Fix CUDA-Linux build: retry without sccache on nvcc device-compile crash - #385
Merged
Merged
Conversation
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
had a problem deploying
to
startgate
August 9, 2026 22:48 — with
GitHub Actions
Error
bernardladenthin
had a problem deploying
to
maven-central
August 9, 2026 22:48 — with
GitHub Actions
Failure
bernardladenthin
had a problem deploying
to
maven-central
August 9, 2026 22:48 — with
GitHub Actions
Failure
|
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. |
|
6 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.



Summary
-virtualarchitecture targets (fatbinary fatal: Could not open input file 'acc.compute_75.ptx'followed bysccache: Compiler killed by signal 1), caught by theCross-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 #384build.shalready 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 backCompiler killed by signal, so this failure mode now falls back to an uncached, release-safe-O3build like every other known sccache/nvcc incompatibilityCLAUDE.md's sccache rollout section (same pattern as the other documented incidents there)Test plan
CLAUDE.md) updated with the incident + fixRelated issues / PRs
Follow-up to #384 (CI regression surfaced on the post-merge publish dispatch)
Checklist
CONTRIBUTING.mdandCODE_OF_CONDUCT.mdhttps://claude.ai/code/session_01CtT7KLNmC8gEoDS1EiGGwD
Generated by Claude Code