1:2.4.6+ds-4 in the archive, ships a compiled extension
tests failed,
ours is 1:2.4.6+ds-4+py315.1.
proves nothing: no (sid-python3.15)
Rebuilding it releases 45 packages waiting on it, as this stage last counted.
published,
1:2.4.6+ds-4+py315.1.
| Question | Answer | Said by | As of |
|---|---|---|---|
| does the release team track this? |
dependency level 4
The release team's own tracker lists it, and their dependency level is how the transition is sequenced: a level cannot be read as done while anything below it is outstanding. Their question is what a package's binaries depend on, not what it ships, so their verdict and ours can disagree without either being wrong. amd64 bad, arm64 bad, armhf bad, i386 bad, loong64 bad, ppc64el bad, riscv64 bad, s390x bad |
release.debian.org, from their ben tracker | 2026-09-16 04:49 UTC |
| what did triage decide about the test? |
infrastructure
Somebody's decision about this failure, not a measurement. The runs beside it are free to disagree with it, which is why both are kept. Disk space, not memory: OSError [Errno 28] ENOSPC in numpy/lib/tests/test_io.py::TestSaveTxt::test_large_zip, which writes a >4GB zip. Fails identically on BOTH arches - amd64 WR 985225 and arm64 WR 985226 - so not arm64-specific. python3.15 itself passes clean (51287 passed amd64, 49172 arm64); the failure lands on whichever interpreter runs 3rd of 3 (python3.14 here). Runs are sequential (114-419s, 422-727s, 730-1035s), so space appears to accumulate across them rather than 3 files coexisting; tempdir() does rmtree in a finally, so the test does not leak its own file - leaking path unidentified. Testbed is an incus container: requires_memory() checks only RAM via /proc/meminfo (which in a container reports the host, a 48-core EPYC) and never checks free disk, so the 7GB guard passes trivially. debian/rules SKIP_TESTS already excludes test_big_arrays/test_large_file_support under the comment 'some buildds and Salsa do not like us creating 5 GB files' but omits test_large_zip. DEFERRED 2026-08-03: packages build and publish fine, not a 3.15 blocker. Revisit: add test_large_zip to SKIP_TESTS (covers build+autopkgtest, all arches, via PYBUILD_TEST_ARGS) and probe df in the testbed to confirm whether disk accumulates. |
whoever triaged it | 2026-08-16 10:17 UTC |
| Version | Arch | Kind | Result | Binaries | Interpreters | Closure | Finished |
|---|---|---|---|---|---|---|---|
1:2.4.6+ds-4+py315.1
|
arm64 | test | fail | rebuilt | 3.13, 3.14, 3.15 | archive-copy-installed | 2026-08-03 10:48 |
1:2.4.6+ds-4+py315.1
|
amd64 | test | fail | rebuilt | 3.13, 3.14, 3.15 | archive-copy-installed | 2026-08-03 10:12 |