[WIP] Attempt to reproduce Windows compiler crashes

[fhanau]

Not ready for review, just trying to reproduce the compiler crash/see if changing compiler options could mitigate it

[ask-bonk]

Quick review of the WIP changes — flagging two issues I noticed.

[ask-bonk]

Unmatched quote: there's a " before /O2 but no closing ". This will likely cause Bazel to parse the remainder of the file as part of this flag value, breaking everything after this line.

The stray quote should either be removed (to match the style used by the existing v8-codegen-opt-windows entries on lines 99-105) or a closing quote added at the end.

Suggested change

	build:windows --per_file_copt=.*capnp/rpc\.c++@"/O2,/Gy,/Gw,-fstrict-aliasing,/clang:-O3
	build:windows --per_file_copt=.*capnp/rpc\.c++@/O2,/Gy,/Gw,-fstrict-aliasing,/clang:-O3

[ask-bonk]

Intentional? Moving debug_memory=False and the per-file optimizations from config:release_windows to config:windows means they now apply to all Windows build modes (fastbuild, dbg, etc.), not just release builds. In particular:

• debug_memory=False disables KJ debug-memory checks in all Windows builds, which line 200 explicitly enables for all platforms.
• The per-file /O2 + /clang:-O3 opts on lines 435-437 compile rpc.c++, server.c++, and container-client.c++ at high optimization even in debug builds, which can make debugging those files difficult.

If this is intentional for reproducing the crash, ignore this — just flagging it so it doesn't accidentally land as-is.

[ask-bonk]

I'm Bonk, and I've done a quick review of your PR.

This PR moves some Windows build config flags around in .bazelrc to try to reproduce compiler crashes, as noted in the WIP title.

Two issues found, posted as review comments on PR #6416:

1. Unmatched quote (line 435) — The per_file_copt for rpc.c++ has an opening " before /O2 with no closing quote. This will likely break Bazel's parsing of subsequent lines. Suggestion provided.

2. Config scope widened (lines 434-437) — debug_memory=False and the per-file /O2+/clang:-O3 optimizations moved from config:release_windows to config:windows, meaning they now apply to all Windows builds (fastbuild, dbg, etc.), not just release. Flagged as potentially unintentional.

github run

[codecov-commenter]

Codecov Report

❌ Patch coverage is 38.31418% with 161 lines in your changes missing coverage. Please review.
✅ Project coverage is 70.74%. Comparing base (47f5816) to head (7b2dbe8).
⚠️ Report is 3 commits behind head on main.

Files with missing lines	Patch %	Lines
src/workerd/server/container-client.c++	0.00%	127 Missing ⚠️
src/workerd/server/server.c++	74.62%	12 Missing and 5 partials ⚠️
src/workerd/server/alarm-scheduler.c++	61.53%	8 Missing and 2 partials ⚠️
src/workerd/server/server-test.c++	50.00%	0 Missing and 3 partials ⚠️
src/workerd/api/sockets.c++	81.81%	0 Missing and 2 partials ⚠️
src/workerd/io/worker-entrypoint.c++	95.45%	0 Missing and 1 partial ⚠️
src/workerd/server/alarm-scheduler.h	50.00%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #6416      +/-   ##
==========================================
- Coverage   70.79%   70.74%   -0.05%     
==========================================
  Files         427      427              
  Lines      117869   117385     -484     
  Branches    18935    18891      -44     
==========================================
- Hits        83447    83048     -399     
+ Misses      23154    23090      -64     
+ Partials    11268    11247      -21     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[codspeed-hq]

Merging this PR will not alter performance

✅ 70 untouched benchmarks
⏩ 129 skipped benchmarks¹

---

_{Comparing felix/032426-reproduce-windows-crashes (7b2dbe8) with main (f966235)}

Footnotes

1. 129 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