continue` through
+non-algebraic higher-order effect combinators such as `tryCatch`.
+-/
+
+/-- A wrapper around `ExceptT` signifying early return. -/
+@[expose]
+abbrev EarlyReturnT (ρ m α) := ExceptT ρ m α
+
+/-- Exit a computation by returning a value `r : ρ` early. -/
+@[always_inline, inline, expose]
+abbrev EarlyReturnT.return {ρ m α} [Monad m] (r : ρ) : EarlyReturnT ρ m α :=
+  throw r
+
+/-- A specialization of `Except.casesOn`. -/
+@[always_inline, inline, expose]
+abbrev EarlyReturn.runK {ρ α : Type u} {β : Type v} (x : Except ρ α) (ret : ρ → β) (pure : α → β) : β :=
+  x.casesOn ret pure
+
+/-- A wrapper around `OptionT` signifying `break` in a loop. -/
+@[expose]
+abbrev BreakT := OptionT
+
+/-- Exit a loop body via `break`. -/
+@[always_inline, inline, expose]
+abbrev BreakT.break {m : Type w → Type x} [Monad m] : BreakT m α := failure
+
+/-- A specialization of `Option.casesOn`. -/
+@[always_inline, inline, expose]
+abbrev Break.runK {α : Type u} {β : Type v} (x : Option α) (breakK : Unit → β) (successK : α → β) : β :=
+  -- Note: The matcher below is used in the elaborator targeting `forIn` loops.
+  -- If you change the order of match arms here, you may need to adjust the elaborator.
+  match x with
+  | some a => successK a
+  | none => breakK ()
+
+/-- A wrapper around `OptionT` signifying `continue` in a loop. -/
+@[expose]
+abbrev ContinueT := OptionT
+
+/-- Exit a loop body via `continue`. -/
+@[always_inline, inline, expose]
+abbrev ContinueT.continue {m : Type w → Type x} [Monad m] : ContinueT m α := failure
+
+/-- A specialization of `Option.casesOn`. -/
+@[always_inline, inline, expose]
+abbrev Continue.runK {α : Type u} {β : Type v} (x : Option α) (continueK : Unit → β) (successK : α → β) : β :=
+  x.casesOn continueK (fun a _ => successK a) ()

src/Init/Control/Id.lean

@@ -49,6 +49,7 @@ instance : Monad Id where
 /--
 The identity monad has a `bind` operator.
 -/
+@[implicit_reducible]
 def hasBind : Bind Id :=
   inferInstance
 
@@ -58,7 +59,7 @@ Runs a computation in the identity monad.
 This function is the identity function. Because its parameter has type `Id α`, it causes
 `do`-notation in its arguments to use the `Monad Id` instance.
 -/
-@[always_inline, inline, expose]
+@[always_inline, inline, expose, implicit_reducible]
 protected def run (x : Id α) : α := x
 
 instance [OfNat α n] : OfNat (Id α) n :=
@@ -79,3 +80,11 @@ instance : LawfulMonadAttach Id where
     exact x.run.2
 
 end Id
+
+/-- Turn a collection with a pure `ForIn` instance into an array. -/
+def ForIn.toArray {α : Type u} [inst : ForIn Id ρ α] (xs : ρ) : Array α :=
+  ForIn.forIn xs Array.empty (fun a acc => pure (.yield (acc.push a))) |> Id.run
+
+/-- Turn a collection with a pure `ForIn` instance into a list. -/
+def ForIn.toList {α : Type u} [ForIn Id ρ α] (xs : ρ) : List α :=
+  ForIn.toArray xs |>.toList

src/Init/Control/Lawful/Instances.lean

@@ -30,6 +30,8 @@ namespace ExceptT
   simp [run] at h
   assumption
 
+@[simp] theorem stM_eq [Monad m] : stM m (ExceptT ε m) α = Except ε α := rfl
+
 @[simp, grind =] theorem run_mk (x : m (Except ε α)) : run (mk x : ExceptT ε m α) = x := rfl
 
 @[simp, grind =] theorem run_pure [Monad m] (x : α) : run (pure x : ExceptT ε m α) = pure (Except.ok x) := rfl
@@ -118,7 +120,7 @@ instance [Monad m] [LawfulMonad m] : LawfulMonad (ExceptT ε m) where
 
 @[simp] theorem run_controlAt [Monad m] [LawfulMonad m] (f : ({β : Type u} → ExceptT ε m β → m (stM m (ExceptT ε m) β)) → m (stM m (ExceptT ε m) α)) :
     ExceptT.run (controlAt m f) = f fun x => x.run := by
-  simp [controlAt, run_bind, bind_map_left]
+  simp [controlAt, run_bind]
 
 @[simp] theorem run_control [Monad m] [LawfulMonad m] (f : ({β : Type u} → ExceptT ε m β → m (stM m (ExceptT ε m) β)) → m (stM m (ExceptT ε m) α)) :
     ExceptT.run (control f) = f fun x => x.run := run_controlAt f
@@ -437,7 +439,6 @@ instance [Monad m] [LawfulMonad m] : LawfulMonad (StateT σ m) where
 @[simp] theorem run_restoreM [Monad m] [LawfulMonad m] (x : stM m (StateT σ m) α) (s : σ) :
     StateT.run (restoreM x) s = pure x := by
   simp [restoreM, MonadControl.restoreM]
-  rfl
 
 @[simp] theorem run_liftWith [Monad m] [LawfulMonad m] (f : ({β : Type u} → StateT σ m β → m (stM m (StateT σ m) β)) → m α) (s : σ) :
     StateT.run (liftWith f) s = ((·, s) <$> f fun x => x.run s) := by

src/Init/Control/Lawful/MonadAttach/Instances.lean

@@ -15,7 +15,8 @@ public import Init.Ext
 public instance [Monad m] [LawfulMonad m] [MonadAttach m] [WeaklyLawfulMonadAttach m] :
     WeaklyLawfulMonadAttach (ReaderT ρ m) where
   map_attach := by
-    simp only [Functor.map, MonadAttach.attach, Functor.map_map, WeaklyLawfulMonadAttach.map_attach]
+    simp only [Functor.map, MonadAttach.attach, Functor.map_map, WeaklyLawfulMonadAttach.map_attach,
+      MonadAttach.CanReturn]
     intros; rfl
 
 public instance [Monad m] [LawfulMonad m] [MonadAttach m] [LawfulMonadAttach m] :
@@ -30,7 +31,7 @@ public instance [Monad m] [LawfulMonad m] [MonadAttach m] [WeaklyLawfulMonadAtta
   map_attach := by
     intro α x
     simp only [Functor.map, StateT, funext_iff, StateT.map, bind_pure_comp, MonadAttach.attach,
-      Functor.map_map]
+      Functor.map_map, MonadAttach.CanReturn]
     exact fun s => WeaklyLawfulMonadAttach.map_attach
 
 public instance [Monad m] [LawfulMonad m] [MonadAttach m] [LawfulMonadAttach m] :
@@ -45,7 +46,7 @@ public instance [Monad m] [LawfulMonad m] [MonadAttach m] [LawfulMonadAttach m]
 public instance [Monad m] [LawfulMonad m] [MonadAttach m] [WeaklyLawfulMonadAttach m] :
     WeaklyLawfulMonadAttach (ExceptT ε m) where
   map_attach {α} x := by
-    simp only [Functor.map, MonadAttach.attach, ExceptT.map]
+    simp only [Functor.map, MonadAttach.attach, ExceptT.map, MonadAttach.CanReturn]
     simp
     conv => rhs; rw [← WeaklyLawfulMonadAttach.map_attach (m := m) (x := x)]
     simp only [map_eq_pure_bind]
@@ -83,6 +84,6 @@ attribute [local instance] MonadAttach.trivial
 
 public instance [Monad m] [LawfulMonad m] :
     WeaklyLawfulMonadAttach m where
-  map_attach := by simp [MonadAttach.attach]
+  map_attach := by simp [MonadAttach.attach, MonadAttach.CanReturn]
 
 end

src/Init/Control/MonadAttach.lean

@@ -70,7 +70,7 @@ information to the return value, except a trivial proof of {name}`True`.
 This instance is used whenever no more useful {name}`MonadAttach` instance can be implemented.
 It always has a {name}`WeaklyLawfulMonadAttach`, but usually no {name}`LawfulMonadAttach` instance.
 -/
-@[expose, instance_reducible]
+@[expose, implicit_reducible]
 public protected def MonadAttach.trivial {m : Type u → Type v} [Monad m] : MonadAttach m where
   CanReturn _ _ := True
   attach x := (⟨·, .intro⟩) <$> x

src/Init/Core.lean

@@ -1339,10 +1339,10 @@ transitive and contains `r`. `TransGen r a z` if and only if there exists a sequ
 -/
 inductive Relation.TransGen {α : Sort u} (r : α → α → Prop) : α → α → Prop
   /-- If `r a b`, then `TransGen r a b`. This is the base case of the transitive closure. -/
-  | single {a b} : r a b → TransGen r a b
+  | single {a b : α} : r a b → TransGen r a b
   /-- If `TransGen r a b` and `r b c`, then `TransGen r a c`.
   This is the inductive case of the transitive closure. -/
-  | tail {a b c} : TransGen r a b → r b c → TransGen r a c
+  | tail {a b c : α} : TransGen r a b → r b c → TransGen r a c
 
 /-- The transitive closure is transitive. -/
 theorem Relation.TransGen.trans {α : Sort u} {r : α → α → Prop} {a b c} :
@@ -2313,6 +2313,13 @@ instance Pi.instSubsingleton {α : Sort u} {β : α → Sort v} [∀ a, Subsingl
 
 /-! # Squash -/
 
+theorem equivalence_true (α : Sort u) : Equivalence fun _ _ : α => True :=
+  ⟨fun _ => trivial, fun _ => trivial, fun _ _ => trivial⟩
+
+/-- Always-true relation as a `Setoid`. -/
+protected def Setoid.trivial (α : Sort u) : Setoid α :=
+  ⟨_, equivalence_true α⟩
+
 /--
 The quotient of `α` by the universal relation. The elements of `Squash α` are those of `α`, but all
 of them are equal and cannot be distinguished.
@@ -2326,8 +2333,11 @@ and its representation in compiled code is identical to that of `α`.
 
 Consequently, `Squash.lift` may extract an `α` value into any subsingleton type `β`, while
 `Nonempty.rec` can only do the same when `β` is a proposition.
+
+`Squash` is defined in terms of `Quotient`, so `Squash` can be used when a `Quotient` argument is
+expected.
 -/
-def Squash (α : Sort u) := Quot (fun (_ _ : α) => True)
+def Squash (α : Sort u) := Quotient (Setoid.trivial α)
 
 /--
 Places a value into its squash type, in which it cannot be distinguished from any other.
@@ -2583,3 +2593,11 @@ class Trichotomous (r : α → α → Prop) : Prop where
   trichotomous (a b : α) : ¬ r a b → ¬ r b a → a = b
 
 end Std
+
+@[simp] theorem flip_flip {α : Sort u} {β : Sort v} {φ : Sort w} {f : α → β → φ} :
+    flip (flip f) = f := by
+  apply funext
+  intro a
+  apply funext
+  intro b
+  rw [flip, flip]

src/Init/Data/Array.lean

@@ -34,3 +34,4 @@ public import Init.Data.Array.MinMax
 public import Init.Data.Array.Nat
 public import Init.Data.Array.Int
 public import Init.Data.Array.Count
+public import Init.Data.Array.Sort

src/Init/Data/Array/Basic.lean

@@ -6,6 +6,7 @@ Authors: Leonardo de Moura
 module
 
 prelude
+public import Init.Control.Do
 public import Init.GetElem
 public import Init.Data.List.ToArrayImpl
 import all Init.Data.List.ToArrayImpl
@@ -170,6 +171,15 @@ This avoids overhead due to unboxing a `Nat` used as an index.
 def uget (xs : @& Array α) (i : USize) (h : i.toNat < xs.size) : α :=
   xs[i.toNat]
 
+/--
+Version of `Array.uget` that does not increment the reference count of its result.
+
+This is only intended for direct use by the compiler.
+-/
+@[extern "lean_array_uget_borrowed"]
+unsafe opaque ugetBorrowed (xs : @& Array α) (i : USize) (h : i.toNat < xs.size) : α :=
+  xs.uget i h
+
 /--
 Low-level modification operator which is as fast as a C array write. The modification is performed
 in-place when the reference to the array is unique.
@@ -273,7 +283,7 @@ Examples:
  * `#[1, 2].isEmpty = false`
  * `#[()].isEmpty = false`
 -/
-@[expose]
+@[expose, inline]
 def isEmpty (xs : Array α) : Bool :=
   xs.size = 0
 
@@ -367,6 +377,7 @@ Returns the last element of an array, or panics if the array is empty.
 Safer alternatives include `Array.back`, which requires a proof the array is non-empty, and
 `Array.back?`, which returns an `Option`.
 -/
+@[inline]
 def back! [Inhabited α] (xs : Array α) : α :=
   xs[xs.size - 1]!
 
@@ -376,6 +387,7 @@ Returns the last element of an array, given a proof that the array is not empty.
 See `Array.back!` for the version that panics if the array is empty, or `Array.back?` for the
 version that returns an option.
 -/
+@[inline]
 def back (xs : Array α) (h : 0 < xs.size := by get_elem_tactic) : α :=
   xs[xs.size - 1]'(Nat.sub_one_lt_of_lt h)
 
@@ -385,6 +397,7 @@ Returns the last element of an array, or `none` if the array is empty.
 See `Array.back!` for the version that panics if the array is empty, or `Array.back` for the version
 that requires a proof the array is non-empty.
 -/
+@[inline]
 def back? (xs : Array α) : Option α :=
   xs[xs.size - 1]?
 
