Upgrade cypari2 to 2.1.3

[kliem]

This is needed for python 3.11 support (#33842) and to make cypari compatible with current pari version

https://groups.google.com/g/sage-devel/c/jqmr6bDjDrk/m/XE2GlB_GBgAJ

Necessary follow ups:

• make cypari thread safe: see cypari2 #116
• remove optional build time dependency of cysignals on pari: see cypari2 #130

Depends on #33864

CC: @videlec @dimpase

Component: build

Author: Matthias Koeppe

Branch/Commit: 0ac9905

Reviewer: Dima Pasechnik

Issue created by migration from https://trac.sagemath.org/ticket/33878

[kliem]

Dependencies: #33864

[kliem]

Branch: public/33878

[kliem]

Commit: f3f1b5d

[kliem]

New commits:

92e9cff	build/pkgs/cython: Update to 0.29.30
f3f1b5d	test new changes to pari

[kiwifb]

comment:3

Is there any reason you are using snapshots from your fork of cypari2? Or is it just for testing and development?

[kliem]

comment:4

Just for testing and development. My own GitHub account is lots faster with the GitHub actions than sagemath. Also I use this trac ticket for cysignal's CI. Once cypari and possibly cysignal are ready, we switch back to the official sources.

[mkoeppe]

comment:5

Still getting

[cypari-2.1.3b1]   [7/7] Cythonizing cypari2/string_utils.pyx
[cypari-2.1.3b1]   cypari2/convert.c:3347:21: error: cannot take the address of an rvalue of type 'Py_ssize_t' (aka 'long')
[cypari-2.1.3b1]     __pyx_v_sizeptr = &Py_SIZE(((PyObject *)__pyx_v_x));
[cypari-2.1.3b1]                       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

with this ticket

[kliem]

comment:6

Sorry about that. I messed up with the tags. I'm rather confident the fix is working, it's just that the fix is not contained in the tag I created.

[kliem]

comment:7

The new commit contains the recent changes.

The test release contains basically the following:

• Python 3.10 compatibility cypari2#103 (requires cython 0.29.20, which videlec does not like: Python 3.10 compatibility cypari2#103 (comment);

  an alternative approach without raised cython requirement could be add test for python 3.11 cypari2#115)

• provide PARI_SIGINT_block/pending address for cysignals cypari2#109 (videlec doesn't like this approach much and rather wants cypari2 to push to some signal handling stack of cysignals)

• a PariThreadPool to handle multithreading via Python cypari2#116 (doesn't work for python 3.6 yet).

[sagetrac-git]

Changed commit from f3f1b5d to 481d782

[sagetrac-git]

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

481d782

test cypari update

[sagetrac-git]

Changed commit from 481d782 to b494326

[sagetrac-git]

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

11a0e57	build/bin/write-dockerfile.sh: Add bootstrap-conda
b494326	Merge branch 'u/mkoeppe/fix_tox_docker_builds' of git://trac.sagemath.org/sage into public/33878

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

0ac9905

build/pkgs/cypari: Update to 2.1.3

[sagetrac-git]

Changed commit from b494326 to 0ac9905

[videlec]

comment:16

Regarding the ticket description: to make pari threadsafe, the only way I explored so far was to use a "thread pool executor", see sagemath/cypari2#116. Because of its stack management, pari needs some care when used within threads.

[mkoeppe]

comment:18

I don't remember the details about this, but there is also a separate related ticket: #28800

[mkoeppe]

Author: Matthias Koeppe

[mkoeppe]

Reviewer: https://github.com/mkoeppe/sage/actions/runs/3448209182

[mkoeppe]

comment:21

Tests look OK, let's get this in

[dimpase]

comment:22

lgtm - tested with Pari 2.,15.1

[dimpase]

Changed reviewer from https://github.com/mkoeppe/sage/actions/runs/3448209182 to Dima Pasechnik

[mkoeppe]

comment:23

Thanks!

[vbraun]

Changed branch from public/33878 to 0ac9905