TEST: 1.6.x + blas variants#153
Conversation
|
Hi! This is the friendly automated conda-forge-linting service. I just wanted to let you know that I linted all conda-recipes in your PR ( |
NOTE: no pypy builds at the momentGoing from 1.5.4 to 1.6.0From 13 failures & 6 timeouts out of 80 combinations, there are now (after removing 12 pypy builds and 16 py36 builds) 9 failures & 4 timeouts out of 52 combinations. The remaining worst offenders:
Details
Build logs (before #151 got merged) win + blis+ cpython 3.7 & 3.9: 6-7 failed tests
win + blis+ cpython 3.8: 7 failed testswin + blis+ cpython 3.8 (before rebase): 436 failed testswin + openblas + cpython: 1 failed test |
6717394 to
3b74bb8
Compare
d3ef10d to
bb314cb
Compare
|
Seems that somehow, the new blis breaks scipy (or scipy is broken with the new blis) under osx & windows (though it works under linux), and only for python 3.7 & 3.8 (3.9 works): OSX: Win: |
|
|
NOTE: no pypy builds at the momentUpdate with blis 0.8.0From 9 failures & 4 timeouts out of 52 combinations, we're at 7 failures & 4 timeouts out of 52 (still missing 12 pypy builds, pending the migration from conda-forge/conda-forge-pinning-feedstock#1024 to arrive at this feedstock) The good news:
The biggest remaining offenders:
Details
Build logs (before conda-forge/blis-feedstock#21 got merged) linux + x86 + blis + cpython 3.8: 1 failed testAppeared for 3.7 & 3.8 here win + blis: 438-439 failed testswin + blis (before): 1 failed testsSee here win + openblas + cpython: 1 failed test |
05e0541 to
24f3e08
Compare
|
@isuruf, now that conda-forge/blis-feedstock#21 and conda-forge/blas-feedstock#63 are in, blis is somehow back to ~440 failures, despite having been down to 1 here. That run was based on the Based on this, I went back to an environment not containing |
1575173 to
4b5800c
Compare
Update with pypy back in playFrom 7 failures & 4 timeouts out of 52 combinations (previously 9 failures & 4 timeouts), we're now at 10 failures & 2 time outs out of 64 combinations. The bad news:
The biggest remaining offenders:
Details
Build logs (restarted in error, drone/travis are the old links, new azure run has exact same errors as described): linux + x86 + blis: 1 failed testAll except py39 job osx + x86 + mkl + pypy: 1 failed testswin + blis + cpython 3.7: 1 segfaultwin + blis + cpython 3.8: 439 failed testswin + blis + cpython 3.9: new and exciting crashwin + openblas + cpython: 1 failed test |
4b5800c to
1b582b5
Compare
1e40e14 to
95014bd
Compare
|
Hi! This is the friendly automated conda-forge-linting service. I was trying to look for recipes to lint for you, but it appears we have a merge conflict. Please ping the 'conda-forge/core' team (using the @ notation in a comment) if you believe this is a bug. |
95014bd to
270c02f
Compare
|
Hi! This is the friendly automated conda-forge-linting service. I just wanted to let you know that I linted all conda-recipes in your PR ( |
270c02f to
ff36ce3
Compare
Update with numpy 1.20.1 & JIT re-enabled for PyPyFrom 10 failures & 4 time outs out of 64 combinations, we're now (pretty unchanged) at 11 failures & 3 time outs out of 64 combinations The good news:
The bad news:
The biggest remaining offenders:
Details
Build logs (restarted in error, drone/travis are the old links, new azure run has exact same errors as described): osx + x86 + mkl + pypy: 2 failed testswin + blis + cpython 3.8 / 3.9: ~440 failed testswin + openblas + cpython: 1 failed test |
ff36ce3 to
2de3b23
Compare
Update for 1.6.2From 11 failures & 3 time outs out of 64 combinations we're effectively unchanged at 10 failures & 5 time outs. The bad news:
The biggest remaining offenders:
Details
Build logs (restarted in error, drone/travis are the old links, new azure run has exact same errors as described): osx + x86 + mkl + pypy: 2 failed testswin + blis + cpython 3.7:win + blis + cpython 3.8 / 3.9: ~440 failed testsAs before... win + openblas + cpython: 1 failed test |
…nda-forge-pinning 2021.05.07.21.54.12
Update for 1.6.3 & new blas buildsFrom 10 failures & 5 time outs out of 64 combinations we're now at 7 failures & 3 time outs. The good news:
The bad news:
Details
Build logs: linux + x86 + blis + cpython 3.7: 1 failure
|
|
Time to close this, I think... Work continues in #172 |
Following the same scheme as #130, but for the 1.6 branch. Should not be merged due to conda/conda-build#3947.
Currently not building for pypy, pending resolution of #151, see also this comment.