BUG: detect buggy windows version and raise at import#17553
Conversation
|
Does what it says. |
Co-authored-by: Eric Wieser <wieser.eric@gmail.com>
Co-authored-by: Eric Wieser <wieser.eric@gmail.com>
|
Maybe you need to make the exception more explicit? |
|
@bashtage if you could try it again that would be great. There were various typos and mistakes in the original version. |
|
@mattip Still does the job. |
|
Maybe this is OK to merge, even if also gh-17547 gets in? Import-time sanity check probably can't hurt, in case similar bug gets reintroduced in another Windows update. |
|
Since OpenBLAS 0.3.12 avoids using the messed-up registers, this check is no longer relevant. Closing. |
| _win_os_check() | ||
|
|
||
| del _win_os_check | ||
|
|
There was a problem hiding this comment.
I think there are some hard tabs in here.
| msg = ("The current Numpy installation ({!r}) fails to " | ||
| "pass a sanity check due to a bug in the windows runtime. " | ||
| "See this issue for more information " | ||
| "https://developercommunity.visualstudio.com/content/problem/1208774/fpu-exception-in-fmod0-x-in-windows-10-version-200.html") |
There was a problem hiding this comment.
The URL is correct, despite missing the 4.
|
Hi, I came here from a pip installation that is refusing to work now, looking for a workaround, but did not find anything. Am I correct in saying, that this is not a problme on 32 bit python? If so, I would ask you to change the error message so that it links to this wiki page I just made instead of to the visual studio bugtracker. |
|
You are correct that there is no issue on 32 but windows. The best workarounds are to either use WSL or 32 bit. Neither has the fmod bug. |
xref gh-16744. Use a check at import to make sure we are not sensitive to the fmod bug in windows. Alternative to gh-17547.
@bashtage could you check if these PRs works as advertised?