Upgrade to givaro-4.2.0 fflas-ffpack-2.5.0 linbox-1.7.0 #32959

ClementPernet · 2021-12-02T10:49:33Z

New packages are available here (with the corresponding checksums):

givaro: https://github.com/linbox-team/givaro/releases/tag/v4.2.0

fflas-ffpack: https://github.com/linbox-team/fflas-ffpack/releases/tag/v2.5.0

linbox: https://github.com/linbox-team/linbox/releases/tag/v1.7.0

The new versions require "a C++ compiler supporting C++11 standard. More precisely g++ v5 or greater, clang++ v3.4 or greater".
So this ticket depends on dropping support for some systems, see #32074.

Depends on #33316

CC: @kiwifb @orlitzky

Component: packages: standard

Branch/Commit: u/cpernet/upgrade_to_givaro_4_2_0_fflas_ffpack_2_5_0_linbox_1_7_0 @ 4006e10

Issue created by migration from https://trac.sagemath.org/ticket/32959

The text was updated successfully, but these errors were encountered:

mkoeppe · 2021-12-02T18:32:55Z

comment:3

I see numerous build errors in https://github.com/linbox-team/fflas-ffpack/runs/4383290866?check_suite_focus=true that will need fixing before we can do these updates

ClementPernet · 2021-12-03T08:35:38Z

comment:5

Of course. I just created this ticket to start working on a its branch, which will help us fix the last issues before cutting the final releases.
Btw, I did not manage to access to the full artifact of the givaro install log givaro which caused the failures in these github action builds (only the tail of the givaro-log is displayed in the sage build log in your link). Is there any way to access it? (I could not find it in the whole log archive).

ClementPernet · 2021-12-03T10:03:26Z

Branch: u/cpernet/upgrade_to_givaro_4_2_0_fflas_ffpack_2_5_0_linbox_1_7_0

ClementPernet · 2021-12-03T14:13:09Z

New commits:

`55bd2fa`	`update to givaro-4.2.0rc0`
`be8b20d`	`fflas-ffpack-2.5.0rc1`
`bd36004`	`fflas-ffpack-2.5.0rc1 fingerprints`

ClementPernet · 2021-12-03T14:13:09Z

Commit: bd36004

mkoeppe · 2021-12-03T18:27:43Z

comment:8

Replying to @ClementPernet:

I did not manage to access to the full artifact of the givaro install log givaro which caused the failures in these github action builds (only the tail of the givaro-log is displayed in the sage build log in your link). Is there any way to access it? (I could not find it in the whole log archive).

In each run, such as https://github.com/linbox-team/fflas-ffpack/runs/4383290866?check_suite_focus=true, there is a section "Print out logs for immediate inspection" in the output, which shows all logs of failing package builds in full.

When all jobs have completed, the summary page, such as https://github.com/linbox-team/fflas-ffpack/actions/runs/1526437644, gives access to the log artifacts.
For example, logs-commit-3e42b4a1aaa585bce832738d5a8885dcf7c4273e-tox-docker-ubuntu-xenial-standard (https://github.com/linbox-team/fflas-ffpack/suites/4521078418/artifacts/121757925) gives a zip file within which you will find logs/pkgs/givaro-git.log

mkoeppe · 2021-12-03T18:28:33Z

comment:9

We have a bit of documentation on this at https://doc.sagemath.org/html/en/developer/portability_testing.html#automatic-parallel-tox-runs-on-github-actions; if you feel anything needs to be added, let me know

ClementPernet · 2021-12-04T08:44:35Z

comment:10

Thanks a lot for the pointers.

sagetrac-git · 2021-12-07T13:13:36Z

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

`6eca64f`	`fix with LinBox new API`
`5a968ca`	`LinBox new API for DenseMatrix`
`78308b7`	`checksums`

sagetrac-git · 2021-12-07T13:13:36Z

Changed commit from bd36004 to 78308b7

mkoeppe · 2021-12-08T00:31:43Z

comment:13

The checksums.ini files should get upstream_url lines, see https://doc.sagemath.org/html/en/developer/packaging.html#upstream-urls

sagetrac-git · 2021-12-14T10:30:22Z

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

`065177a`	`update checksums and versions to latest released linbox, fflas-ffpack and givaro`

sagetrac-git · 2021-12-14T10:30:22Z

Changed commit from 78308b7 to 065177a

ClementPernet · 2021-12-14T10:42:36Z

comment:15

I just released the three libraries and updated this branch with the corresponding checksums, and added the upstream_url as requested.

