gh-130317: Skip test_pack_unpack_roundtrip_for_nans() on x86 #133155
Conversation
|
cc @skirpichev |
|
Example of failure with 32-bit float on x86:
There is a similar problem with I'm not sure why it works with |
There was a problem hiding this comment.
LGTM (as a temporary fix). Don't close issue.
0xf2 in _testcapi.float_pack() becomes 0xfa in PyFloat_Pack4(). The sNaN becomes a qNaN just by the calling convention, there is nothing that we can do.
I'm not sure why it works with PyFloat_Pack8().
It shouldn't, if your explanation is valid.
Hmm, my tests show that on 32-bit sNaN becomes qNaN after calling. Are you sure that size=8 tests works? |
|
!buildbot x86 |
|
🤖 New build scheduled with the buildbot fleet by @vstinner for commit 2d62e7b 🤖 Results will be shown at: https://buildbot.python.org/all/#/grid?branch=refs%2Fpull%2F133155%2Fmerge The command will test the builders whose names match following regular expression: The builders matched are:
|
It works on Python built with |
Reduce also the number of iterations from 1000 to 10 to ease debugging failures and prevent "command line too line" error when tests are re-run.
2d62e7b to
c39b3c7
Compare
I guess, optimizer avoid using FPU registers As a quick fix I suggest to disable testing in 32-bit mode. And also reduce number of tests as suggested, 10 should be enough. In a follow-up PR I'll do cleanup. Probably float_set_snan() is not helpful, as well as other attempts to workaround passing of sNaN. |
|
@skirpichev: I rewrote my PR to simply skip the test on x86 Debian, but not skip the test on x86 Windows. Please review the updated PR. |
|
!buildbot x86 Debian |
|
🤖 New build scheduled with the buildbot fleet by @vstinner for commit c39b3c7 🤖 Results will be shown at: https://buildbot.python.org/all/#/grid?branch=refs%2Fpull%2F133155%2Fmerge The command will test the builders whose names match following regular expression: The builders matched are:
|
|
Merged, thanks for the review. Sadly, I don't think that we can work around "compiler bugs" if sNaN becomes qNaN as soon as we call a simple function taking a |
Reduce also the number of iterations from 1000 to 10 to ease
debugging failures and prevent "command line too line" error when
tests are re-run.
nanfloats is non-invertible #130317