Re: qhull external symbols clash

On Thu, Jan 12, 2023 at 1:33 AM Sam Gillingham <[email protected]>
wrote:

> Ralf,
>
> I just stumbled across this: https://github.com/scipy/scipy/issues/15996.
>
> Would this have the potential to resolve this issue?


Probably - it will remove SciPy's global symbols, but I'm not quite sure
what happens if another package already loads its global symbols first.
Also, as mentioned in
https://github.com/scipy/scipy/pull/17052#issuecomment-1251395901, the
Apple linker works differently and we don't implement symbol hiding for it.


> I just checked with a build of main and all the qhull symbols are now
> local so AFAICT this shouldn't cause a problem. They are still global
> symbols in scipy 1.10.0 however, so maybe we need to wait for 1.11?
>

The relevant patches made it into 1.10.0. This difference is most likely
due to conda-forge packages still being built with distutils rather than
with Meson, and those distutils builds doing less extensive symbol hiding.
Or you're on macOS and looking at a build with the Apple linker. If you
could find out what the difference you're seeing is caused by, that'd be
very useful.

Cheers,
Ralf


> Sorry for the noise, hopefully this can be resolved without too much pain
> and without scrambled function names!
>
> Cheers,
> Sam.
> _______________________________________________
> SciPy-Dev mailing list -- [email protected]
> To unsubscribe send an email to [email protected]
> https://mail.python.org/mailman3/lists/scipy-dev.python.org/
> Member address: [email protected]
>

_______________________________________________
SciPy-Dev mailing list -- [email protected]
To unsubscribe send an email to [email protected]
https://mail.python.org/mailman3/lists/scipy-dev.python.org/
Member address: [email protected]