-
Notifications
You must be signed in to change notification settings - Fork 2.1k
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
sagemath: update to 10.1. #45708
sagemath: update to 10.1. #45708
Conversation
I know how to fix musl. This is an issue in giac with musl libc:
It's already ack by upstream and it will be fixed, meanwhile I'll include a workaround so it works for us. |
The i686 failure seems transient 🤷 I still want to run this locally a few more days, I'll rerun CI before we merge. |
Looks good to me, thanks. I won't be able to run tests on aarch64 for three more weeks. |
As reported in void-linux#45708, including <giac/giac.h> fails in musl because the HAVE_TGAMMAF macro from autotools is not available at this time. Here we include a patch adapted from void-linux#45708 (comment) which is the solution upstream will include in next release of giac. We need to fix this now so we can build sagemath in musl!
As reported in void-linux#45708, including <giac/giac.h> fails in musl because the HAVE_TGAMMAF macro from autotools is not available at this time. Here we include a patch adapted from geogebra/giac@618a5de which is the solution upstream will include in next release of giac. We need to fix this now so we can build sagemath in musl!
Sure. How would that work? What would we gain? I imagine:
As a matter of fact, I don't see what would be the point of An important point is: how would we test that the "depends" line is correct (i.e. that we don't remove something that is needed, e.g., to pass the testsuite). |
As reported in void-linux#45708, including <giac/giac.h> fails in musl because the HAVE_TGAMMAF macro from autotools is not available at this time. Here we include a patch adapted from geogebra/giac@618a5de which is the solution upstream will include in next release of giac. We need to fix this now so we can build sagemath in musl!
We could remove
and then see which tests fail from within installed sage; |
Nothing will fail inside the build chroot, since as you said, these are already installed via (host)makedepends. That's precisely the issue I am worried about. The only easy way I see is to have a separate meta package to test sagemath, with a checkdepends on sagemath, which runs the sagemath testsuite (on the installed package). |
As reported in #45708, including <giac/giac.h> fails in musl because the HAVE_TGAMMAF macro from autotools is not available at this time. Here we include a patch adapted from geogebra/giac@618a5de which is the solution upstream will include in next release of giac. We need to fix this now so we can build sagemath in musl!
As reported in void-linux#45708, including <giac/giac.h> fails in musl because the HAVE_TGAMMAF macro from autotools is not available at this time. Here we include a patch adapted from geogebra/giac@618a5de which is the solution upstream will include in next release of giac. We need to fix this now so we can build sagemath in musl!
@dkwo all dependencies you singled out are needed to use either cython or fortran in sagemath, which are not optional (at least the testsuite fails without them). I can imagine both features to be made optional, but I'm wary, something needs to happen so these are made optional and the testsuite works fine (a "feature" is detected, etc). What is the need / advantage of this? I guess some download or installed size, but how much? Is it really worth it? |
7dae82a
to
831d5d0
Compare
I think this will become possible soon, but not yet. The sagemath doctests are adding "feature flags" which detect features in the environment, but this is not yet completely polished for cython / external libraries / fortran code. At this time, I'd rather focus on getting sagemath working on cython 3 so we can update cython (see #45086 and ahesford#2). After 10.2 or 10.3 it may be easier to do this. In any case, I'd like to see a detail of how much is saved (say how much download/disk it takes now an install of sagemath with all deps not standard in void vs. what would it take if we get this "savings"). If the savings is not really very significant, I'm not sure I see the point... |
@leahneukirchen @dkwo This is ready to merge IMO. Note that my last push is after the last big update (boost, python, ssl, etc) so CI here is tested with the up-to-date packages in the repo. |
I agree, I can do more experiments for one of the next point releases. |
Testing the changes
I will install and test a little bit. I'm pushing now to get an idea if CI works ok.
CC: @dkwo