I could check that the testsuite passes on 2 machines (broadwell and skylake-X both running Linux mint).

I only had to comment out local/include/factory/factory.h line 92 (#define IntegerDomain 1), because this define by Singular gets in the way of Givaro's Givaro::IntegerDomain. See https://groups.google.com/g/sage-devel/c/lC941Lab5ck

Dima suggested to pacth this file, but I am not sure to see where this should be done in sage install process. Any help would be most welcome.

dimpase · 2021-12-14T12:03:26Z

comment:16

Does Singular/Singular@18bb792
do the job? Or they made an error there, it doesn't fix anything for you?

dimpase · 2021-12-14T12:10:45Z

comment:17

Replying to @dimpase:

Does Singular/Singular@18bb792
do the job? Or they made an error there, it doesn't fix anything for you?

Note that the patched file is actually meant to be preprocessed by Singular's ./factory/bin/makeheader

orlitzky · 2021-12-14T13:12:14Z

comment:18

Replying to @ClementPernet:

Dima suggested to pacth this file, but I am not sure to see where this should be done in sage install process. Any help would be most welcome.

I don't think there's such an easy solution. The straightforward one is to wait for a new singular release, and then upgrade singular and givaro in sage at the same time, after enough distributions have packaged them both.

mkoeppe · 2021-12-14T16:38:33Z

comment:19

I would suggest to work around this clash in Givaro.

mkoeppe · 2021-12-14T16:44:34Z

comment:20

The latest run of the integration testing workflow - https://github.com/linbox-team/fflas-ffpack/runs/4511137082?check_suite_focus=true - fails badly on all platforms

ClementPernet · 2021-12-14T17:31:54Z

comment:22

Replying to @mkoeppe:

The latest run of the integration testing workflow - https://github.com/linbox-team/fflas-ffpack/runs/4511137082?check_suite_focus=true - fails badly on all platforms

This run does not correspond to the released package and the failure there has been fixed in a more recent commit which is in the release.

mkoeppe · 2021-12-14T17:35:11Z

comment:23

You may want to cancel the runs that are no longer needed so that the current one gets a chance to run

ClementPernet · 2022-02-07T07:41:05Z

comment:42

Replying to @mkoeppe:

This looks like a good solution for now, but this will become an obstacle to modularization pretty soon.

Should I revert this commit, as the fixed Singular is now released ? I am not sure what's the best move to get thinks rolling. Just let me know.

mkoeppe · 2022-02-07T16:48:39Z

comment:43

This would depend on

actually making the Singular update (update Singular to 4.3.1 #33160, which needs help)
tightening our version requirement in build/pkgs/singular/spkg-configure.m4

mkoeppe · 2022-02-07T16:52:22Z

comment:44

By the way, I've marked this ticket for Sage 9.7 because of the increased compiler requirements (see #32074).

Also before we can merge this upgrade, Cygwin support needs to repaired (https://github.com/linbox-team/fflas-ffpack/runs/4658698318)

mkoeppe · 2022-02-07T16:54:00Z

comment:45

Also, 32bit platforms seem broken

mkoeppe · 2022-02-08T21:23:05Z

Dependencies: #33316

mkoeppe · 2022-08-25T20:16:13Z

comment:48

@ClementPernet Is it planned to release new versions of the packages that include patches you may have received from distribution packaging?

ClementPernet · 2022-08-26T12:15:39Z

comment:49

Yes, I'm finally getting back to it, and would like to cut a minor version in a near future (say september or october) fixing the various issues raised here and in other PRs.

isuruf · 2023-04-27T20:05:31Z

@ClementPernet, any plans for a new minor version?

@ClementPernet