On 2020-07-10 02:55, Thorsten Glaser wrote:
...
This still fails, even on x32, which is an x86 architecture õÕ
I’ve commented upstream. The funniest thing is that, when I run
the test that fails inside the build chroot, manually, (just with
more debugging output) it passes!
https://github.com/pyFFTW/pyFFTW/issues/128#issuecomment-656290711
I therefore propose to skip the problematic tests on most
architectures.
Another interesting difference is that the build has 2 failues
whereas my local build has 12 so the binary build might be
dependent on the system it was built on, which is… dangerous,
and, for a binary distribution, obviously inappropriate.
It's tricky problem to resolve. Let's encourage upstream to take up a
guest account on our Debian test machines so he can probe directly.
In the mean time I can skip tests as a work-around, and see if problems
show up in actual usage.
Drew
