Upgrade to Maxima 5.49 and ecl 26.3.27#41879
Upgrade to Maxima 5.49 and ecl 26.3.27#41879cxzhong wants to merge 1 commit intosagemath:developfrom
Conversation
|
@antonio-rojas please review this now. I have upgrade the conda-lock files |
|
Documentation preview for this PR (built with commit 8844bed; changes) is ready! 🎉 |
|
You need to make sure that system maxima < 5.48 is not accepted, because the And we also need to decide what to do with system 5.48. Functionally it should work, but some tests will fail with different output. So we either
|
This error is very strange |
I think we just ignore this. Because in fact, it is not mathematically wrong, just the expressions are different but they are equal |
|
It seems the bug of increment docker build. after we build new docker image, this will be fixed. |
The rebuild of docker is OK. |
|
@antonio-rojas I think this can be merged |
You still haven't addressed #41879 (comment). According to https://repology.org/project/maxima/versions there aren't many distros shipping 5.48, so maybe we can allow that with a few broken tests (although I'm not 100% happy about it), but you definitely need to exclude < 5.48. In any case, I can't approve this myself, it's 90% my code. |
So we block <5.48? |
|
|
What's with all the closing and reopening? Please note that every time you do that everybody gets spammed with two emails. |
Sorry for this. I do not know it will sent email. The ci is not stable. merge pr always fails |
|
It seems meson.build missing ecl's path caused docker build failed. I have editted it. @tobiasdiez |
|
This PR is completed |
|
CC: @dimpase This can be reviewed now. |
tobiasdiez
left a comment
There was a problem hiding this comment.
What exactly is the error that needs compilation with the ecl include dir now?
Increase docker build seems will incorrect, if we update ecl for different path include (before using system's ecl, now using sage's ecl). I think we just needs a warning about this. If the user do not counter the increase build for different path of ecl, they do not encounter this problem. |
|
The main problem is in test-long |
|
That looks more like a CI bug/shortcoming to me. If the CI / linux runs are okay, then I don't see any need to fix the test-long. Or is there any realistic scenario where users would run into the same issue? |
If you first compile with system's ecl, then second increasely compile with sage's ecl (change path). then it will cause problem. |
|
and I think it is a much special case. We just add a warning if we can not find ecl_config |
|
|
@nbruin - interface to ECL is being changed. Can you have a look? |
|
I do not edit ecl.pyx |
This seems also trigger on ecl 24.5.10. It is not a problem caused by ecl 26.3.27 |
I never saw this with 24.5.10. I'm seeing it occasionally on 26.3.27. Here's another issue with 26.3.27 that I never saw before: https://github.com/void-linux/void-packages/actions/runs/24376377943/job/71190688685?pr=59932#step:7:36197 |
In CI https://github.com/sagemath/sage/actions/runs/24401641135 |
- Upgrade Maxima to 5.49 and ECL to 26.3.27 - Enhance robustness of *maxima-objdir* directory creation in maxima_lib.py - Add warnings for fallback to ecl_lib in meson.build to improve ECL build reliability - Format meson.build - Align ECL first_env to 65536 bytes to fix flaky mprotect EINVAL crash (ECL 26.3.27 changed the main thread env from mmap-allocated to a static struct; mprotect() requires page alignment, which mmap() guaranteed but a static variable does not) Co-authored-by: arojas <arojas@archlinux.org>
|
Maybe this needs to report to ecl devel |
📝 Checklist
⌛ Dependencies