@@ -2125,7 +2138,7 @@ Examples:
 
 /-! ### Repr and ToString -/
 
-protected def Array.repr {α : Type u} [Repr α] (xs : Array α) : Std.Format :=
+protected def repr {α : Type u} [Repr α] (xs : Array α) : Std.Format :=
   let _ : Std.ToFormat α := ⟨repr⟩
   if xs.size == 0 then
     "#[]"

src/Init/Data/Array/Bootstrap.lean

@@ -52,7 +52,9 @@ theorem foldrM_eq_reverse_foldlM_toList.aux [Monad m]
   unfold foldrM.fold
   match i with
   | 0 => simp
-  | i+1 => rw [← List.take_concat_get h]; simp [← aux]
+  | i+1 =>
+    set_option backward.isDefEq.respectTransparency false in
+    rw [← List.take_concat_get h]; simp [← aux]
 
 theorem foldrM_eq_reverse_foldlM_toList [Monad m] {f : α → β → m β} {init : β} {xs : Array α} :
     xs.foldrM f init = xs.toList.reverse.foldlM (fun x y => f y x) init := by

src/Init/Data/Array/Count.lean

@@ -117,11 +117,13 @@ grind_pattern Std.Internal.Array.not_of_countP_eq_zero_of_mem => xs.countP p, x
 theorem countP_replicate {a : α} {n : Nat} : countP p (replicate n a) = if p a then n else 0 := by
   simp [← List.toArray_replicate, List.countP_replicate]
 
+set_option backward.isDefEq.respectTransparency false in
 theorem boole_getElem_le_countP {xs : Array α} {i : Nat} (h : i < xs.size) :
     (if p xs[i] then 1 else 0) ≤ xs.countP p := by
   rcases xs with ⟨xs⟩
   simp [List.boole_getElem_le_countP]
 
+set_option backward.isDefEq.respectTransparency false in
 @[grind =]
 theorem countP_set {xs : Array α} {i : Nat} {a : α} (h : i < xs.size) :
     (xs.set i a).countP p = xs.countP p - (if p xs[i] then 1 else 0) + (if p a then 1 else 0) := by

src/Init/Data/Array/DecidableEq.lean

@@ -76,7 +76,7 @@ theorem isEqv_eq_decide (xs ys : Array α) (r) :
     simpa [isEqv_iff_rel] using h'
 
 @[simp, grind =] theorem isEqv_toList [BEq α] (xs ys : Array α) : (xs.toList.isEqv ys.toList r) = (xs.isEqv ys r) := by
-  simp [isEqv_eq_decide, List.isEqv_eq_decide, Array.size]
+  simp [isEqv_eq_decide, List.isEqv_eq_decide, Array.size]; rfl
 
 theorem eq_of_isEqv [DecidableEq α] (xs ys : Array α) (h : Array.isEqv xs ys (fun x y => x = y)) : xs = ys := by
   have ⟨h, h'⟩ := rel_of_isEqv h
@@ -87,6 +87,7 @@ private theorem isEqvAux_self (r : α → α → Bool) (hr : ∀ a, r a a) (xs :
   induction i with
   | zero => simp [Array.isEqvAux]
   | succ i ih =>
+    set_option backward.isDefEq.respectTransparency false in
     simp_all only [isEqvAux, Bool.and_self]
 
 theorem isEqv_self_beq [BEq α] [ReflBEq α] (xs : Array α) : Array.isEqv xs xs (· == ·) = true := by
@@ -153,7 +154,7 @@ theorem beq_eq_decide [BEq α] (xs ys : Array α) :
   simp [BEq.beq, isEqv_eq_decide]
 
 @[simp, grind =] theorem beq_toList [BEq α] (xs ys : Array α) : (xs.toList == ys.toList) = (xs == ys) := by
-  simp [beq_eq_decide, List.beq_eq_decide, Array.size]
+  simp [beq_eq_decide, List.beq_eq_decide, Array.size]; rfl
 
 end Array
 

src/Init/Data/Array/Erase.lean

@@ -329,7 +329,7 @@ theorem eraseIdx_eq_take_drop_succ {xs : Array α} {i : Nat} (h) :
   rcases xs with ⟨xs⟩
   simp only [List.size_toArray] at h
   simp only [List.eraseIdx_toArray, List.eraseIdx_eq_take_drop_succ, take_eq_extract,
-    List.extract_toArray, List.extract_eq_drop_take, Nat.sub_zero, List.drop_zero, drop_eq_extract,
+    List.extract_toArray, List.extract_eq_take_drop, Nat.sub_zero, List.drop_zero, drop_eq_extract,
     List.size_toArray, List.append_toArray, mk.injEq, List.append_cancel_left_eq]
   rw [List.take_of_length_le]
   simp

src/Init/Data/Array/Find.lean

@@ -83,6 +83,10 @@ theorem findSome?_eq_some_iff {f : α → Option β} {xs : Array α} {b : β} :
   · rintro ⟨xs, a, ys, h₀, h₁, h₂⟩
     exact ⟨xs.toList, a, ys.toList, by simpa using congrArg toList h₀, h₁, by simpa⟩
 
+theorem isSome_findSome? {xs : Array α} {f : α → Option β} :
+    (xs.findSome? f).isSome = xs.any (f · |>.isSome) := by
+  simp [← findSome?_toList, List.isSome_findSome?]
+
 @[simp, grind =] theorem findSome?_guard {xs : Array α} : findSome? (Option.guard p) xs = find? p xs := by
   cases xs; simp
 
@@ -197,6 +201,10 @@ theorem find?_eq_some_iff_append {xs : Array α} :
     exact ⟨as.toList, ⟨l, by simpa using congrArg Array.toList h'⟩,
       by simpa using h⟩
 
+theorem isSome_find? {xs : Array α} {f : α → Bool} :
+    (xs.find? f).isSome = xs.any (f ·) := by
+  simp [← find?_toList, List.isSome_find?]
+
 theorem find?_push {xs : Array α} : (xs.push a).find? p = (xs.find? p).or (if p a then some a else none) := by
   cases xs; simp
 
@@ -425,6 +433,7 @@ theorem lt_findIdx_of_not {p : α → Bool} {xs : Array α} {i : Nat} (h : i < x
   simp only [Nat.not_lt] at f
   exact absurd (@findIdx_getElem _ p xs (Nat.lt_of_le_of_lt f h)) (h2 (xs.findIdx p) f)
 
+set_option backward.isDefEq.respectTransparency false in
 /-- `xs.findIdx p = i` iff `p xs[i]` and `¬ p xs [j]` for all `j < i`. -/
 theorem findIdx_eq {p : α → Bool} {xs : Array α} {i : Nat} (h : i < xs.size) :
     xs.findIdx p = i ↔ p xs[i] ∧ ∀ j (hji : j < i), p (xs[j]'(Nat.lt_trans hji h)) = false := by
@@ -613,12 +622,12 @@ theorem findIdx?_eq_some_le_of_findIdx?_eq_some {xs : Array α} {p q : α → Bo
 /-! ### findFinIdx? -/
 
 @[grind =]
-theorem findFinIdx?_empty {p : α → Bool} : findFinIdx? p #[] = none := by simp
+theorem findFinIdx?_empty {p : α → Bool} : findFinIdx? p #[] = none := by simp; rfl
 
 @[grind =]
 theorem findFinIdx?_singleton {a : α} {p : α → Bool} :
     #[a].findFinIdx? p = if p a then some ⟨0, by simp⟩ else none := by
-  simp
+  simp; rfl
 
 -- We can't mark this as a `@[congr]` lemma since the head of the RHS is not `findFinIdx?`.
 theorem findFinIdx?_congr {p : α → Bool} {xs ys : Array α} (w : xs = ys) :
@@ -792,7 +801,7 @@ theorem idxOf?_eq_map_finIdxOf?_val [BEq α] {xs : Array α} {a : α} :
     xs.idxOf? a = (xs.finIdxOf? a).map (·.val) := by
   simp [idxOf?, finIdxOf?]
 
-@[grind =] theorem finIdxOf?_empty [BEq α] : (#[] : Array α).finIdxOf? a = none := by simp
+@[grind =] theorem finIdxOf?_empty [BEq α] : (#[] : Array α).finIdxOf? a = none := by simp; rfl
 
 @[simp, grind =] theorem finIdxOf?_eq_none_iff [BEq α] [LawfulBEq α] {xs : Array α} {a : α} :
     xs.finIdxOf? a = none ↔ a ∉ xs := by

src/Init/Data/Array/Lemmas.lean

@@ -72,6 +72,9 @@ theorem toArray_eq : List.toArray as = xs ↔ as = xs.toList := by
 
 /-! ### size -/
 
+theorem size_singleton {x : α} : #[x].size = 1 := by
+  simp
+
 theorem eq_empty_of_size_eq_zero (h : xs.size = 0) : xs = #[] := by
   cases xs
   simp_all
@@ -170,6 +173,7 @@ theorem getD_getElem? {xs : Array α} {i : Nat} {d : α} :
 
 @[simp] theorem getElem?_empty {i : Nat} : (#[] : Array α)[i]? = none := rfl
 
+set_option backward.isDefEq.respectTransparency false in
 theorem getElem_push_lt {xs : Array α} {x : α} {i : Nat} (h : i < xs.size) :
     have : i < (xs.push x).size := by simp [*, Nat.lt_succ_of_le, Nat.le_of_lt]
     (xs.push x)[i] = xs[i] := by
@@ -3482,6 +3486,21 @@ theorem foldl_eq_foldr_reverse {xs : Array α} {f : β → α → β} {b} :
 theorem foldr_eq_foldl_reverse {xs : Array α} {f : α → β → β} {b} :
     xs.foldr f b = xs.reverse.foldl (fun x y => f y x) b := by simp
 
+theorem foldl_eq_apply_foldr {xs : Array α} {f : α → α → α}
+    [Std.Associative f] [Std.LawfulRightIdentity f init] :
+    xs.foldl f x = f x (xs.foldr f init) := by
+  simp [← foldl_toList, ← foldr_toList, List.foldl_eq_apply_foldr]
+
+theorem foldr_eq_apply_foldl {xs : Array α} {f : α → α → α}
+    [Std.Associative f] [Std.LawfulLeftIdentity f init] :
+    xs.foldr f x = f (xs.foldl f init) x := by
+  simp [← foldl_toList, ← foldr_toList, List.foldr_eq_apply_foldl]
+
+theorem foldr_eq_foldl {xs : Array α} {f : α → α → α}
+    [Std.Associative f] [Std.LawfulIdentity f init] :
+    xs.foldr f init = xs.foldl f init := by
+  simp [foldl_eq_apply_foldr, Std.LawfulLeftIdentity.left_id]
+
 @[simp] theorem foldr_push_eq_append {as : Array α} {bs : Array β} {f : α → β} (w : start = as.size) :
     as.foldr (fun a xs => Array.push xs (f a)) bs start 0 = bs ++ (as.map f).reverse := by
   subst w
@@ -3974,6 +3993,7 @@ theorem all_filterMap {xs : Array α} {f : α → Option β} {p : β → Bool} :
   · simp only [Id.run_pure]
     rw [if_neg (mt (by rintro rfl; exact h) (by simp_all))]
 
+set_option backward.isDefEq.respectTransparency false in
 @[simp, grind =] theorem toList_modify {xs : Array α} {f : α → α} {i : Nat} :
     (xs.modify i f).toList = xs.toList.modify i f := by
   apply List.ext_getElem
@@ -4146,7 +4166,7 @@ variable [LawfulBEq α]
     (xs.replace a b)[i]? = if xs[i]? == some a then if a ∈ xs.take i then some a else some b else xs[i]? := by
   rcases xs with ⟨xs⟩
   simp only [List.replace_toArray, List.getElem?_toArray, List.getElem?_replace, take_eq_extract,
-    List.extract_toArray, List.extract_eq_drop_take, Nat.sub_zero, List.drop_zero, List.mem_toArray]
+    List.extract_toArray, List.extract_eq_take_drop, Nat.sub_zero, List.drop_zero, List.mem_toArray]
 
 theorem getElem?_replace_of_ne {xs : Array α} {i : Nat} (h : xs[i]? ≠ some a) :
     (xs.replace a b)[i]? = xs[i]? := by
@@ -4259,6 +4279,7 @@ private theorem getElem_ofFn_go {f : Fin n → α} {acc i k} (h : i ≤ n) (w₁
     · simp
       omega
 
+set_option backward.isDefEq.respectTransparency false in
 @[simp] theorem getElem_ofFn {f : Fin n → α} {i : Nat} (h : i < (ofFn f).size) :
     (ofFn f)[i] = f ⟨i, size_ofFn (f := f) ▸ h⟩ := by
   unfold ofFn
@@ -4332,16 +4353,33 @@ def sum_eq_sum_toList := @sum_toList
 
 @[simp, grind =]
 theorem sum_append [Zero α] [Add α] [Std.Associative (α := α) (· + ·)]
-    [Std.LeftIdentity (α := α) (· + ·) 0] [Std.LawfulLeftIdentity (α := α) (· + ·) 0]
+    [Std.LawfulLeftIdentity (α := α) (· + ·) 0]
     {as₁ as₂ : Array α} : (as₁ ++ as₂).sum = as₁.sum + as₂.sum := by
   simp [← sum_toList, List.sum_append]
 
+@[simp, grind =]
+theorem sum_singleton [Add α] [Zero α] [Std.LawfulRightIdentity (· + ·) (0 : α)] {x : α} :
+    #[x].sum = x := by
+  simp [Array.sum_eq_foldr, Std.LawfulRightIdentity.right_id x]
+
+@[simp, grind =]
+theorem sum_push [Add α] [Zero α] [Std.Associative (α := α) (· + ·)]
+    [Std.LawfulIdentity (· + ·) (0 : α)] {xs : Array α} {x : α} :
+    (xs.push x).sum = xs.sum + x := by
+  simp [Array.sum_eq_foldr, Std.LawfulRightIdentity.right_id, Std.LawfulLeftIdentity.left_id,
+    ← Array.foldr_assoc]
+
 @[simp, grind =]
 theorem sum_reverse [Zero α] [Add α] [Std.Associative (α := α) (· + ·)]
     [Std.Commutative (α := α) (· + ·)]
     [Std.LawfulLeftIdentity (α := α) (· + ·) 0] (xs : Array α) : xs.reverse.sum = xs.sum := by
   simp [← sum_toList, List.sum_reverse]
 
+theorem sum_eq_foldl [Zero α] [Add α] [Std.Associative (α := α) (· + ·)]
+    [Std.LawfulIdentity (· + ·) (0 : α)] {xs : Array α} :
+    xs.sum = xs.foldl (init := 0) (· + ·) := by
+  simp [← sum_toList, List.sum_eq_foldl]
+
 theorem foldl_toList_eq_flatMap {l : List α} {acc : Array β}
     {F : Array β → α → Array β} {G : α → List β}
     (H : ∀ acc a, (F acc a).toList = acc.toList ++ G a) :
@@ -4490,11 +4528,13 @@ theorem getElem?_push_eq {xs : Array α} {x : α} : (xs.push x)[xs.size]? = some
   cases xs
   simp
 
+set_option backward.isDefEq.respectTransparency false in
 @[simp, grind =] theorem finIdxOf?_toList [BEq α] {a : α} {xs : Array α} :
     xs.toList.finIdxOf? a = (xs.finIdxOf? a).map (Fin.cast (by simp)) := by
   cases xs
   simp
 
+set_option backward.isDefEq.respectTransparency false in
 @[simp, grind =] theorem findFinIdx?_toList {p : α → Bool} {xs : Array α} :
     xs.toList.findFinIdx? p = (xs.findFinIdx? p).map (Fin.cast (by simp)) := by
   cases xs
@@ -4619,6 +4659,7 @@ namespace List
     as.toArray.unzip = Prod.map List.toArray List.toArray as.unzip := by
   ext1 <;> simp
 
+set_option backward.isDefEq.respectTransparency false in
 @[simp, grind =] theorem firstM_toArray [Alternative m] {as : List α} {f : α → m β} :
     as.toArray.firstM f = as.firstM f := by
   unfold Array.firstM

src/Init/Data/Array/MinMax.lean

@@ -89,7 +89,7 @@ public theorem _root_.List.min_toArray [Min α] {l : List α} {h} :
     · rename_i x xs
       simp only [List.getElem_toArray, List.getElem_cons_zero, List.size_toArray, List.length_cons]
       rw [List.toArray_cons, foldl_eq_foldl_extract]
-      rw [← Array.foldl_toList, Array.toList_extract, List.extract_eq_drop_take]
+      rw [← Array.foldl_toList, Array.toList_extract, List.extract_eq_take_drop]
       simp [List.min]
 
 public theorem _root_.List.min_eq_min_toArray [Min α] {l : List α} {h} :
@@ -129,7 +129,7 @@ public theorem _root_.List.max_toArray [Max α] {l : List α} {h} :
     · rename_i x xs
       simp only [List.getElem_toArray, List.getElem_cons_zero, List.size_toArray, List.length_cons]
       rw [List.toArray_cons, foldl_eq_foldl_extract]
-      rw [← Array.foldl_toList, Array.toList_extract, List.extract_eq_drop_take]
+      rw [← Array.foldl_toList, Array.toList_extract, List.extract_eq_take_drop]
       simp [List.max]
 
 public theorem _root_.List.max_eq_max_toArray [Max α] {l : List α} {h} :

src/Init/Data/Array/Perm.lean

@@ -126,6 +126,14 @@ theorem swap_perm {xs : Array α} {i j : Nat} (h₁ : i < xs.size) (h₂ : j < x
   simp only [swap, perm_iff_toList_perm, toList_set]
   apply set_set_perm
 
+theorem Perm.pairwise_iff {R : α → α → Prop} (S : ∀ {x y}, R x y → R y x) {xs ys : Array α}
+    : ∀ _p : xs.Perm ys, xs.toList.Pairwise R ↔ ys.toList.Pairwise R := by
+  simpa only [perm_iff_toList_perm] using List.Perm.pairwise_iff S
+
+theorem Perm.pairwise {R : α → α → Prop} {xs ys : Array α} (hp : xs ~ ys)
+    (hR : xs.toList.Pairwise R) (hsymm : ∀ {x y}, R x y → R y x) :
+    ys.toList.Pairwise R := (hp.pairwise_iff hsymm).mp hR
+
 namespace Perm
 
 set_option linter.indexVariables false in
@@ -135,7 +143,7 @@ theorem extract {xs ys : Array α} (h : xs ~ ys) {lo hi : Nat}
   rcases xs with ⟨xs⟩
   rcases ys with ⟨ys⟩
   simp_all only [perm_iff_toList_perm, List.getElem?_toArray, List.extract_toArray,
-    List.extract_eq_drop_take]
+    List.extract_eq_take_drop]
   apply List.Perm.take_of_getElem? (w := fun i h => by simpa using whi (lo + i) (by omega))
   apply List.Perm.drop_of_getElem? (w := wlo)
   exact h

src/Init/Data/Array/Sort.lean

@@ -0,0 +1,10 @@
+/-
+Copyright (c) 2026 Lean FRO. All rights reserved.
+Released under Apache 2.0 license as described in the file LICENSE.
+Authors: Paul Reichert
+-/
+module
+
+prelude
+public import Init.Data.Array.Sort.Basic
+public import Init.Data.Array.Sort.Lemmas

src/Init/Data/Array/Sort/Basic.lean

@@ -0,0 +1,55 @@
+/-
+Copyright (c) 2026 Lean FRO. All rights reserved.
+Released under Apache 2.0 license as described in the file LICENSE.
+Authors: Paul Reichert
+-/
+module
+
+prelude
+public import Init.Data.Array.Subarray.Split
+public import Init.Data.Slice.Array
+import Init.Omega
+
+public section
+
+private def Array.MergeSort.Internal.merge (xs ys : Array α) (le : α → α → Bool := by exact (· ≤ ·)) :
+    Array α :=
+  if hxs : 0 < xs.size then
+    if hys : 0 < ys.size then
+      go xs[*...*] ys[*...*] (by simp only [Array.size_mkSlice_rii]; omega) (by simp only [Array.size_mkSlice_rii]; omega) (Array.emptyWithCapacity (xs.size + ys.size))
+    else
+      xs
+  else
+    ys
+where
+  go (xs ys : Subarray α) (hxs : 0 < xs.size) (hys : 0 < ys.size) (acc : Array α) : Array α :=
+    let x := xs[0]
+    let y := ys[0]
+    if le x y then
+      if hi : 1 < xs.size then
+        go (xs.drop 1) ys (by simp only [Subarray.size_drop]; omega) hys (acc.push x)
+      else
+        ys.foldl (init := acc.push x) (fun acc y => acc.push y)
+    else
+      if hj : 1 < ys.size then
+        go xs (ys.drop 1) hxs (by simp only [Subarray.size_drop]; omega) (acc.push y)
+      else
+        xs.foldl (init := acc.push y) (fun acc x => acc.push x)
+  termination_by xs.size + ys.size
+
+def Subarray.mergeSort (xs : Subarray α) (le : α → α → Bool := by exact (· ≤ ·)) : Array α :=
+    if h : 1 < xs.size then
+      let splitIdx := (xs.size + 1) / 2 -- We follow the same splitting convention as `List.mergeSort`
+      let left := xs[*...splitIdx]
+      let right := xs[splitIdx...*]
+      Array.MergeSort.Internal.merge (mergeSort left le) (mergeSort right le) le
+    else
+      xs.toArray
+termination_by xs.size
+decreasing_by
+  · simp only [Subarray.size_mkSlice_rio]; omega
+  · simp only [Subarray.size_mkSlice_rci]; omega
+
+@[inline]
+def Array.mergeSort (xs : Array α) (le : α → α → Bool := by exact (· ≤ ·)) : Array α :=
+    xs[*...*].mergeSort le

src/Init/Data/Array/Sort/Lemmas.lean

@@ -0,0 +1,240 @@
+/-
+Copyright (c) 2026 Lean FRO. All rights reserved.
+Released under Apache 2.0 license as described in the file LICENSE.
+Authors: Paul Reichert
+-/
+module
+
+prelude
+public import Init.Data.Array.Sort.Basic
+public import Init.Data.List.Sort.Basic
+public import Init.Data.Array.Perm
+import all Init.Data.Array.Sort.Basic
+import all Init.Data.List.Sort.Basic
+import Init.Data.List.Sort.Lemmas
+import Init.Data.Slice.Array.Lemmas
+import Init.Data.Slice.List.Lemmas
+import Init.Data.Array.Bootstrap
+import Init.Data.Array.Lemmas
+import Init.Data.Array.MapIdx
+import Init.ByCases
+
+public section
+
+private theorem Array.MergeSort.merge.go_eq_listMerge {xs ys : Subarray α} {hxs hys le acc} :
+    (Array.MergeSort.Internal.merge.go le xs ys hxs hys acc).toList = acc.toList ++ List.merge xs.toList ys.toList le := by
+  fun_induction Array.MergeSort.Internal.merge.go le xs ys hxs hys acc
+  · rename_i xs ys _ _ _ _ _ _ _ _
+    rw [List.merge.eq_def]
+    split
+    · have : xs.size = 0 := by simp [← Subarray.length_toList, *]
+      omega
+    · have : ys.size = 0 := by simp [← Subarray.length_toList, *]
+      omega
+    · rename_i x' xs' y' ys' _ _
+      simp +zetaDelta only at *
+      have h₁ : x' = xs[0] := by simp [Subarray.getElem_eq_getElem_toList, *]
+      have h₂ : y' = ys[0] := by simp [Subarray.getElem_eq_getElem_toList, *]
+      cases h₁
+      cases h₂
+      simp [Subarray.toList_drop, *]
+  · rename_i xs ys _ _ _ _ _ _ _
+    rw [List.merge.eq_def]
+    split
+    · have : xs.size = 0 := by simp [← Subarray.length_toList, *]
+      omega
+    · have : ys.size = 0 := by simp [← Subarray.length_toList, *]
+      omega
+    · rename_i x' xs' y' ys' _ _
+      simp +zetaDelta only at *
+      have h₁ : x' = xs[0] := by simp [Subarray.getElem_eq_getElem_toList, *]
+      have h₂ : y' = ys[0] := by simp [Subarray.getElem_eq_getElem_toList, *]
+      cases h₁
+      cases h₂
+      simp [*]
+      have : xs.size = xs'.length + 1 := by simp [← Subarray.length_toList, *]
+      have : xs' = [] := List.eq_nil_of_length_eq_zero (by omega)
+      simp only [this]
+      rw [← Subarray.foldl_toList]
+      simp [*]
+  · rename_i xs ys _ _ _ _ _ _ _ _
+    rw [List.merge.eq_def]
+    split
+    · have : xs.size = 0 := by simp [← Subarray.length_toList, *]
+      omega
+    · have : ys.size = 0 := by simp [← Subarray.length_toList, *]
+      omega
+    · rename_i x' xs' y' ys' _ _
+      simp +zetaDelta only at *
+      have h₁ : x' = xs[0] := by simp [Subarray.getElem_eq_getElem_toList, *]
+      have h₂ : y' = ys[0] := by simp [Subarray.getElem_eq_getElem_toList, *]
+      cases h₁
+      cases h₂
+      simp [Subarray.toList_drop, *]
+  · rename_i xs ys _ _ _ _ _ _ _
+    rw [List.merge.eq_def]
+    split
+    · have : xs.size = 0 := by simp [← Subarray.length_toList, *]
+      omega
+    · have : ys.size = 0 := by simp [← Subarray.length_toList, *]
+      omega
+    · rename_i x' xs' y' ys' _ _
+      simp +zetaDelta only at *
+      have h₁ : x' = xs[0] := by simp [Subarray.getElem_eq_getElem_toList, *]
+      have h₂ : y' = ys[0] := by simp [Subarray.getElem_eq_getElem_toList, *]
+      cases h₁
+      cases h₂
+      simp [*]
+      have : ys.size = ys'.length + 1 := by simp [← Subarray.length_toList, *]
+      have : ys' = [] := List.eq_nil_of_length_eq_zero (by omega)
+      simp [this]
+      rw [← Subarray.foldl_toList]
+      simp [*]
+
+private theorem Array.MergeSort.merge_eq_listMerge {xs ys : Array α} {le} :
+    (Array.MergeSort.Internal.merge xs ys le).toList = List.merge xs.toList ys.toList le := by
+  rw [Array.MergeSort.Internal.merge]
+  split <;> rename_i heq₁
+  · split <;> rename_i heq₂
+    · simp [Array.MergeSort.merge.go_eq_listMerge]
+    · have : ys.toList = [] := by simp_all
+      simp [this]
+  · have : xs.toList = [] := by simp_all
+    simp [this]
+
+private theorem List.mergeSort_eq_merge_mkSlice {xs : List α} :
+    xs.mergeSort le =
+      if 1 < xs.length then
+        merge (xs[*...((xs.length + 1) / 2)].toList.mergeSort le) (xs[((xs.length + 1) / 2)...*].toList.mergeSort le) le
+      else
+        xs := by
+  fun_cases xs.mergeSort le
+  · simp
+  · simp
+  · rename_i x y ys lr hl hr
+    simp [lr]
+
+theorem Subarray.toList_mergeSort {xs : Subarray α} {le : α → α → Bool} :
+    (xs.mergeSort le).toList = xs.toList.mergeSort le := by
+  fun_induction xs.mergeSort le
+  · rw [List.mergeSort_eq_merge_mkSlice]
+    simp +zetaDelta [Array.MergeSort.merge_eq_listMerge, *]
+  · simp [List.mergeSort_eq_merge_mkSlice, *]
+
+@[simp, grind =]
+theorem Subarray.mergeSort_eq_mergeSort_toArray {xs : Subarray α} {le : α → α → Bool} :
+    xs.mergeSort le = xs.toArray.mergeSort le := by
+  simp [← Array.toList_inj, toList_mergeSort, Array.mergeSort]
+
+theorem Subarray.mergeSort_toArray {xs : Subarray α} {le : α → α → Bool} :
+    xs.toArray.mergeSort le = xs.mergeSort le := by
+  simp
+
+theorem Array.toList_mergeSort {xs : Array α} {le : α → α → Bool} :
+    (xs.mergeSort le).toList = xs.toList.mergeSort le := by
+  rw [Array.mergeSort, Subarray.toList_mergeSort, Array.toList_mkSlice_rii]
+
+theorem Array.mergeSort_eq_toArray_mergeSort_toList {xs : Array α} {le : α → α → Bool} :
+    xs.mergeSort le = (xs.toList.mergeSort le).toArray := by
+  simp [← toList_mergeSort]
+
+/-!
+# Basic properties of `Array.mergeSort`.
+
+* `pairwise_mergeSort`: `mergeSort` produces a sorted array.
+* `mergeSort_perm`: `mergeSort` is a permutation of the input array.
+* `mergeSort_of_pairwise`: `mergeSort` does not change a sorted array.
+* `sublist_mergeSort`: if `c` is a sorted sublist of `l`, then `c` is still a sublist of `mergeSort le l`.
+-/
+
+namespace Array
+
+-- Enable this instance locally so we can write `Pairwise le` instead of `Pairwise (le · ·)` everywhere.
+attribute [local instance] boolRelToRel
+
+@[simp] theorem mergeSort_empty : (#[] : Array α).mergeSort r = #[] := by
+  simp [mergeSort_eq_toArray_mergeSort_toList]
+
+@[simp] theorem mergeSort_singleton {a : α} : #[a].mergeSort r = #[a] := by
+  simp [mergeSort_eq_toArray_mergeSort_toList]
+
+theorem mergeSort_perm {xs : Array α} {le} : (xs.mergeSort le).Perm xs := by
+  simpa [mergeSort_eq_toArray_mergeSort_toList, Array.perm_iff_toList_perm] using List.mergeSort_perm _ _
+
+@[simp] theorem size_mergeSort {xs : Array α} : (mergeSort xs le).size = xs.size := by
+  simp [mergeSort_eq_toArray_mergeSort_toList]
+
+@[simp] theorem mem_mergeSort {a : α} {xs : Array α} : a ∈ mergeSort xs le ↔ a ∈ xs := by
+  simp [mergeSort_eq_toArray_mergeSort_toList]
+
+/--
+The result of `Array.mergeSort` is sorted,
+as long as the comparison function is transitive (`le a b → le b c → le a c`)
+and total in the sense that `le a b || le b a`.
+
+The comparison function need not be irreflexive, i.e. `le a b` and `le b a` is allowed even when `a ≠ b`.
+-/
+theorem pairwise_mergeSort
+    (trans : ∀ (a b c : α), le a b → le b c → le a c)
+    (total : ∀ (a b : α), le a b || le b a)
+    {xs : Array α} :
+    (mergeSort xs le).toList.Pairwise (le · ·) := by
+  simpa [mergeSort_eq_toArray_mergeSort_toList] using List.pairwise_mergeSort trans total _
+
+/--
+If the input array is already sorted, then `mergeSort` does not change the array.
+-/
+theorem mergeSort_of_pairwise {le : α → α → Bool} {xs : Array α} (_ : xs.toList.Pairwise (le · ·)) :
+    mergeSort xs le = xs := by
+  simpa [mergeSort_eq_toArray_mergeSort_toList, List.toArray_eq_iff] using List.mergeSort_of_pairwise ‹_›
+
+/--
+This merge sort algorithm is stable,
+in the sense that breaking ties in the ordering function using the position in the array
+has no effect on the output.
+
+That is, elements which are equal with respect to the ordering function will remain
+in the same order in the output array as they were in the input array.
+
+See also:
+* `sublist_mergeSort`: if `c <+ l` and `c.Pairwise le`, then `c <+ (mergeSort le l).toList`.
+* `pair_sublist_mergeSort`: if `[a, b] <+ l` and `le a b`, then `[a, b] <+ (mergeSort le l).toList`)
+-/
+theorem mergeSort_zipIdx {xs : Array α} :
+    (mergeSort (xs.zipIdx.map fun (a, i) => (a, i)) (List.zipIdxLE le)).map (·.1) = mergeSort xs le := by
+  simpa [mergeSort_eq_toArray_mergeSort_toList, Array.toList_zipIdx] using List.mergeSort_zipIdx
+
+/--
+Another statement of stability of merge sort.
+If `c` is a sorted sublist of `xs.toList`,
+then `c` is still a sublist of `(mergeSort le xs).toList`.
+-/
+theorem sublist_mergeSort {le : α → α → Bool}
+    (trans : ∀ (a b c : α), le a b → le b c → le a c)
+    (total : ∀ (a b : α), le a b || le b a)
+    {ys : List α} (_ : ys.Pairwise (le · ·)) (_ : List.Sublist ys xs.toList) :
+    List.Sublist ys (mergeSort xs le).toList := by
+  simpa [mergeSort_eq_toArray_mergeSort_toList, Array.toList_zipIdx] using
+    List.sublist_mergeSort trans total ‹_› ‹_›
+
+/--
+Another statement of stability of merge sort.
+If a pair `[a, b]` is a sublist of `xs.toList` and `le a b`,
+then `[a, b]` is still a sublist of `(mergeSort le xs).toList`.
+-/
+theorem pair_sublist_mergeSort
+    (trans : ∀ (a b c : α), le a b → le b c → le a c)
+    (total : ∀ (a b : α), le a b || le b a)
+    (hab : le a b) (h : List.Sublist [a, b] xs.toList) :
+    List.Sublist [a, b] (mergeSort xs le).toList := by
+  simpa [mergeSort_eq_toArray_mergeSort_toList, Array.toList_zipIdx] using
+    List.pair_sublist_mergeSort trans total ‹_› ‹_›
+
+theorem map_mergeSort {r : α → α → Bool} {s : β → β → Bool} {f : α → β}
+    {xs : Array α} (hxs : ∀ a ∈ xs, ∀ b ∈ xs, r a b = s (f a) (f b)) :
+    (xs.mergeSort r).map f = (xs.map f).mergeSort s := by
+  simp only [mergeSort_eq_toArray_mergeSort_toList, List.map_toArray, toList_map, mk.injEq]
+  apply List.map_mergeSort
+  simpa
+
+end Array

src/Init/Data/Array/Subarray.lean

@@ -7,7 +7,7 @@ module
 
 prelude
 public import Init.Data.Array.Basic
-public import Init.Data.Slice.Basic
+public import Init.Data.Slice.Operations
 
 public section
 
@@ -76,15 +76,17 @@ def Subarray.stop_le_array_size (xs : Subarray α) : xs.stop ≤ xs.array.size :
 
 namespace Subarray
 
-/--
-Computes the size of the subarray.
--/
-def size (s : Subarray α) : Nat :=
-  s.stop - s.start
+instance : SliceSize (Internal.SubarrayData α) where
+  size s := s.internalRepresentation.stop - s.internalRepresentation.start
+
+@[grind =, suggest_for Subarray.size]
+public theorem size_eq {xs : Subarray α} :
+    xs.size = xs.stop - xs.start := by
+  simp [Std.Slice.size, SliceSize.size, start, stop]
 
 theorem size_le_array_size {s : Subarray α} : s.size ≤ s.array.size := by
   let ⟨{array, start, stop, start_le_stop, stop_le_array_size}⟩ := s
-  simp only [size, ge_iff_le]
+  simp only [ge_iff_le, size_eq]
   apply Nat.le_trans (Nat.sub_le stop start)
   assumption
 

src/Init/Data/BitVec/Basic.lean

@@ -210,7 +210,7 @@ protected def toHex {n : Nat} (x : BitVec n) : String :=
   String.Internal.append t s
 
 /-- `BitVec` representation. -/
-protected def BitVec.repr (a : BitVec n) : Std.Format :=
+protected def repr (a : BitVec n) : Std.Format :=
   "0x" ++ (a.toHex : Std.Format) ++ "#" ++ repr n
 
 instance : Repr (BitVec n) where

src/Init/Data/BitVec/Bitblast.lean

@@ -2192,6 +2192,7 @@ def uppcRec {w} (x : BitVec w) (s : Nat) (hs : s < w) : Bool :=
   | 0 => x.msb
   | i + 1 =>  x[w - 1 - i] || uppcRec x i (by omega)
 
+set_option backward.isDefEq.respectTransparency false in
 /-- The unsigned parallel prefix of `x` at `s` is `true` if and only if x interpreted
   as a natural number is greater or equal than `2 ^ (w - 1 - (s - 1))`. -/
 @[simp]
@@ -2392,4 +2393,412 @@ theorem fastUmulOverflow (x y : BitVec w) :
         simp [← Nat.pow_add, show w + 1 - (k - 1) + k = w + 1 + 1 by omega] at this
         omega
 
