-
-
Notifications
You must be signed in to change notification settings - Fork 526
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Upgrade giac to 1.6.0-47 #30537
Comments
comment:1
In Debian we already have giac 1.6.0-7 and one doctest either times out or segfaults. If it does not time out it looks as below. With the giac 1.5.0-87 Debian package the test works fine.
|
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
comment:6
1.6.0-17 is out. And my goodness they finally moved from configure.in to configure.ac. |
comment:7
After the update, it should be checked whether the workarounds for underlinked libgiac on Cygwin (#30396) are still necessary. |
comment:9
Build fails on macOS with
Hoping that upstream can improve the autoconf tests a bit by the time that we are ready to do the giac 1.6 upgrade in the Sage 9.3 cycle. We also have a number of patches in New commits:
|
Commit: |
comment:10
Replying to @mkoeppe:
To browse them online, see for instance one of these: |
comment:11
I have a failure when I ran the testsuite on Gentoo
I think it is a bit concerning. |
comment:12
And by the way, this is the same test for which we already have a small patch for test results with pari 2.11+ https://github.com/sagemath/sage-prod/blob/develop/build/pkgs/giac/patches/pari_2_11.patch |
comment:13
1.6.0-23 is out. The problematic test is still there for me. |
comment:14
Can you reproduce the same crash with 1 thread (export GIAC_THREADS=1) |
comment:15
I am not using a threaded pari.
|
comment:16
But maybe it's a bug in giac despite the error message. I don't get a crash in a xubuntu 20.04 VM with 1 thread (pari version is 2.11.2) |
comment:17
coredump
|
comment:18
I can no longer reproduce the crashes mentioned in comment:1 with 1.6.0-25. The tests still fail, though
|
comment:19
I still have the crash in giac's testsuite with 1.6.0.25. |
comment:20
Replying to @kiwifb:
I can't reproduce that one. I only have one test failure in the TP11 test suite (the already known failure mentioned in comment:12) |
comment:21
Replying to @antonio-rojas:
Is your pari using threads? |
comment:22
Replying to @kiwifb:
Yes |
comment:23
Replying to @kiwifb:
What do you know. You now need to have pari compiled with threads for that test to pass - and probably some specific calls to pari too. |
comment:24
merge failed |
comment:25
See #30868 for an issue with giac 1.6 that has affected Fedora packaging |
comment:26
We seem to have stumbled across the issue mentioned in comment 23: with giac >= 1.6.0-25 a threaded pari is necessary. |
comment:91
Replying to @kiwifb:
I think we should |
comment:92
OK, the real issues between the two are the doctests in A different test should fix this. The formatting is problematic. |
Branch pushed to git repo; I updated commit sha1. New commits:
|
comment:94
OK, so the last bit of difference seems to be that internal representation in giac has changed. Looking at the larger context of the doctest
The line |
Branch pushed to git repo; I updated commit sha1. New commits:
|
comment:96
Your changes look good to me |
Reviewer: ..., Matthias Koeppe |
comment:97
OK, so for that last line which will differ between 1.5.0 and 1.6.0. Pool:
|
comment:98
either way is fine with me, or make it |
comment:99
That is a viable solution, I will do it now and then we can call it done. |
Branch pushed to git repo; I updated commit sha1. New commits:
|
Changed reviewer from ..., Matthias Koeppe to François Bissey, Matthias Koeppe |
comment:101
I approve of your changes :) do we want an independent person to chime in? |
comment:102
I think we can set it to positive review |
comment:103
For the record, the giac testsuite fails on KuCalc (Debian 9 x86_64)
it has always been flaky on that buildbot, but now seems to fail consistently. I'll exclude running the giac testsuite on kucalc from now on. |
comment:104
I'd have to double check that. Some (mathematical) tests return different results depending on whether you build the GUI (linked to fltk) or not. The results distributed with giac are done with GUI. I officially gave up on patching for tests without GUI in sage-on-gentoo for giac-1.6.0-x, because it would have been a massive patch. Very much like what would be needed given your results above. I just cannot ascertain that it was |
Changed branch from u/fbissey/giac-1.6.0.47 to |
Changed commit from |
comment:108
Follow-up - speed regression because pari is forced to be --mt=pthread - in #31572. |
Source:
Previous updates:
See also:
spkg-configure.m4
for giacTarball in attachment
https://github.com/sagemath/sage/files/ticket30537/giac-1.6.0.47p2.tar.bz2.gz
CC: @slel @frederichan-IMJPRG @sagetrac-parisse @antonio-rojas @jamesjer @videlec @orlitzky
Component: packages: standard
Keywords: giac
Author: Matthias Koeppe, François Bissey
Branch:
f05720b
Reviewer: François Bissey, Matthias Koeppe
Issue created by migration from https://trac.sagemath.org/ticket/30537
The text was updated successfully, but these errors were encountered: