upgrade to pynac-0.3.9.1

[rws]

Pynac-0.3.9.1 has:

• fix for part of #18630: Expression.is_positive/negative incomplete
• many abs() additions and fixes (#12588)
• pynac now compiles with Python3 headers too (#15530)
• from GiNaC: Fix pow(+(...),2).expand() (#18568)
• from GiNaC: power::expand(): (x*p)^c -> x^c * p^c, if p>0
• merge numeric and Number_T classes

​https://github.com/pynac/pynac/releases/download/pynac-0.3.9.1/pynac-0.3.9.1.tar.bz2

[rws]

New commits:

​4457a3a

18537: pynac-0.3.9

[kcrisman]

I can't test the Python3 thing, or rather don't know how. I think it might be useful to have doctests for some of the changes, notably the expand and the abs stuff not in #12588 (there seem to be a few commits including from upstream). Do you anticipate any platforms on which there would be problems installing with these changes? (0.4.0 seems to have more clang stuff so not sure if that would impact Mac or not, since hopefully by the time we hit Pynac only gcc would be in use.)

Waiting on documentation to build... sigh.

[kcrisman]

See also my comments on #12588. Patchbot isn't speaking here so I will run long doctests on my own, but just on an older Mac.

[kcrisman]

All tests pass. I guess I'd want someone to look at the Pynac-specific changes, since I don't think anyone has reviewed them yet - I may get to that, but don't have time today.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​466837c	Merge branch 'develop' into t/18537/upgrade_to_pynac_0_3_9
​ac542da	18537: bump to 0.3.9.1; fix doctest

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​9c092c7

18537: doctest for a GiNaC improvement

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​7590e94

18257: previous doctest should reset variable domains

[jdemeyer]

Found local metadata for pynac-0.3.9.1
Invalid checksum for cached file /usr/local/src/sage-git/upstream/pynac-0.3.9.1.tar.bz2, deleting

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​57380ae

18537: reset checksum

[jdemeyer]

Can you explain why the checksum was wrong before? I'm suspicious of checksums changing without justification...

[rws]

I was too so I compared: the sources were correct. It's not the first time either. But no final explanation atm.

[rws]

What I can say is that the github file matches exactly what I get locally with make dist, so the checksum back then was from something different.

[rws]

The most likely explanation is that my work process (which gives bleeding edge tarballs locally the same version as the last release), because of switching back and forth between last release and development, lead to a development tarball being the base of the release checksum commit. The change that would least affect the speed of the process would be to double-check before checksumming for the release.

[vbraun]

lgtm