+/-! ### Population Count -/
+
+/-- Extract the `k`-th bit from `x` and extend it to have length `len`. -/
+def extractAndExtendBit (idx len : Nat) (x : BitVec w) : BitVec len :=
+  BitVec.zeroExtend len (BitVec.extractLsb' idx 1 x)
+
+
+/-- Recursively extract one bit at a time and extend it to width `w` -/
+def extractAndExtendAux (k len : Nat) (x : BitVec w) (acc : BitVec (k * len)) (hle : k ≤ w) :
+    BitVec (w * len) :=
+  match hwi : w - k with
+  | 0 => acc.cast (by simp [show w = k by omega])
+  | n' + 1 =>
+    let acc' := extractAndExtendBit k len x ++ acc
+    extractAndExtendAux (k + 1) len x (acc'.cast (by simp [Nat.add_mul]; omega)) (by omega)
+termination_by w - k
+
+/-- We instantiate `extractAndExtendAux` to extend each bit to `len`, extending
+  each bit in `x` to have width `w` and returning a `BitVec (w * w)`. -/
+def extractAndExtend (len : Nat) (x : BitVec w) : BitVec (w * len) :=
+  extractAndExtendAux 0 len x ((0#0).cast (by simp)) (by omega)
+
+/--
+  Construct a layer of the parallel-prefix-sum tree by summing two-by-two all the
+  `w`-long words in `oldLayer`, returning a bitvector containing `(oldLen + 1) / 2`
+  flattened `w`-long words, each resulting from an addition.
+-/
+def cpopLayer (oldLayer : BitVec (len * w)) (newLayer : BitVec (iterNum * w))
+    (hold : 2 * (iterNum - 1) < len) : BitVec (((len + 1)/2) * w) :=
+  if hlen : len - (iterNum * 2) = 0 then
+    have : ((len + 1)/2) = iterNum := by omega
+    newLayer.cast (by simp [this])
+  else
+    let op1 := oldLayer.extractLsb' ((2 * iterNum) * w) w
+    let op2 := oldLayer.extractLsb' ((2 * iterNum + 1) * w) w
+    let newLayer' := (op1 + op2) ++ newLayer
+    have hcast : w + iterNum * w = (iterNum + 1) * w := by simp [Nat.add_mul]; omega
+    cpopLayer oldLayer (newLayer'.cast hcast) (by omega)
+termination_by len - (iterNum * 2)
+
+/--
+  Given a `BitVec (len * w)` of `len` flattened `w`-long words,
+  construct a binary tree that sums two-by-two the `w`-long words in the previous layer,
+  ultimately returning a single `w`-long words corresponding to the whole addition.
+-/
+def cpopTree (l : BitVec (len * w)) : BitVec w :=
+  if h : len = 0 then 0#w
+  else if h : len = 1 then
+    l.cast (by simp [h])
+  else
+    cpopTree (cpopLayer l 0#(0 * w) (by omega))
+termination_by len
+
+/--
+  Given flattened bitvector `x : BitVec w` and a length `l : Nat`,
+  construct a parallel prefix sum circuit adding each available `l`-long word in `x`.
+-/
+def cpopRec (x : BitVec w) : BitVec w :=
+  if hw : 1 < w then
+    let extendedBits := x.extractAndExtend w
+    (cpopTree extendedBits).cast (by simp)
+  else if hw' : 0 < w then
+    x
+  else
+    0#w
+
+/-- Recursive addition of the elements in a flattened bitvec, starting from the `rem`-th element. -/
+private def addRecAux (x : BitVec (l * w)) (rem : Nat) (acc : BitVec w) : BitVec w :=
+  match rem with
+  | 0 => acc
+  | n + 1 => x.addRecAux n (acc + x.extractLsb' (n * w) w)
+
+/-- Recursive addition of the elements in a flattened bitvec. -/
+private def addRec (x : BitVec (l * w)) : BitVec w := addRecAux x l 0#w
+
+theorem getLsbD_extractAndExtendBit {x : BitVec w} :
+    (extractAndExtendBit k len x).getLsbD i =
+    (decide (i = 0) && decide (0 < len) && x.getLsbD k) := by
+  simp only [extractAndExtendBit, truncate_eq_setWidth, getLsbD_setWidth, getLsbD_extractLsb',
+    Nat.lt_one_iff]
+  by_cases hi : i = 0
+  <;> simp [hi]
+
+@[simp]
+private theorem extractAndExtendAux_zero {k len : Nat} {x : BitVec w}
+    {acc : BitVec (k * len)} (heq : w = k) :
+    extractAndExtendAux k len x acc (by omega) = acc.cast (by simp [heq]) := by
+  unfold extractAndExtendAux
+  split
+  · simp
+  · omega
+
+private theorem extractLsb'_extractAndExtendAux {k len : Nat} {x : BitVec w}
+    (acc : BitVec (k * len)) (hle : k ≤ w) :
+    (∀ i (_ : i < k), acc.extractLsb' (i * len) len = (x.extractLsb' i 1).setWidth len) →
+    (extractAndExtendAux k len x acc (by omega)).extractLsb' (i * len) len =
+    (x.extractLsb' i 1).setWidth len := by
+  intros hacc
+  induction hwi : w - k generalizing acc k
+  · case zero =>
+    rw [extractAndExtendAux_zero (by omega)]
+    by_cases hj : i < k
+    · apply hacc
+      exact hj
+    · ext l hl
+      have := mul_le_mul_right (n := k) (m := i) len (by omega)
+      simp [← getLsbD_eq_getElem, getLsbD_extractLsb', hl, getLsbD_setWidth,
+        show w ≤ i + l by omega, getLsbD_of_ge acc (i * len + l) (by omega)]
+  · case succ n' ihn' =>
+    rw [extractAndExtendAux]
+    split
+    · omega
+    · apply ihn'
+      · intros i hi
+        have hcast : len + k * len = (k + 1) * len := by
+            simp [Nat.mul_comm, Nat.mul_add, Nat.add_comm]
+
+        by_cases hi' : i < k
+        · have heq : extractLsb' (i * len) len (BitVec.cast hcast (extractAndExtendBit k len x ++ acc))  =
+              extractLsb' (i * len) len ((extractAndExtendBit k len x ++ acc)) := by
+            ext; simp
+          rw [heq, extractLsb'_append_of_lt hi']
+          apply hacc
+          exact hi'
+        · have heq : extractLsb' (i * len) len (BitVec.cast hcast (extractAndExtendBit k len x ++ acc))   =
+              extractLsb' (i * len) len ((extractAndExtendBit k len x ++ acc)) := by
+            ext; simp
+          rw [heq, extractLsb'_append_of_eq (by omega)]
+          simp [show i = k by omega, extractAndExtendBit]
+      · omega
+
+theorem extractLsb'_cpopLayer {w iterNum i oldLen : Nat} {oldLayer : BitVec (oldLen * w)}
+    {newLayer : BitVec (iterNum * w)} (hold : 2 * (iterNum - 1) < oldLen) :
+    (∀ i (_hi: i < iterNum),
+      newLayer.extractLsb' (i * w) w =
+      oldLayer.extractLsb' ((2 * i) * w) w + (oldLayer.extractLsb' ((2 * i + 1) * w) w)) →
+    extractLsb' (i * w) w (oldLayer.cpopLayer newLayer hold) =
+      extractLsb' (2 * i * w) w oldLayer + extractLsb' ((2 * i + 1) * w) w oldLayer := by
+  intro proof_addition
+  rw [cpopLayer]
+  split
+  · by_cases hi : i < iterNum
+    · simp only [extractLsb'_cast]
+      apply proof_addition
+      exact hi
+    · ext j hj
+      have : iterNum * w ≤ i * w := by refine mul_le_mul_right w (by omega)
+      have : oldLen * w ≤ (2 * i) * w := by refine mul_le_mul_right w (by omega)
+      have : oldLen * w ≤ (2 * i + 1) * w := by refine mul_le_mul_right w (by omega)
+      have hz : extractLsb' (2 * i * w) w oldLayer = 0#w := by
+        ext j hj
+        simp [show oldLen * w ≤ 2 * i * w + j by omega]
+      have hz' : extractLsb' ((2 * i + 1) * w) w oldLayer = 0#w := by
+        ext j hj
+        simp [show oldLen * w ≤ (2 * i + 1) * w + j by omega]
+      simp [show iterNum * w ≤ i * w + j by omega, hz, hz']
+  · generalize hop1 : oldLayer.extractLsb' ((2 * iterNum) * w) w = op1
+    generalize hop2 : oldLayer.extractLsb' ((2 * iterNum + 1) * w) w = op2
+    have hcast : w + iterNum * w = (iterNum + 1) * w := by simp [Nat.add_mul]; omega
+    apply extractLsb'_cpopLayer
+    intros i hi
+    by_cases hlt : i < iterNum
+    · rw [extractLsb'_cast, extractLsb'_append_eq_of_add_le]
+      · apply proof_addition
+        exact hlt
+      · rw [show i * w + w = i * w + 1 * w by omega, ← Nat.add_mul]
+        exact mul_le_mul_right w hlt
+    · rw [extractLsb'_cast, show i = iterNum by omega, extractLsb'_append_eq_left, hop1, hop2]
+termination_by oldLen - 2 * (iterNum + 1 - 1)
+
+theorem getLsbD_cpopLayer {w iterNum: Nat} {oldLayer : BitVec (oldLen * w)}
+    {newLayer : BitVec (iterNum * w)} (hold : 2 * (iterNum - 1) < oldLen) :
+    (∀ i (_hi: i < iterNum),
+          newLayer.extractLsb' (i * w) w =
+          oldLayer.extractLsb' ((2 * i) * w) w + (oldLayer.extractLsb' ((2 * i + 1) * w) w)) →
+    (oldLayer.cpopLayer newLayer hold).getLsbD k =
+      (extractLsb' (2 * ((k - k % w) / w) * w) w oldLayer +
+        extractLsb' ((2 * ((k - k % w) / w) + 1) * w) w oldLayer).getLsbD (k % w) := by
+  intro proof_addition
+  by_cases hw0 : w = 0
+  · subst hw0
+    simp
+  · simp only [← extractLsb'_cpopLayer (hold := by omega) proof_addition,
+      Nat.mod_lt (x := k) (y := w) (by omega), getLsbD_eq_getElem, getElem_extractLsb']
+    congr
+    by_cases hmod : k % w = 0
+    · rw [hmod, Nat.sub_zero, Nat.add_zero, Nat.div_mul_cancel (by omega)]
+    · rw [Nat.div_mul_cancel (by exact dvd_sub_mod k), Nat.sub_add_cancel (by exact mod_le k w)]
+
+@[simp]
+private theorem addRecAux_zero {x : BitVec (l * w)} {acc : BitVec w} :
+    x.addRecAux 0 acc = acc := rfl
+
+@[simp]
+private theorem addRecAux_succ {x : BitVec (l * w)} {n : Nat} {acc : BitVec w} :
+    x.addRecAux (n + 1) acc = x.addRecAux n (acc + extractLsb' (n * w) w x) := rfl
+
+private theorem addRecAux_eq {x : BitVec (l * w)} {n : Nat} {acc : BitVec w} :
+     x.addRecAux n acc = x.addRecAux n 0#w + acc := by
+  induction n generalizing acc
+  · case zero =>
+    simp
+  · case succ n ihn =>
+    simp only [addRecAux_succ, BitVec.zero_add, ihn (acc := extractLsb' (n * w) w x),
+      BitVec.add_assoc, ihn (acc := acc + extractLsb' (n * w) w x), BitVec.add_right_inj]
+    rw [BitVec.add_comm (x := acc)]
+
+private theorem extractLsb'_addRecAux_of_le {x : BitVec (len * w)} (h : r ≤ k):
+    (extractLsb' 0 (k * w) x).addRecAux r 0#w = x.addRecAux r 0#w := by
+  induction r generalizing x len k
+  · case zero =>
+    simp [addRecAux]
+  · case succ diff ihdiff =>
+    simp only [addRecAux_succ, BitVec.zero_add]
+    have hext : diff * w + w ≤ k * w := by
+      simp only [show diff * w + w = (diff + 1) * w by simp [Nat.add_mul]]
+      exact Nat.mul_le_mul_right w h
+    rw [extractLsb'_extractLsb'_of_le hext, addRecAux_eq (x := x),
+        addRecAux_eq (x := extractLsb' 0 (k * w) x), ihdiff (x := x) (by omega) (k := k)]
+
+private theorem extractLsb'_extractAndExtend_eq {i len : Nat} {x : BitVec w} :
+    (extractAndExtend len x).extractLsb' (i * len) len = extractAndExtendBit i len x := by
+  unfold extractAndExtend
+  by_cases hilt : i < w
+  · ext j hj
+    simp [extractLsb'_extractAndExtendAux, extractAndExtendBit]
+  · ext k hk
+    have := Nat.mul_le_mul_right (n := w) (k := len) (m := i) (by omega)
+    simp only [extractAndExtendBit, cast_ofNat, getElem_extractLsb', truncate_eq_setWidth,
+      getElem_setWidth, getLsbD_extractLsb', Nat.lt_one_iff]
+    rw [getLsbD_of_ge, getLsbD_of_ge]
+    · simp
+    · omega
+    · omega
+
+private theorem addRecAux_append_extractLsb' {x : BitVec (len * w)} (ha : 0 < len) :
+    ((x.extractLsb' ((len - 1) * w) w ++
+      x.extractLsb' 0 ((len - 1) * w)).cast (m := len * w) hcast).addRecAux len 0#w =
+    x.extractLsb' ((len - 1) * w) w +
+      (x.extractLsb' 0 ((len - 1) * w)).addRecAux (len - 1) 0#w := by
+  simp only [extractLsb'_addRecAux_of_le (k := len - 1) (r := len - 1) (by omega),
+    BitVec.append_extractLsb'_of_lt (hcast := hcast)]
+  have hsucc := addRecAux_succ (x := x) (acc := 0#w) (n := len - 1)
+  rw [BitVec.zero_add, Nat.sub_one_add_one (by omega)] at hsucc
+  rw [hsucc, addRecAux_eq, BitVec.add_comm]
+
+private theorem Nat.mul_add_le_mul_of_succ_le {a b c : Nat} (h : a + 1 ≤ c) :
+    a * b + b ≤ c * b := by
+  rw [← Nat.succ_mul]
+  exact mul_le_mul_right b h
+
+/--
+  The recursive addition of `w`-long words on two flattened bitvectors `x` and `y` (with different
+  number of words `len` and `len'`, respectively) returns the same value, if we can prove
+  that each `w`-long word in `x` results from the addition of two `w`-long words in `y`,
+  using exactly all `w`-long words in `y`.
+-/
+private theorem addRecAux_eq_of {x : BitVec (len * w)} {y : BitVec (len' * w)}
+    (hlen : len = (len' + 1) / 2) :
+    (∀ (i : Nat) (_h : i < (len' + 1) / 2),
+      extractLsb' (i * w) w x = extractLsb' (2 * i * w) w y + extractLsb' ((2 * i + 1) * w) w y) →
+    x.addRecAux len 0#w = y.addRecAux len' 0#w := by
+  intro hadd
+  induction len generalizing len' y
+  · case zero =>
+    simp [show len' = 0 by omega]
+  · case succ len ih =>
+    have hcast : w + (len + 1 - 1) * w = (len + 1) * w := by
+      simp [Nat.add_mul, Nat.add_comm]
+    have hcast' :  w + (len' - 1) * w = len' * w := by
+      rw [Nat.sub_mul, Nat.one_mul,
+        ← Nat.add_sub_assoc (by refine Nat.le_mul_of_pos_left w (by omega)), Nat.add_comm]
+      simp
+    rw [addRecAux_succ, ← BitVec.append_extractLsb'_of_lt (x := x) (hcast := hcast)]
+    have happ := addRecAux_append_extractLsb' (len := len + 1) (x := x) (hcast := hcast) (by omega)
+    simp only [Nat.add_one_sub_one, addRecAux_succ, BitVec.zero_add] at happ
+    simp only [Nat.add_one_sub_one, BitVec.zero_add, happ]
+    have := Nat.succ_mul (n := len' - 1) (m := w)
+    rw [succ_eq_add_one, Nat.sub_one_add_one (by omega)] at this
+    by_cases hmod : len' % 2 = 0
+    · /- `sum` results from the addition of the two last elements in `y`, `sum = op1 + op2` -/
+      have := Nat.mul_le_mul_right (n := len' - 1 - 1) (m := len' - 1) (k := w) (by omega)
+      have := Nat.succ_mul (n := len' - 1 - 1) (m := w)
+      have hcast'' :  w + (len' - 1 - 1) * w = (len' - 1) * w := by
+        rw [Nat.sub_mul, Nat.one_mul,
+          ← Nat.add_sub_assoc (k := w) (by refine Nat.le_mul_of_pos_left w (by omega))]
+        simp
+      rw [succ_eq_add_one, Nat.sub_one_add_one (by omega)] at this
+      rw [← BitVec.append_extractLsb'_of_lt (x := y) (hcast := hcast'),
+        addRecAux_append_extractLsb' (by omega),
+        ← BitVec.append_extractLsb'_of_lt (x := extractLsb' 0 ((len' - 1) * w) y) (hcast := hcast''),
+        addRecAux_append_extractLsb' (by omega),
+        extractLsb'_extractLsb'_of_le (by exact Nat.mul_add_le_mul_of_succ_le (by omega)),
+        extractLsb'_extractLsb'_of_le (by omega), ← BitVec.add_assoc, hadd (_h := by omega)]
+      congr 1
+      · rw [show len = (len' + 1) / 2 - 1 by omega, BitVec.add_comm]
+        congr <;> omega
+      · apply ih
+        · omega
+        · intros
+          rw [extractLsb'_extractLsb'_of_le (by exact Nat.mul_add_le_mul_of_succ_le (by omega)),
+            extractLsb'_extractLsb'_of_le (by exact Nat.mul_add_le_mul_of_succ_le (by omega)),
+            extractLsb'_extractLsb'_of_le (by exact Nat.mul_add_le_mul_of_succ_le (by omega)),
+            hadd (_h := by omega)]
+    · /- `sum` results from the addition of the last elements in `y` with `0#w` -/
+      have : len' * w ≤ (len' - 1 + 1) * w := by exact mul_le_mul_right w (by omega)
+      rw [← BitVec.append_extractLsb'_of_lt (x := y) (hcast := hcast'),
+        addRecAux_append_extractLsb' (by omega), hadd (_h := by omega),
+        show 2 * len = len' - 1 by omega]
+      congr 1
+      · rw [BitVec.add_right_eq_self]
+        ext k hk
+        simp only [getElem_extractLsb', getElem_zero]
+        apply getLsbD_of_ge y ((len' - 1 + 1) * w + k) (by omega)
+      · apply ih
+        · omega
+        · intros
+          rw [extractLsb'_extractLsb'_of_le (by exact Nat.mul_add_le_mul_of_succ_le (by omega)),
+            extractLsb'_extractLsb'_of_le (by exact Nat.mul_add_le_mul_of_succ_le (by omega)),
+            extractLsb'_extractLsb'_of_le (by exact Nat.mul_add_le_mul_of_succ_le (by omega)),
+            hadd (_h := by omega)]
+
+private theorem getLsbD_extractAndExtend_of_lt {x : BitVec w} (hk : k < v) :
+    (x.extractAndExtend v).getLsbD (pos * v + k) = (extractAndExtendBit pos v x).getLsbD k := by
+  simp [← extractLsb'_extractAndExtend_eq (w := w) (len := v) (i := pos) (x := x)]
+  omega
+
+/--
+  Extracting a bit from a `BitVec.extractAndExtend` is the same as extracting a bit
+  from a zero-extended bit at a certain position in the original bitvector.
+-/
+theorem getLsbD_extractAndExtend {x : BitVec w} (hv : 0 < v) :
+    (BitVec.extractAndExtend v x).getLsbD k =
+    (BitVec.extractAndExtendBit ((k - (k % v)) / v) v x).getLsbD (k % v):= by
+  rw [← getLsbD_extractAndExtend_of_lt (by exact mod_lt k hv)]
+  congr
+  by_cases hmod : k % v = 0
+  · simp only [hmod, Nat.sub_zero, Nat.add_zero]
+    rw [Nat.div_mul_cancel (by omega)]
+  · rw [← Nat.div_eq_sub_mod_div]
+    exact Eq.symm (div_add_mod' k v)
+
+private theorem addRecAux_extractAndExtend_eq_cpopNatRec {x : BitVec w} :
+    (extractAndExtend w x).addRecAux n 0#w = x.cpopNatRec n 0 := by
+  induction n
+  · case zero =>
+    simp
+  · case succ n' ihn' =>
+    rw [cpopNatRec_succ, Nat.zero_add, natCast_eq_ofNat, addRecAux_succ, BitVec.zero_add,
+      addRecAux_eq, cpopNatRec_eq, ihn', ofNat_add, natCast_eq_ofNat, BitVec.add_right_inj,
+      extractLsb'_extractAndExtend_eq]
+    ext k hk
+    simp only [extractAndExtendBit, ← getLsbD_eq_getElem, getLsbD_ofNat, hk, decide_true,
+      Bool.true_and, truncate_eq_setWidth, getLsbD_setWidth, getLsbD_extractLsb', Nat.lt_one_iff]
+    by_cases hk0 : k = 0
+    · simp only [hk0, testBit_zero, decide_true, Nat.add_zero, Bool.true_and]
+      cases x.getLsbD n' <;> simp
+    · simp only [show ¬k = 0 by omega, decide_false, Bool.false_and]
+      symm
+      apply testBit_lt_two_pow ?_
+      have : (x.getLsbD n').toNat ≤ 1 := by
+        cases x.getLsbD n' <;> simp
+      have : 1 < 2 ^ k := by exact Nat.one_lt_two_pow hk0
+      omega
+
+private theorem addRecAux_extractAndExtend_eq_cpop {x : BitVec w} :
+    (extractAndExtend w x).addRecAux w 0#w = x.cpop := by
+  simp only [cpop]
+  apply addRecAux_extractAndExtend_eq_cpopNatRec
+
+private theorem addRecAux_cpopTree {x : BitVec (len * w)} :
+    addRecAux ((cpopTree x).cast (m := 1 * w) (by simp)) 1 0#w = addRecAux x len 0#w := by
+  unfold cpopTree
+  split
+  · case _ h =>
+    subst h
+    simp [addRecAux]
+  · case _ h =>
+    split
+    · case _ h' =>
+      simp only [addRecAux_succ, Nat.zero_mul, BitVec.zero_add, addRecAux_zero, h']
+      ext; simp
+    · rw [addRecAux_cpopTree]
+      apply BitVec.addRecAux_eq_of (x := cpopLayer x 0#(0 * w) (by omega)) (y := x)
+      · rfl
+      · intros j hj
+        simp [extractLsb'_cpopLayer]
+termination_by len
+
+private theorem addRecAux_eq_cpopTree {x : BitVec (len * w)} :
+    x.addRecAux len 0#w = (x.cpopTree).cast (by simp) := by
+  rw [← addRecAux_cpopTree, addRecAux_succ, Nat.zero_mul, BitVec.zero_add, addRecAux_zero]
+  ext k hk
+  simp [← getLsbD_eq_getElem, hk]
+
+theorem cpop_eq_cpopRec {x : BitVec w} :
+    BitVec.cpop x = BitVec.cpopRec x := by
+  unfold BitVec.cpopRec
+  split
+  · simp [← addRecAux_extractAndExtend_eq_cpop, addRecAux_eq_cpopTree (x := extractAndExtend w x)]
+  · split
+    · ext k hk
+      cases hx : x.getLsbD 0
+      <;> simp [hx, cpop, ← getLsbD_eq_getElem, show k = 0 by omega, show w = 1 by omega]
+    · have hw : w = 0 := by omega
+      subst hw
+      simp [of_length_zero]
+
 end BitVec

src/Init/Data/BitVec/Lemmas.lean

@@ -2581,6 +2581,19 @@ theorem msb_signExtend {x : BitVec w} :
   · simp [h, BitVec.msb, getMsbD_signExtend, show v - w = 0 by omega]
   · simp [h, BitVec.msb, getMsbD_signExtend, show ¬ (v - w = 0) by omega]
 
+/-- Sign-extending to `w + n` bits, extracting bits `[w - 1 + n..n]`, and setting width
+back to `w` is equivalent to arithmetic right shift by `n`, since both sides discard the `n`
+least significant bits and replicate the sign bit into the upper bits. -/
+@[simp]
+theorem signExtend_extractLsb_setWidth {x : BitVec w} {n : Nat} :
+    ((x.signExtend (w + n)).extractLsb (w - 1 + n) n).setWidth w = x.sshiftRight n := by
+  ext i hi
+  simp only [getElem_sshiftRight, getElem_setWidth, getLsbD_extract,
+    Nat.add_sub_cancel, show i ≤ w - 1 by omega, decide_true, getLsbD_signExtend,
+    Bool.true_and]
+  by_cases hni : n + i < w
+  <;> (simp [hni]; omega)
+
 /-- Sign extending to a width smaller than the starting width is a truncation. -/
 theorem signExtend_eq_setWidth_of_le (x : BitVec w) {v : Nat} (hv : v ≤ w) :
   x.signExtend v = x.setWidth v := by
@@ -2773,6 +2786,14 @@ theorem msb_append {x : BitVec w} {y : BitVec v} :
   rw [getElem_append] -- Why does this not work with `simp [getElem_append]`?
   simp
 
+theorem append_of_zero_width (x : BitVec w) (y : BitVec v) (h : w = 0) :
+    (x ++ y) = y.cast (by simp [h]) := by
+  ext i ih
+  subst h
+  simp [← getLsbD_eq_getElem, getLsbD_append]
+  omega
+
+set_option backward.isDefEq.respectTransparency false in
 @[grind =]
 theorem toInt_append {x : BitVec n} {y : BitVec m} :
     (x ++ y).toInt = if n == 0 then y.toInt else (2 ^ m) * x.toInt + y.toNat := by
@@ -2999,6 +3020,34 @@ theorem extractLsb'_append_extractLsb'_eq_extractLsb' {x : BitVec w} (h : start
   congr 1
   omega
 
+theorem append_extractLsb'_of_lt {x : BitVec (x_len * w)} :
+    (x.extractLsb' ((x_len - 1) * w) w ++ x.extractLsb' 0 ((x_len - 1) * w)).cast hcast = x := by
+  ext i hi
+  simp only [getElem_cast, getElem_append, getElem_extractLsb', Nat.zero_add, dite_eq_ite]
+  rw [← getLsbD_eq_getElem, ite_eq_left_iff, Nat.not_lt]
+  intros
+  simp only [show (x_len - 1) * w + (i - (x_len - 1) * w) = i by omega]
+
+
+theorem extractLsb'_append_of_lt {x : BitVec (k * w)} {y : BitVec w} (hlt : i < k) :
+    extractLsb' (i * w) w (y ++ x) = extractLsb' (i * w) w x := by
+  ext j hj
+  simp only [← getLsbD_eq_getElem, getLsbD_extractLsb', hj, decide_true, getLsbD_append,
+    Bool.true_and, ite_eq_left_iff, Nat.not_lt]
+  intros h
+  by_cases hw0 : w = 0
+  · subst hw0
+    simp
+  · have : i * w ≤ (k - 1) * w := Nat.mul_le_mul_right w (by omega)
+    have h' : i * w + j < (k - 1 + 1) * w := by simp [Nat.add_mul]; omega
+    rw [Nat.sub_one_add_one (by omega)] at h'
+    omega
+
+theorem extractLsb'_append_of_eq {x : BitVec (k * w)} {y : BitVec w} (heq : i = k) :
+    extractLsb' (i * w) w (y ++ x) = y := by
+  ext j hj
+  simp [← getLsbD_eq_getElem, getLsbD_append, hj, heq]
+
 /-- Combine adjacent `~~~ (extractLsb _)'` operations into a single `~~~ (extractLsb _)'`. -/
 theorem not_extractLsb'_append_not_extractLsb'_eq_not_extractLsb' {x : BitVec w} (h : start₂ = start₁ + len₁) :
     (~~~ (x.extractLsb' start₂ len₂) ++ ~~~ (x.extractLsb' start₁ len₁)) =

src/Init/Data/Bool.lean

@@ -629,6 +629,7 @@ export Bool (cond_eq_if cond_eq_ite xor and or not)
 This should not be turned on globally as an instance because it degrades performance in Mathlib,
 but may be used locally.
 -/
+@[implicit_reducible]
 def boolPredToPred : Coe (α → Bool) (α  → Prop) where
   coe r := fun a => Eq (r a) true
 
@@ -636,7 +637,7 @@ def boolPredToPred : Coe (α → Bool) (α  → Prop) where
 This should not be turned on globally as an instance because it degrades performance in Mathlib,
 but may be used locally.
 -/
-@[expose, instance_reducible] def boolRelToRel : Coe (α → α → Bool) (α → α → Prop) where
+@[expose, implicit_reducible] def boolRelToRel : Coe (α → α → Bool) (α → α → Prop) where
   coe r := fun a b => Eq (r a b) true
 
 /-! ### subtypes -/

src/Init/Data/ByteArray/Lemmas.lean

@@ -111,13 +111,13 @@ theorem getElem_eq_getElem_data {a : ByteArray} {i : Nat} {h : i < a.size} :
 theorem getElem_append_left {i : Nat} {a b : ByteArray} {h : i < (a ++ b).size}
     (hlt : i < a.size) : (a ++ b)[i] = a[i] := by
   simp only [getElem_eq_getElem_data, data_append]
-  rw [Array.getElem_append_left (by simpa)]
+  rw [Array.getElem_append_left (by simpa)]; rfl
 
 theorem getElem_append_right {i : Nat} {a b : ByteArray} {h : i < (a ++ b).size}
     (hle : a.size ≤ i) : (a ++ b)[i] = b[i - a.size]'(by simp_all; omega) := by
   simp only [getElem_eq_getElem_data, data_append]
   rw [Array.getElem_append_right (by simpa)]
-  simp
+  simp; rfl
 
 @[simp]
 theorem _root_.List.getElem_toByteArray {l : List UInt8} {i : Nat} {h : i < l.toByteArray.size} :
@@ -223,7 +223,7 @@ theorem getElem_extract_aux {xs : ByteArray} {start stop : Nat} (h : i < (xs.ext
 
 theorem getElem_extract {i : Nat} {b : ByteArray} {start stop : Nat}
     (h) : (b.extract start stop)[i]'h = b[start + i]'(getElem_extract_aux h) := by
-  simp [getElem_eq_getElem_data]
+  simp [getElem_eq_getElem_data]; rfl
 
 theorem extract_eq_extract_left {a : ByteArray} {i i' j : Nat} :
     a.extract i j = a.extract i' j ↔ min j a.size - i = min j a.size - i' := by
@@ -236,25 +236,25 @@ theorem extract_add_one {a : ByteArray} {i : Nat} (ha : i + 1 ≤ a.size) :
     omega
   · rename_i j hj hj'
     obtain rfl : j = 0 := by simpa using hj'
-    simp [ByteArray.getElem_eq_getElem_data]
+    simp [ByteArray.getElem_eq_getElem_data]; rfl
 
 theorem extract_add_two {a : ByteArray} {i : Nat} (ha : i + 2 ≤ a.size) :
     a.extract i (i + 2) = [a[i], a[i + 1]].toByteArray := by
   rw [extract_eq_extract_append_extract (i + 1) (by simp) (by omega),
     extract_add_one (by omega), extract_add_one (by omega)]
-  simp [← List.toByteArray_append]
+  simp [← List.toByteArray_append]; rfl
 
 theorem extract_add_three {a : ByteArray} {i : Nat} (ha : i + 3 ≤ a.size) :
     a.extract i (i + 3) = [a[i], a[i + 1], a[i + 2]].toByteArray := by
   rw [extract_eq_extract_append_extract (i + 1) (by simp) (by omega),
     extract_add_one (by omega), extract_add_two (by omega)]
-  simp [← List.toByteArray_append]
+  simp [← List.toByteArray_append]; rfl
 
 theorem extract_add_four {a : ByteArray} {i : Nat} (ha : i + 4 ≤ a.size) :
     a.extract i (i + 4) = [a[i], a[i + 1], a[i + 2], a[i + 3]].toByteArray := by
   rw [extract_eq_extract_append_extract (i + 1) (by simp) (by omega),
     extract_add_one (by omega), extract_add_three (by omega)]
-  simp [← List.toByteArray_append]
+  simp [← List.toByteArray_append]; rfl
 
 theorem append_assoc {a b c : ByteArray} : a ++ b ++ c = a ++ (b ++ c) := by
   ext1

src/Init/Data/Char/Basic.lean

@@ -129,6 +129,14 @@ The ASCII digits are the following: `0123456789`.
 @[inline] def isDigit (c : Char) : Bool :=
   c.val ≥ '0'.val && c.val ≤ '9'.val
 
+/--
+Returns `true` if the character is an ASCII hexadecimal digit.
+
+The ASCII hexadecimal digits are the following: `0123456789abcdefABCDEF`.
+-/
+@[inline] def isHexDigit (c : Char) : Bool :=
+  c.isDigit || (c.val ≥ 'a'.val && c.val ≤ 'f'.val) || (c.val ≥ 'A'.val && c.val ≤ 'F'.val)
+
 /--
 Returns `true` if the character is an ASCII letter or digit.
 

src/Init/Data/Char/Lemmas.lean

@@ -50,7 +50,7 @@ instance ltTrans : Trans (· < · : Char → Char → Prop) (· < ·) (· < ·)
   trans := Char.lt_trans
 
 -- This instance is useful while setting up instances for `String`.
-@[instance_reducible]
+@[implicit_reducible]
 def notLTTrans : Trans (¬ · < · : Char → Char → Prop) (¬ · < ·) (¬ · < ·) where
   trans h₁ h₂ := by simpa using Char.le_trans (by simpa using h₂) (by simpa using h₁)
 
@@ -62,7 +62,7 @@ instance ltTrichotomous : Std.Trichotomous (· < · : Char → Char → Prop) wh
   trichotomous _ _ h₁ h₂ := Char.le_antisymm (by simpa using h₂) (by simpa using h₁)
 
 @[deprecated ltTrichotomous (since := "2025-10-27")]
-def notLTAntisymm : Std.Antisymm (¬ · < · : Char → Char → Prop) where
+theorem notLTAntisymm : Std.Antisymm (¬ · < · : Char → Char → Prop) where
   antisymm := Char.ltTrichotomous.trichotomous
 
 instance ltAsymm : Std.Asymm (· < · : Char → Char → Prop) where
@@ -73,7 +73,7 @@ instance leTotal : Std.Total (· ≤ · : Char → Char → Prop) where
 
 -- This instance is useful while setting up instances for `String`.
 @[deprecated ltAsymm (since := "2025-08-01")]
-def notLTTotal : Std.Total (¬ · < · : Char → Char → Prop) where
+theorem notLTTotal : Std.Total (¬ · < · : Char → Char → Prop) where
   total := fun x y => by simpa using Char.le_total y x
 
