Skip to content

Fix CUDA-Linux build: retry without sccache on nvcc device-compile crash - #385

Merged
bernardladenthin merged 1 commit into
mainfrom
claude/cuda-toolkit-action-update-0675dk
Aug 9, 2026
Merged

Fix CUDA-Linux build: retry without sccache on nvcc device-compile crash#385
bernardladenthin merged 1 commit into
mainfrom
claude/cuda-toolkit-action-update-0675dk

Conversation

@bernardladenthin

Copy link
Copy Markdown
Owner

Summary

  • CUDA 13.3's nvcc broke sccache's nvcc wrapping for -virtual architecture targets (fatbinary fatal: Could not open input file 'acc.compute_75.ptx' followed by sccache: Compiler killed by signal 1), caught by the Cross-Compile manylinux_2_28 x86_64 (CUDA) job on the b10333/CUDA-13.3 publish dispatch that followed Upgrade llama.cpp to b10333 and CUDA to 13.3 #384
  • build.sh already retries the build without sccache when it detects an sccache/nvcc cache error, but its trigger regex didn't cover this failure's wording, so the job went red instead of falling back
  • Extend the regex to also match Compiler killed by signal, so this failure mode now falls back to an uncached, release-safe -O3 build like every other known sccache/nvcc incompatibility
  • Document the incident in CLAUDE.md's sccache rollout section (same pattern as the other documented incidents there)

Test plan

  • Regex verified locally against the exact captured log lines (now matches)
  • CI is green on this branch (CUDA-Linux job re-runs the full uncached fallback path)
  • Docs (CLAUDE.md) updated with the incident + fix

Related issues / PRs

Follow-up to #384 (CI regression surfaced on the post-merge publish dispatch)

Checklist

  • I have read CONTRIBUTING.md and CODE_OF_CONDUCT.md
  • My commits follow Conventional Commits
  • No security-sensitive changes

https://claude.ai/code/session_01CtT7KLNmC8gEoDS1EiGGwD


Generated by Claude Code

CUDA 13.3's nvcc broke sccache's nvcc wrapping for -virtual architecture
targets ("fatbinary fatal: Could not open input file 'acc.compute_75.ptx'"
followed by "sccache: Compiler killed by signal 1"), caught by the
CUDA-Linux job on the b10333/CUDA-13.3 publish dispatch (run
31339594933). The existing sccache-error retry-without-cache fallback in
build.sh didn't trigger because its regex didn't cover this wording.
Extend the regex so this failure mode falls back to an uncached build
like every other known sccache/nvcc incompatibility, instead of redding
the job.
@bernardladenthin
bernardladenthin merged commit 7aa3d66 into main Aug 9, 2026
9 of 15 checks passed
@bernardladenthin
bernardladenthin deleted the claude/cuda-toolkit-action-update-0675dk branch August 9, 2026 22:49
@claude

claude Bot commented Aug 9, 2026

Copy link
Copy Markdown

This PR addresses a CUDA 13.3 nvcc regression correctly. The regex extension to catch Compiler killed by signal is proper and the documentation is thorough. No issues found. Recommend approval.

@sonarqubecloud

sonarqubecloud Bot commented Aug 9, 2026

Copy link
Copy Markdown

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants