Proposal: downgrade some packages to experimental

[John H Palmieri]

The following packages do not build for me on OS X 12.5 (Monterrey), Intel chip:

• polylib (#33758)
• symengine_py (#34141)
• p_group_cohomology (#30787)
• r_jupyter
• rubiks
  

I propose that these be downgraded from optional to experimental until the issues are fixed. What do people think?

--

John

[Matthias Koeppe]

Also on various Linux platforms (https://github.com/sagemath/sage/actions/runs/2777704587) I see failures for

symengine_py, 

p_group_cohomology, 

r_jupyter

So these 3 definitely need to be changed to "experimental".

 

[John H Palmieri]

Also, the various sage_numerical_backend_* packages fail to build, with errors like

  Error compiling Cython file:
  ------------------------------------------------------------
  ...
  from sage.numerical.backends.generic_backend cimport GenericBackend
  ^
  ------------------------------------------------------------

  check_add_col_untyped_args.pyx:1:0: 'sage/numerical/backends/generic_backend.pxd' not found

Also downgrade? Is there a ticket for this?

[Isuru Fernando]

I haven't seen any reports about symengine.py failing to build. Can you open an issue?

Isuru

--
You received this message because you are subscribed to the Google Groups "sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/c6e5754e-68eb-4c9e-8848-ec23b51d63b3n%40googlegroups.com.

[Matthias Koeppe]

This one is https://trac.sagemath.org/ticket/34221

[Matthias Koeppe]

The symengine ticket is https://trac.sagemath.org/ticket/34141

[John H Palmieri]

Thank you for the link. With this ticket, the gurobi and cplex backends still fail because I don't have gurobi and cplex installed. It would be nice if that was detected earlier in the process and the build failed right away. Actually, it is detected early in the process:

  /bin/sh: gurobi.sh: No such file or directory
  GUROBI_HOME is not set, or it does not point to a directory with a Gurobi installation.  Trying to link against -lgurobi

but the installation proceeds anyway. But this is a separate issue.

[Matthias Koeppe]

The sage_numerical_backends_* packages are on their way out anyway.

Coin support will switch to go through CVXPy in https://trac.sagemath.org/ticket/34251, and Gurobi/CPLEX likewise in a follow up ticket

[Travis Scrimshaw]

I would like to see the p_group_cohomology fixed. Progress got stalled on that, but I am happy to work on that (while I have time for August).

Best,

Travis

[Marc Culler]

The p_group_cohomology package is included in the macOS binary release.  it requires a patch to its spkg-install,sh.  Symengine_py is also included in the macOS binary and also needs a patch.  I am attaching a tarball of the build/pkg directories for those two packages.

- Marc

[Marc Culler]

Yes, there is a ticket for this.  It is #34221.  The current version of Cython ignores .pxd files in directories with no __init__.py.  The ticket backports a cython fix for that problem which will appear in a future Cython release.

- Marc

[Marc Culler]

Incidentally, the patch for p_group_cohomology that I uploaded above also appeared in #30787 .

- Marc

[John H Palmieri]

I would like to see all of them fixed, p_group_cohomology especially, but I think we are coming up on a release and if the package doesn't build on supported platforms, we can't really call it optional. Once a package is fixed, then it should immediately get promoted back to optional again.

Regards,

  John

[Dima Pasechnik]

On Sun, 7 Aug 2022, 20:48 John H Palmieri, <jhpalm...@gmail.com> wrote:

I would like to see all of them fixed, p_group_cohomology especially,

there is a regression to solve with the package, which needs understanding of the algorithm behind. (build problems are solved, I think).

--

You received this message because you are subscribed to the Google Groups "sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/53c18b02-c5fa-4dcf-9dd0-6c57955ccdf4n%40googlegroups.com.

[Matthias Koeppe]

On Sunday, August 7, 2022 at 12:48:21 PM UTC-7 John H Palmieri wrote:

I would like to see all of them fixed, p_group_cohomology especially, but I think we are coming up on a release and if the package doesn't build on supported platforms, we can't really call it optional. Once a package is fixed, then it should immediately get promoted back to optional again.

+1