 @[simp] theorem ofNat_toNat (c : Char) : Char.ofNat c.toNat = c := by

src/Init/Data/Fin/Fold.lean

@@ -4,7 +4,6 @@ Released under Apache 2.0 license as described in the file LICENSE.
 Authors: François G. Dorais
 -/
 module
-
 prelude
 public import Init.Control.Lawful.Basic
 public import Init.Ext
@@ -13,7 +12,7 @@ import Init.Data.Nat.Lemmas
 import Init.Omega
 import Init.TacticsExtra
 import Init.WFTactics
-
+import Init.Hints
 public section
 
 namespace Fin

src/Init/Data/Fin/Iterate.lean

@@ -4,14 +4,12 @@ Released under Apache 2.0 license as described in the file LICENSE.
 Authors: Joe Hendrix
 -/
 module
-
 prelude
 public import Init.Data.Fin.Basic
 import Init.PropLemmas
 import Init.WFTactics
-
+import Init.Hints
 public section
-
 namespace Fin
 
 /--
@@ -71,7 +69,7 @@ private theorem hIterateFrom_elim {P : Nat → Sort _}(Q : ∀(i : Nat), P i →
     have g : ¬ (i < n) := by simp at p; simp [p]
     have r : Q n (_root_.cast (congrArg P p) s) :=
       @Eq.rec Nat i (fun k eq => Q k (_root_.cast (congrArg P eq) s)) init n p
-    simp only [g, r, dite_false]
+    simp only [g, dite_false]; exact r
   | succ j inv =>
     unfold hIterateFrom
     have d : Nat.succ i + j = n := by simp [Nat.succ_add]; exact p

src/Init/Data/Fin/Lemmas.lean

@@ -4,7 +4,6 @@ Released under Apache 2.0 license as described in the file LICENSE.
 Authors: Mario Carneiro, Leonardo de Moura
 -/
 module
-
 prelude
 public import Init.Ext
 public import Init.Data.Nat.Div.Basic
@@ -15,7 +14,7 @@ import Init.Data.Nat.Lemmas
 import Init.Data.Nat.Linear
 import Init.Omega
 import Init.TacticsExtra
-
+import Init.Hints
 @[expose] public section
 
 open Std
@@ -124,7 +123,7 @@ For example, for `x : Fin k` and `n : Nat`,
 it causes `x < n` to be elaborated as `x < ↑n` rather than `↑x < n`,
 silently introducing wraparound arithmetic.
 -/
-@[expose, instance_reducible]
+@[expose, implicit_reducible]
 def instNatCast (n : Nat) [NeZero n] : NatCast (Fin n) where
   natCast a := Fin.ofNat n a
 
@@ -146,7 +145,7 @@ This is not a global instance, but may be activated locally via `open Fin.IntCas
 
 See the doc-string for `Fin.NatCast.instNatCast` for more details.
 -/
-@[expose, instance_reducible]
+@[expose, implicit_reducible]
 def instIntCast (n : Nat) [NeZero n] : IntCast (Fin n) where
   intCast := Fin.intCast
 
@@ -998,7 +997,7 @@ For the induction:
 
 @[simp, grind =] theorem reverseInduction_last {n : Nat} {motive : Fin (n + 1) → Sort _} {zero succ} :
     (reverseInduction zero succ (Fin.last n) : motive (Fin.last n)) = zero := by
-  rw [reverseInduction, reverseInduction.go]; simp
+  rw [reverseInduction, reverseInduction.go]; simp; rfl
 
 private theorem reverseInduction_castSucc_aux {n : Nat} {motive : Fin (n + 1) → Sort _} {succ}
     (i : Fin n) (j : Nat) (h) (h2 : i.1 < j) (zero : motive ⟨j, h⟩) :
@@ -1009,9 +1008,9 @@ private theorem reverseInduction_castSucc_aux {n : Nat} {motive : Fin (n + 1)
   | succ j ih =>
     rw [reverseInduction.go, dif_neg (by exact Nat.ne_of_lt h2)]
     by_cases hij : i = j
-    · subst hij; simp [reverseInduction.go]
-    dsimp only
-    rw [ih _ _ (by omega), eq_comm, reverseInduction.go, dif_neg (by change i.1 + 1 ≠ _; omega)]
+    · subst hij; simp [reverseInduction.go]; rfl
+    · dsimp only
+      rw [ih _ _ (by omega), eq_comm, reverseInduction.go, dif_neg (by change i.1 + 1 ≠ _; omega)]
 
 @[simp, grind =] theorem reverseInduction_castSucc {n : Nat} {motive : Fin (n + 1) → Sort _} {zero succ}
     (i : Fin n) : reverseInduction (motive := motive) zero succ (castSucc i) =

src/Init/Data/Format/Basic.lean

@@ -414,7 +414,7 @@ Renders a `Format` to a string.
 -/
 def pretty (f : Format) (width : Nat := defWidth) (indent : Nat := 0) (column := 0) : String :=
   let act : StateM State Unit := prettyM f width indent
-  State.out <| act (State.mk "" column) |>.snd
+  State.out <| act.run (State.mk "" column) |>.snd
 
 end Format
 

src/Init/Data/Hashable.lean

@@ -6,7 +6,7 @@ Authors: Leonardo de Moura
 module
 
 prelude
-public import Init.Data.String.PosRaw
+import Init.Data.Array.Basic
 public import Init.Data.UInt.Basic
 
 public section
@@ -15,9 +15,6 @@ universe u
 instance : Hashable Nat where
   hash n := UInt64.ofNat n
 
-instance : Hashable String.Pos.Raw where
-  hash p := UInt64.ofNat p.byteIdx
-
 instance [Hashable α] [Hashable β] : Hashable (α × β) where
   hash | (a, b) => mixHash (hash a) (hash b)
 

src/Init/Data/Iterators/Basic.lean

@@ -315,7 +315,7 @@ of another state. Having this proof bundled up with the step is important for te
 See `IterM.Step` and `Iter.Step` for the concrete choice of the plausibility predicate.
 -/
 @[expose]
-def PlausibleIterStep (IsPlausibleStep : IterStep α β → Prop) := Subtype IsPlausibleStep
+abbrev PlausibleIterStep (IsPlausibleStep : IterStep α β → Prop) := Subtype IsPlausibleStep
 
 /--
 Match pattern for the `yield` case. See also `IterStep.yield`.
@@ -379,6 +379,8 @@ class Iterator (α : Type w) (m : Type w → Type w') (β : outParam (Type w)) w
   -/
   step : (it : IterM (α := α) m β) → m (Shrink <| PlausibleIterStep <| IsPlausibleStep it)
 
+attribute [reducible] Iterator.IsPlausibleStep
+
 section Monadic
 
 /-- The constructor has been renamed. -/
@@ -424,7 +426,6 @@ theorem IterM.toIter_mk {α β} {it : α} :
 Asserts that certain step is plausibly the successor of a given iterator. What "plausible" means
 is up to the `Iterator` instance but it should be strong enough to allow termination proofs.
 -/
-@[expose]
 abbrev IterM.IsPlausibleStep {α : Type w} {m : Type w → Type w'} {β : Type w} [Iterator α m β] :
     IterM (α := α) m β → IterStep (IterM (α := α) m β) β → Prop :=
   Iterator.IsPlausibleStep (α := α) (m := m)
@@ -493,7 +494,7 @@ Asserts that certain step is plausibly the successor of a given iterator. What "
 is up to the `Iterator` instance but it should be strong enough to allow termination proofs.
 -/
 @[expose]
-def Iter.IsPlausibleStep {α : Type w} {β : Type w} [Iterator α Id β]
+abbrev Iter.IsPlausibleStep {α : Type w} {β : Type w} [Iterator α Id β]
     (it : Iter (α := α) β) (step : IterStep (Iter (α := α) β) β) : Prop :=
   it.toIterM.IsPlausibleStep (step.mapIterator Iter.toIterM)
 
@@ -549,7 +550,7 @@ The type of the step object returned by `Iter.step`, containing an `IterStep`
 and a proof that this is a plausible step for the given iterator.
 -/
 @[expose]
-def Iter.Step {α : Type w} {β : Type w} [Iterator α Id β] (it : Iter (α := α) β) :=
+abbrev Iter.Step {α : Type w} {β : Type w} [Iterator α Id β] (it : Iter (α := α) β) :=
   PlausibleIterStep (Iter.IsPlausibleStep it)
 
 /--

src/Init/Data/Iterators/Combinators/Monadic/FlatMap.lean

@@ -362,8 +362,7 @@ def Flatten.instProductivenessRelation [Monad m] [Iterator α m (IterM (α := α
     case innerDone =>
       apply Flatten.productiveRel_of_right₂
 
-@[no_expose]
-public def Flatten.instProductive [Monad m] [Iterator α m (IterM (α := α₂) m β)] [Iterator α₂ m β]
+public theorem Flatten.instProductive [Monad m] [Iterator α m (IterM (α := α₂) m β)] [Iterator α₂ m β]
     [Finite α m] [Productive α₂ m] : Productive (Flatten α α₂ β m) m :=
   .of_productivenessRelation instProductivenessRelation
 

src/Init/Data/Iterators/Consumers/Loop.lean

@@ -35,7 +35,7 @@ A `ForIn'` instance for iterators. Its generic membership relation is not easy t
 so this is not marked as `instance`. This way, more convenient instances can be built on top of it
 or future library improvements will make it more comfortable.
 -/
-@[always_inline, inline]
+@[always_inline, inline, expose, implicit_reducible]
 def Iter.instForIn' {α : Type w} {β : Type w} {n : Type x → Type x'} [Monad n]
     [Iterator α Id β] [IteratorLoop α Id n] :
     ForIn' n (Iter (α := α) β) β ⟨fun it out => it.IsPlausibleIndirectOutput out⟩ where
@@ -53,7 +53,7 @@ instance (α : Type w) (β : Type w) (n : Type x → Type x') [Monad n]
 /--
 An implementation of `for h : ... in ... do ...` notation for partial iterators.
 -/
-@[always_inline, inline]
+@[always_inline, inline, expose, implicit_reducible]
 def Iter.Partial.instForIn' {α : Type w} {β : Type w} {n : Type x → Type x'} [Monad n]
     [Iterator α Id β] [IteratorLoop α Id n] :
     ForIn' n (Iter.Partial (α := α) β) β ⟨fun it out => it.it.IsPlausibleIndirectOutput out⟩ where
@@ -71,7 +71,7 @@ instance (α : Type w) (β : Type w) (n : Type x → Type x') [Monad n]
 A `ForIn'` instance for iterators that is guaranteed to terminate after finitely many steps.
 It is not marked as an instance because the membership predicate is difficult to work with.
 -/
-@[always_inline, inline]
+@[always_inline, inline, expose, implicit_reducible]
 def Iter.Total.instForIn' {α : Type w} {β : Type w} {n : Type x → Type x'} [Monad n]
     [Iterator α Id β] [IteratorLoop α Id n] [Finite α Id] :
     ForIn' n (Iter.Total (α := α) β) β ⟨fun it out => it.it.IsPlausibleIndirectOutput out⟩ where

src/Init/Data/Iterators/Consumers/Monadic/Loop.lean

@@ -159,7 +159,7 @@ This is the default implementation of the `IteratorLoop` class.
 It simply iterates through the iterator using `IterM.step`. For certain iterators, more efficient
 implementations are possible and should be used instead.
 -/
-@[always_inline, inline, expose]
+@[always_inline, inline, expose, implicit_reducible]
 def IteratorLoop.defaultImplementation {α : Type w} {m : Type w → Type w'} {n : Type x → Type x'}
     [Monad n] [Iterator α m β] :
     IteratorLoop α m n where
@@ -211,7 +211,7 @@ theorem IteratorLoop.wellFounded_of_productive {α β : Type w} {m : Type w →
 /--
 This `ForIn'`-style loop construct traverses a finite iterator using an `IteratorLoop` instance.
 -/
-@[always_inline, inline]
+@[always_inline, inline, expose, implicit_reducible]
 def IteratorLoop.finiteForIn' {m : Type w → Type w'} {n : Type x → Type x'}
     {α : Type w} {β : Type w} [Iterator α m β] [IteratorLoop α m n] [Monad n]
     (lift : ∀ γ δ, (γ → n δ) → m γ → n δ) :
@@ -224,7 +224,7 @@ A `ForIn'` instance for iterators. Its generic membership relation is not easy t
 so this is not marked as `instance`. This way, more convenient instances can be built on top of it
 or future library improvements will make it more comfortable.
 -/
-@[always_inline, inline]
+@[always_inline, inline, expose, implicit_reducible]
 def IterM.instForIn' {m : Type w → Type w'} {n : Type w → Type w''}
     {α : Type w} {β : Type w} [Iterator α m β] [IteratorLoop α m n] [Monad n]
     [MonadLiftT m n] :
@@ -239,7 +239,7 @@ instance IterM.instForInOfIteratorLoop {m : Type w → Type w'} {n : Type w →
   instForInOfForIn'
 
 /-- Internal implementation detail of the iterator library. -/
-@[always_inline, inline]
+@[always_inline, inline, expose, implicit_reducible]
 def IterM.Partial.instForIn' {m : Type w → Type w'} {n : Type w → Type w''}
     {α : Type w} {β : Type w} [Iterator α m β] [IteratorLoop α m n] [MonadLiftT m n] [Monad n] :
     ForIn' n (IterM.Partial (α := α) m β) β ⟨fun it out => it.it.IsPlausibleIndirectOutput out⟩ where
@@ -247,7 +247,7 @@ def IterM.Partial.instForIn' {m : Type w → Type w'} {n : Type w → Type w''}
     haveI := @IterM.instForIn'; forIn' it.it init f
 
 /-- Internal implementation detail of the iterator library. -/
-@[always_inline, inline]
+@[always_inline, inline, expose, implicit_reducible]
 def IterM.Total.instForIn' {m : Type w → Type w'} {n : Type w → Type w''}
     {α : Type w} {β : Type w} [Iterator α m β] [IteratorLoop α m n] [MonadLiftT m n] [Monad n]
     [Finite α m] :

src/Init/Data/Iterators/Internal/LawfulMonadLiftFunction.lean

@@ -70,7 +70,7 @@ theorem LawfulMonadLiftFunction.lift_seqRight [LawfulMonad m] [LawfulMonad n]
 abbrev idToMonad [Monad m] ⦃α : Type u⦄ (x : Id α) : m α :=
     pure x.run
 
-def LawfulMonadLiftFunction.idToMonad [Monad m] [LawfulMonad m] :
+theorem LawfulMonadLiftFunction.idToMonad [LawfulMonad m] :
     LawfulMonadLiftFunction (m := Id) (n := m) idToMonad where
   lift_pure := by simp [Internal.idToMonad]
   lift_bind := by simp [Internal.idToMonad]
@@ -95,7 +95,7 @@ instance [LawfulMonadLiftBindFunction (n := n) (fun _ _ f x => lift x >>= f)] [L
     simpa using LawfulMonadLiftBindFunction.liftBind_bind (n := n)
       (liftBind := fun _ _ f x => lift x >>= f) (β := β) (γ := γ) (δ := γ) pure x g
 
-def LawfulMonadLiftBindFunction.id [Monad m] [LawfulMonad m] :
+theorem LawfulMonadLiftBindFunction.id [LawfulMonad m] :
     LawfulMonadLiftBindFunction (m := Id) (n := m) (fun _ _ f x => f x.run) where
   liftBind_pure := by simp
   liftBind_bind := by simp

src/Init/Data/Iterators/Lemmas/Combinators/FilterMap.lean

@@ -765,6 +765,7 @@ theorem Iter.anyM_eq_anyM_mapM_pure {α β : Type} {m : Type → Type w'} [Itera
   rw [forIn_eq_match_step, IterM.forIn_eq_match_step, bind_assoc, step_mapM]
   cases it.step using PlausibleIterStep.casesOn
   · rename_i out _
+    simp only
     simp only [bind_assoc, pure_bind, map_eq_pure_bind, Shrink.inflate_deflate,
       liftM, monadLift]
     have {x : m Bool} : x = MonadAttach.attach (pure out) >>= (fun _ => x) := by
@@ -777,7 +778,7 @@ theorem Iter.anyM_eq_anyM_mapM_pure {α β : Type} {m : Type → Type w'} [Itera
     apply bind_congr; intro px
     split
     · simp
-    · simp [ihy ‹_›]
+    · simp [ihy ‹_›, monadLift]
   · simp [ihs ‹_›]
   · simp
 

src/Init/Data/Iterators/Lemmas/Combinators/FlatMap.lean

@@ -232,6 +232,7 @@ public theorem Iter.toArray_flatMapM {α α₂ β γ : Type w} {m : Type w → T
     (it₁.flatMapM f).toArray = Array.flatten <$> (it₁.mapM fun b => do (← f b).toArray).toArray := by
   simp [flatMapM, toArray_flatMapAfterM]
 
+set_option backward.isDefEq.respectTransparency false in
 public theorem Iter.toList_flatMapAfter {α α₂ β γ : Type w} [Iterator α Id β] [Iterator α₂ Id γ]
     [Finite α Id] [Finite α₂ Id]
     {f : β → Iter (α := α₂) γ} {it₁ : Iter (α := α) β} {it₂ : Option (Iter (α := α₂) γ)} :
@@ -242,6 +243,7 @@ public theorem Iter.toList_flatMapAfter {α α₂ β γ : Type w} [Iterator α I
   simp only [flatMapAfter, Iter.toList, toIterM_toIter, IterM.toList_flatMapAfter]
   cases it₂ <;> simp [map, IterM.toList_map_eq_toList_mapM, - IterM.toList_map]
 
+set_option backward.isDefEq.respectTransparency false in
 public theorem Iter.toArray_flatMapAfter {α α₂ β γ : Type w} [Iterator α Id β] [Iterator α₂ Id γ]
     [Finite α Id] [Finite α₂ Id]
     {f : β → Iter (α := α₂) γ} {it₁ : Iter (α := α) β} {it₂ : Option (Iter (α := α₂) γ)} :

src/Init/Data/Iterators/Lemmas/Combinators/Monadic/FilterMap.lean

@@ -182,11 +182,12 @@ theorem IterM.step_filterMap [Monad m] [LawfulMonad m] {f : β → Option β'} :
       pure <| .deflate <| .skip (it'.filterMap f) (.skip h)
     | .done h =>
       pure <| .deflate <| .done (.done h)) := by
-  simp only [IterM.filterMap, step_filterMapWithPostcondition, pure]
+  simp only [IterM.filterMap]
+  simp only [step_filterMapWithPostcondition, PostconditionT.operation_pure]
   apply bind_congr
   intro step
   split
-  · simp only [PostconditionT.pure, PlausibleIterStep.skip, PlausibleIterStep.yield, pure_bind]
+  · simp only [PlausibleIterStep.skip, PlausibleIterStep.yield, pure_bind]
     split <;> split <;> simp_all
   · simp
   · simp
@@ -361,8 +362,8 @@ theorem IterM.toList_map_eq_toList_mapM {α β γ : Type w}
       bind_map_left]
     conv => rhs; rhs; ext a; rw [← pure_bind (x := a.val) (f := fun _ => _ <$> _)]
     simp only [← bind_assoc, bind_pure_comp, WeaklyLawfulMonadAttach.map_attach]
-    simp [ihy ‹_›]
-  · simp [ihs ‹_›]
+    simpa using ihy ‹_›
+  · simpa using ihs ‹_›
   · simp
 
 theorem IterM.toList_map_eq_toList_filterMapM {α β γ : Type w} {m : Type w → Type w'}
@@ -373,6 +374,7 @@ theorem IterM.toList_map_eq_toList_filterMapM {α β γ : Type w} {m : Type w
   simp [toList_map_eq_toList_mapM, toList_mapM_eq_toList_filterMapM]
   congr <;> simp
 
+set_option backward.whnf.reducibleClassField false in
 /--
 Variant of `toList_filterMapWithPostcondition_filterMapWithPostcondition` that is intended to be
 used with the `apply` tactic. Because neither the LHS nor the RHS determine all implicit parameters,
@@ -600,6 +602,7 @@ theorem IterM.toList_map_mapM {α β γ δ : Type w}
     toList_filterMapM_mapM]
   congr <;> simp
 
+set_option backward.isDefEq.respectTransparency false in
 @[simp]
 theorem IterM.toList_filterMapWithPostcondition {α β γ : Type w} {m : Type w → Type w'}
     [Monad m] [LawfulMonad m]
@@ -623,6 +626,7 @@ theorem IterM.toList_filterMapWithPostcondition {α β γ : Type w} {m : Type w
   · simp [ihs ‹_›, heq]
   · simp [heq]
 
+set_option backward.isDefEq.respectTransparency false in
 @[simp]
 theorem IterM.toList_mapWithPostcondition {α β γ : Type w} {m : Type w → Type w'}
     [Monad m] [LawfulMonad m] [Iterator α Id β] [Finite α Id]
@@ -643,6 +647,7 @@ theorem IterM.toList_mapWithPostcondition {α β γ : Type w} {m : Type w → Ty
   · simp [ihs ‹_›, heq]
   · simp [heq]
 
+set_option backward.isDefEq.respectTransparency false in
 @[simp]
 theorem IterM.toList_filterMapM {α β γ : Type w} {m : Type w → Type w'}
     [Monad m] [MonadAttach m] [LawfulMonad m] [WeaklyLawfulMonadAttach m]
@@ -652,6 +657,7 @@ theorem IterM.toList_filterMapM {α β γ : Type w} {m : Type w → Type w'}
   simp [toList_filterMapM_eq_toList_filterMapWithPostcondition, toList_filterMapWithPostcondition,
     PostconditionT.attachLift, PostconditionT.run_eq_map, WeaklyLawfulMonadAttach.map_attach]
 
+set_option backward.isDefEq.respectTransparency false in
 @[simp]
 theorem IterM.toList_mapM {α β γ : Type w} {m : Type w → Type w'}
     [Monad m] [MonadAttach m] [LawfulMonad m] [WeaklyLawfulMonadAttach m]
@@ -1297,6 +1303,7 @@ theorem IterM.forIn_filterMap
   rw [filterMap, forIn_filterMapWithPostcondition]
   simp [PostconditionT.run_eq_map]
 
+set_option backward.isDefEq.respectTransparency false in
 theorem IterM.forIn_mapWithPostcondition
     [Monad m] [LawfulMonad m] [Monad n] [LawfulMonad n] [Monad o] [LawfulMonad o]
     [MonadLiftT m n] [LawfulMonadLiftT m n] [MonadLiftT n o] [LawfulMonadLiftT n o]

src/Init/Data/Iterators/Lemmas/Combinators/Monadic/FlatMap.lean

@@ -36,7 +36,7 @@ theorem IterM.step_flattenAfter {α α₂ β : Type w} {m : Type w → Type w'}
   cases it₂
   all_goals
   · apply bind_congr; intro step
-    cases step.inflate using PlausibleIterStep.casesOn <;> simp [IterM.flattenAfter]
+    cases step.inflate using PlausibleIterStep.casesOn <;> simp [IterM.flattenAfter] <;> rfl
 
 namespace Iterators.Types
 

src/Init/Data/Iterators/Lemmas/Combinators/Monadic/ULift.lean

@@ -29,7 +29,7 @@ theorem IterM.step_uLift [Iterator α m β] [Monad n] {it : IterM (α := α) m
       | .done h => return .deflate (.done ⟨_, h, rfl⟩)) := by
   simp only [IterM.step, Iterator.step, IterM.uLift]
   apply bind_congr; intro step
-  split <;> simp [Types.ULiftIterator.Monadic.modifyStep, *]
+  split <;> simp [Types.ULiftIterator.Monadic.modifyStep, *] <;> rfl
 
 @[simp]
 theorem IterM.toList_uLift [Iterator α m β] [Monad m] [Monad n] {it : IterM (α := α) m β}

src/Init/Data/Iterators/Lemmas/Combinators/ULift.lean

@@ -42,7 +42,7 @@ theorem Iter.step_uLift [Iterator α Id β] {it : Iter (α := α) β} :
 theorem Iter.toList_uLift [Iterator α Id β] {it : Iter (α := α) β}
     [Finite α Id] :
     it.uLift.toList = it.toList.map ULift.up := by
-  simp only [monadLift, uLift_eq_toIter_uLift_toIterM, IterM.toList_toIter]
+  simp only [uLift_eq_toIter_uLift_toIterM, IterM.toList_toIter]
   rw [IterM.toList_uLift]
   simp [monadLift, Iter.toList_eq_toList_toIterM]
 
@@ -63,7 +63,7 @@ theorem Iter.toArray_uLift [Iterator α Id β] {it : Iter (α := α) β}
 theorem Iter.length_uLift [Iterator α Id β] {it : Iter (α := α) β}
     [Finite α Id] [IteratorLoop α Id Id] [LawfulIteratorLoop α Id Id] :
     it.uLift.length = it.length := by
-  simp only [monadLift, uLift_eq_toIter_uLift_toIterM, length_eq_length_toIterM, toIterM_toIter]
+  simp only [uLift_eq_toIter_uLift_toIterM, length_eq_length_toIterM, toIterM_toIter]
   rw [IterM.length_uLift]
   simp [monadLift]
 

src/Init/Data/Iterators/Lemmas/Consumers/Loop.lean

@@ -32,7 +32,7 @@ theorem Iter.forIn'_eq {α β : Type w} [Iterator α Id β] [Finite α Id]
       IterM.DefaultConsumers.forIn' (n := m) (fun _ _ f x => f x.run) γ (fun _ _ _ => True)
         it.toIterM init _ (fun _ => id)
           (fun out h acc => return ⟨← f out (Iter.isPlausibleIndirectOutput_iff_isPlausibleIndirectOutput_toIterM.mpr h) acc, trivial⟩) := by
-  simp +instances only [instForIn', ForIn'.forIn', IteratorLoop.finiteForIn']
+  simp +instances only [ForIn'.forIn']
   have : ∀ a b c, f a b c = (Subtype.val <$> (⟨·, trivial⟩) <$> f a b c) := by simp
   simp +singlePass only [this]
   rw [hl.lawful (fun _ _ f x => f x.run) (wf := IteratorLoop.wellFounded_of_finite)]
@@ -58,8 +58,7 @@ theorem Iter.forIn_eq {α β : Type w} [Iterator α Id β] [Finite α Id]
     forIn' ita b f = forIn' itb b' g := by
   subst_eqs
   simp only [← funext_iff] at h
-  rw [← h]
-  rfl
+  rw [← h]; rfl
 
 @[congr] theorem Iter.forIn_congr {α β : Type w} {m : Type w → Type w'} [Monad m]
     [Iterator α Id β] [Finite α Id] [IteratorLoop α Id m]
@@ -82,7 +81,7 @@ theorem Iter.forIn'_eq_forIn'_toIterM {α β : Type w} [Iterator α Id β]
       letI : ForIn' m (IterM (α := α) Id β) β _ := IterM.instForIn'
       ForIn'.forIn' it.toIterM init
         (fun out h acc => f out (isPlausibleIndirectOutput_iff_isPlausibleIndirectOutput_toIterM.mpr h) acc) := by
-  simp +instances [ForIn'.forIn', Iter.instForIn', IterM.instForIn', monadLift]
+  simp +instances [ForIn'.forIn', monadLift]
 
 theorem Iter.forIn_eq_forIn_toIterM {α β : Type w} [Iterator α Id β]
     [Finite α Id] {m : Type w → Type w''} [Monad m] [LawfulMonad m]
@@ -276,8 +275,7 @@ theorem Iter.forIn'_eq_forIn'_toList {α β : Type w} [Iterator α Id β]
     {f : (out : β) → _ → γ → m (ForInStep γ)} :
     letI : ForIn' m (Iter (α := α) β) β _ := Iter.instForIn'
     ForIn'.forIn' it init f = ForIn'.forIn' it.toList init (fun out h acc => f out (Iter.mem_toList_iff_isPlausibleIndirectOutput.mp h) acc) := by
-  simp only [forIn'_toList]
-  congr
+  simp only [forIn'_toList]; rfl
 
 theorem Iter.forIn'_eq_forIn'_toArray {α β : Type w} [Iterator α Id β]
     [Finite α Id] {m : Type x → Type x'} [Monad m] [LawfulMonad m]
@@ -287,8 +285,7 @@ theorem Iter.forIn'_eq_forIn'_toArray {α β : Type w} [Iterator α Id β]
     {f : (out : β) → _ → γ → m (ForInStep γ)} :
     letI : ForIn' m (Iter (α := α) β) β _ := Iter.instForIn'
     ForIn'.forIn' it init f = ForIn'.forIn' it.toArray init (fun out h acc => f out (Iter.mem_toArray_iff_isPlausibleIndirectOutput.mp h) acc) := by
-  simp only [forIn'_toArray]
-  congr
+  simp only [forIn'_toArray]; rfl
 
 theorem Iter.forIn_toList {α β : Type w} [Iterator α Id β]
     [Finite α Id] {m : Type x → Type x'} [Monad m] [LawfulMonad m]