Re: qhull external symbols clash
Ralf Gommers <[email protected]>
| Newsgroups | gmane.comp.python.scientific.devel |
|---|---|
| Message-ID | <CABL7CQiN7g+xtASGMQHbn7iTT9B5KZHy1vxz3Y++-n-ZaNgqvg@mail.gmail.com> |
On Fri, Jan 13, 2023 at 7:16 AM Sam Gillingham <[email protected]> wrote: > Ralf, > > You are right about 1.10.0 - I was using the conda build which still uses > distutils. I'm on Ubuntu 22.04. > > I just created a little C project which exercises the local symbols > functionality: https://github.com/gillins/symbols_check > For me (on Linux) it seemed to confirm whichever symbol gets loaded first > is used - unless it marked 'local' in case it only seen by the library that > uses it internally (whatever the order of loading). Can you confirm this is > as you expect? > Yes, that sounds right. And I can confirm that your test code works for me. > So it seems that OSX also supports this but it is not implemented in > SciPy? Seems that -exported_symbols_list on OSX supports wildcards ( > https://www.unix.com/man-page/osx/1/ld/). Would this be a good solution > for OSX as it will work for all vendored libs, plus make importing new > versions of these vendored libs easier (no effort mangling names as in the > GDAL link above etc)? Unfortunately, I no longer have access to an OSX > machine, so probably not much help here in testing. > As I posted in https://github.com/gillins/symbols_check/issues/2, there doesn't seem to be an issue on macOS. I'm out of time for now, if anyone knows the explanation here that'd be great. Otherwise I'll do some digging later. Cheers, Ralf > If anyone is reading this and has access to an OSX machine I'd appreciate > hearing how https://github.com/gillins/symbols_check works (see OSX > instructions in the README) to build and run the programs. Unfortunately > this is untested so I've probably got something simple wrong... > > Interested in your thoughts, > 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]