KhronosGroup / KhronosGroup/SPIRV-LLVM-Translator
>60% test failures on big endian architectures
- Dominant language
- LLVM
- Stars
- 625
- Forks
- 279
- Avg merge
- 3d 5h
- Merged PRs (30d)
- 34
Description
More than 60% of the tests are failing on the three Debian big-endian architectures s390x, ppc64, sparc64:
https://buildd.debian.org/status/fetch.php?pkg=spirv-llvm-translator-17&arch=s390x&ver=17.0.0-4&stamp=1707586764&raw=0
https://buildd.debian.org/status/fetch.php?pkg=spirv-llvm-translator-17&arch=ppc64&ver=17.0.0-4&stamp=1707587023&raw=0
https://buildd.debian.org/status/fetch.php?pkg=spirv-llvm-translator-17&arch=sparc64&ver=17.0.0-4&stamp=1707595393&raw=0
(Testsuite failures are currently fatal only on amd64: issue #1964)
Furthermore, running the testsuite on big-endian architectures requires >40GB of diskspace (while it only needs around 500MB on little-endian architectures). One particularily huge (40GB) file is build/test/test_output/DebugInfo/Generic/Output/two-cus-from-same-file.ll.tmp
Similar results happen on the other branches.
In comparison, on the little-endian architecture ppc64el only 8 tests are failing:
https://buildd.debian.org/status/fetch.php?pkg=spirv-llvm-translator-17&arch=ppc64el&ver=17.0.0-4&stamp=1707585588&raw=0
Contributor guide
Research direction
Start by reading the three linked Debian build logs for s390x, ppc64, and sparc64, then inspect the failures and the oversized build/test/test_output/DebugInfo/Generic/Output/two-cus-from-same-file.ll.tmp file. Done means substantially fewer big-endian test failures and no longer requiring more than 40GB of disk space, but the issue does not identify a specific code path or test fix.
Written by the indexing model from the issue text.
Assessment
- Domain
- compilers, operating-systems, testing-qa
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 30/100