On 2020/12/29 20:34, Stuart Henderson wrote:
> On 2020/12/29 15:36, Stuart Henderson wrote:
> > On 2020/12/28 16:39, Daniel Dickman wrote:
> > > On Mon, Dec 28, 2020 at 4:12 PM Kurt Mosiejczuk <k...@cranky.work> wrote:
> > > >
> > > > Here's an update for Python 3.8 to 3.8.7
> > > >
> > > > https://docs.python.org/release/3.8.7/whatsnew/changelog.html#python-3-8-7
> > > >
> > > > As usual, all tests but the asyncio test complete successfully. (asyncio
> > > > never completes).
> > > >
> > > > (cc maintainer)
> > > >
> > > 
> > > sadly this will break packaging for all python ports with shared
> > > libraries to the change they made for shared libraries :-(
> > 
> > 3.9 will have the same in the next release.
> > 
> > There are approx 178 ports to fix, they'll just need plist regen and
> > bumping (there is already an appropriate variable in python.port.mk).
> > I'm working on it.
> 
> If anyone has time to look at one that isn't just a regen;
> 
> ===> databases/xapian-bindings
> ===>  Building for xapian-bindings-1.4.17
> make  all-recursive
> Making all in .
> Making all in python3
> make  all-am
> mkdir -p xapian
> /usr/bin/libtool --config | grep -v pic > libtoolconfig.tmp
> . ./libtoolconfig.tmp; cp $objdir/_xapian.cpython-38.so xapian
> cp: .libs/_xapian.cpython-38.so: No such file or directory
> *** Error 1 in python3 (Makefile:1293 'xapian/_xapian.cpython-38.so')
> 

So that one is interesting in that it still produces files with just .so
but tries to use .cpython-38.so in parts of the build.

There are more that still produce .so (which presumably won't really work)
but build/packaging doesn't actually fail - the build is ongoing but some
examples of these

textproc/libxml
devel/subversion
editors/py-qscintilla
sysutils/libolecf
sysutils/libregf
sysutils/libsigscan (and i think most/all of the other sysutils/lib*
  with python components)

I'll certainly have more of these by the time it's done.

Some (but not all) of these use swig so maybe an update/change there
might help those (the version of swig that we have isn't supposed
to support py3.8 anyway).

Reply via email to