Chrome builds interrupted by module compilation failure #649
Comments
|
My guess is we don't have the libffi-dev package installed hence we couldn't build |
|
That sounds plausible and more satisfying than my feeble "transitory packaging issue." |
|
Surprisingly, two more builds just failed with the very same error (8 days later). This happened about 3 hours ago, and that directly correlates with the release of cffi 1.12.2. @gsnedders @jgraham I know next to nothing about Python packaging, but this seems like something we ought to be able to avoid. Is there anything we can change in WPT to guard against invalid installations? |
|
Hah.
Adjust the |
|
Thanks, James! That looks like a fairly module-specific solution. Is that an indication of some bad practice in |
|
Ooh, pip seems to have a |
|
Six more builds (3 Firefox, 2 Chrome, and 1 Edge) were interrupted today at 14:05 UTC. There's no new release of |
|
Do you have logs? |
|
Sure do!
All for WPT commit 9f49616e7fa90489fac85dd203e86d7c83ca67da |
|
I'm not sure why it's trying to build it there; maybe something happened to the wheel package. In any case passing |
|
At least the ones I looked at there aren't To expand on my previous comment: note that uploads aren't atomic on PyPI. The common flow is:
Each of these items happens sequentially, and if a source release artifact is uploaded before a (suitable) binary release artifact is anything looking for the latest release will download the source release and build it. As noted, either of the options @jgraham suggested will passing this will change this (preferring an older release with a binary release). |
|
We got hit by both releases of the "cryptography" package (the second release interfered with six builds for Firefox stable). I'm going to look in to patching WPT as you folks recommend. I'm curious, though: have either of you seen this occur in Taskcluster? |
… install, a=testonly Automatic update from web-platform-tests [wpt] Avoid race condition during module install (#15614) See web-platform-tests/results-collection#649 -- wpt-commits: cc4f2960306867e50d5beba4e86749aa75961fe8 wpt-pr: 15614
… install, a=testonly Automatic update from web-platform-tests [wpt] Avoid race condition during module install (#15614) See web-platform-tests/results-collection#649 -- wpt-commits: cc4f2960306867e50d5beba4e86749aa75961fe8 wpt-pr: 15614
… install, a=testonly Automatic update from web-platform-tests [wpt] Avoid race condition during module install (#15614) See web-platform-tests/results-collection#649 -- wpt-commits: cc4f2960306867e50d5beba4e86749aa75961fe8 wpt-pr: 15614 UltraBlame original commit: 63a641c33b4d6264c2226911fb61d7d59f9aacff
… install, a=testonly Automatic update from web-platform-tests [wpt] Avoid race condition during module install (#15614) See web-platform-tests/results-collection#649 -- wpt-commits: cc4f2960306867e50d5beba4e86749aa75961fe8 wpt-pr: 15614 UltraBlame original commit: 63a641c33b4d6264c2226911fb61d7d59f9aacff
… install, a=testonly Automatic update from web-platform-tests [wpt] Avoid race condition during module install (#15614) See web-platform-tests/results-collection#649 -- wpt-commits: cc4f2960306867e50d5beba4e86749aa75961fe8 wpt-pr: 15614 UltraBlame original commit: 63a641c33b4d6264c2226911fb61d7d59f9aacff
Yesterday, 2 Chrome stable builds and 1 Chrome experimental builds failed as the WPT CLI installed Python dependencies. Specifically, the workers were unable to compile the native extensions provided by the
cffimodule (error log included below).I haven't been able to find a conclusive explanation for this failure, but I suspect it was a transitory packaging issue. Here's why:
cffiwas released at approximately that time](https://groups.google.com/forum/#!topic/python-cffi/AuYExAf8LSY)I'll keep my eyes peeled for future occurrences, but I'm hopeful that it was a passing glitch. I've manually re-triggered the failed builds; the corresponding data should be available on wpt.fyi shortly.
Partial error log
The text was updated successfully, but these errors were encountered: