alpine linux: Add system package information / tox.ini, fix flint, giac, cvxopt

[mkoeppe]

To test: tox -e docker-alpine-standard

Depends on #33070
Depends on #33138

Upstream: Reported upstream. No feedback yet.

CC: @dimpase @slel @frederichan-IMJPRG @sagetrac-parisse @dkwo @tornaria @orlitzky

Component: porting

Author: Matthias Koeppe, Dima Pasechnik

Reviewer: Dima Pasechnik

Issue created by migration from https://trac.sagemath.org/ticket/33083

[mkoeppe]

Branch: u/mkoeppe/system_package_information___tox_ini___gh_actions_for_alpine_linux

[mkoeppe]

Commit: 1a49d24

[mkoeppe]

New commits:

1a49d24

tox.ini, build/bin/write-dockerfile.sh, build/bin/sage-print-system-package-command: Add alpine

[mkoeppe]

Author: Matthias Koeppe

[sagetrac-git]

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

d1a9266

Add more alpine.txt

[sagetrac-git]

Changed commit from 1a49d24 to d1a9266

[mkoeppe]

comment:5

  [r-3.6.3]   ********************************************************************************
  [r-3.6.3]   /sage/build/bin/sage-dist-helpers: line 143: fmt: command not found
  [r-3.6.3]   ********************************************************************************

[sagetrac-git]

Changed commit from d1a9266 to 522c452

[sagetrac-git]

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

522c452

build/pkgs/_prereq/distros/alpine.txt: Add coreutils

[sagetrac-git]

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

f3e59a6

build/pkgs/{graphviz,qhull,r}/distros/alpine.txt: Add non-dev packages

[sagetrac-git]

Changed commit from 522c452 to f3e59a6

[mkoeppe]

comment:8

flint

In file included from /sage/local/var/tmp/sage/build/flint-2.7.1/src/thread_support.h:16,
                 from /sage/local/var/tmp/sage/build/flint-2.7.1/src/nmod_mat.h:35,
                 from /sage/local/var/tmp/sage/build/flint-2.7.1/src/nmod_poly.h:34,
                 from /sage/local/var/tmp/sage/build/flint-2.7.1/src/fmpz_poly.h:35,
                 from /sage/local/var/tmp/sage/build/flint-2.7.1/src/aprcl.h:18,
                 from config_gauss.c:12:
/sage/local/var/tmp/sage/build/flint-2.7.1/src/thread_pool.h:55:5: error: unknown type name 'cpu_set_t'
   55 |     cpu_set_t original_affinity;
      |     ^~~~~~~~~

giac:

In file included from History.cc:8:
first.h: In function 'float fgamma(float)':
first.h:546:39: error: 'gammaf' was not declared in this scope; did you mean 'tgammaf'?
  546 | inline float fgamma(float f1){ return gammaf(f1); } // or tgammaf(f1) on some versions of emscripten
      |                                       ^~~~~~
      |                                       tgammaf
make[5]: *** [Makefile:976: History.lo] Error 1

[dimpase]

comment:9

there is also mpfr-dev for mpfr.

[dimpase]

comment:10

see flintlib/flint#1054 for Flint error

Probably as a workaround on can put -D_GNU_SOURCE into CFLAGS.

[dimpase]

Upstream: Reported upstream. No feedback yet.

[dkwo]

comment:12

That's indeed what we do on Void linux, see https://github.com/void-linux/void-packages/blob/master/srcpkgs/flintlib/template

[sagetrac-git]

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

095569b	build/pkgs/mpfr/distros/alpine.txt: New
d7b89f1	build/pkgs/flint/spkg-install.in: Use -D_GNU_SOURCE

[sagetrac-git]

Changed commit from f3e59a6 to d7b89f1

[mkoeppe]

comment:14

The giac build error was previously mentioned in

• https://www.mail-archive.com/sage-devel@googlegroups.com/msg100176.html
• https://groups.google.com/g/sage-devel/c/kiEh8JgbWeQ/m/Oqe77sQTCAAJ

[mkoeppe]

comment:15

From https://www.gnu.org/software/libc/manual/html_node/Special-Functions.html:

Function: double gamma (double x)
Function: float gammaf (float x)
Function: long double gammal (long double x)
Preliminary: | MT-Unsafe race:signgam | AS-Unsafe | AC-Safe | See POSIX Safety Concepts.

These functions exist for compatibility reasons. They are equivalent to lgamma etc. It is better to use lgamma since for one the name reflects better the actual computation, and moreover lgamma is standardized in ISO C99 while gamma is not.

[mkoeppe]

comment:16

Has this patch been submitted to upstream giac?

• https://github.com/void-linux/void-packages/blob/master/srcpkgs/giac/patches/fgamma.patch

[mkoeppe]

comment:17

There seems to be some confusion between gamma and log of gamma here.

[dkwo]

comment:18

Does this test pass on Alpine?

sage -t /usr/lib/sage-9.5.beta9/src/sage/matrix/matrix_mod2_dense.pyx # 1 doctest failed

It's the last one failing on Void with musl 1.2.2,
where it gives

cysignals.signals.SignalError: matrix allocation failed

instead of

RuntimeError: matrix allocation failed

so I wonder if the result is the same on Alpine,
and whether possibly the test result should be updated.

[dimpase]

comment:51

Replying to @mkoeppe:

Replying to @dimpase:

Replying to @sagetrac-git:

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

455e1fa	build/pkgs/numpy/spkg-install.in: Use pip wheel --no-use-pep517
1fe1e85	Merge #33138

This gives a different error

Works fine here. How are you testing this?

running Apline in an lxc container.
Using their Python 3.10, their openblas - I guess that's the only pieces of the config puzzle that matter.

That's the same error as I get while trying to build numpy wheel without Sage using pip,
just by running pip wheel -e . in the unzipped numpy distro. (same error on their master from the repo, too),

[mkoeppe]

comment:52

Your installation is broken somehow

[dimpase]

comment:53

Maybe, but "No such file or directory" error ought to be traceable...

[dimpase]

Reviewer: Dima Pasechnik

[dimpase]

comment:54

OK, let's get this in. (the lxc container setup might be to blame for the numpy issue I keep seeing)

[mkoeppe]

comment:55

Looks like the added flags break the build of arb on Debian variants - https://github.com/mkoeppe/sage/runs/4785815685?check_suite_focus=true

[mkoeppe]

comment:57

Replying to @orlitzky:

Replying to @mkoeppe:

This problem is again - our build/pkgs/suitesparse/spkg-configure.m4 finds suitesparse but forgets to tell anyone where it was found.

Suitesparse found in the system library/header path but cvxopt tries to guess it anyway?

This has been reported a few times,

• try using pkgconfig in setup.py to determine various libs and dirs cvxopt/cvxopt#156
• cannot build on ARM Debian cvxopt/cvxopt#174

For now we'll probably be adding another special case to

https://github.com/cvxopt/cvxopt/blob/master/setup.py#L55

See also #31905

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

2575abf	tox.ini, build/bin/write-dockerfile.sh, build/bin/sage-print-system-package-command: Add alpine
fc4ea60	Add more alpine.txt
8766a06	build/pkgs/_prereq/distros/alpine.txt: Add coreutils
0ebfc9e	build/pkgs/{graphviz,qhull,r}/distros/alpine.txt: Add non-dev packages
b3abd3b	build/pkgs/mpfr/distros/alpine.txt: New
30988ea	build/pkgs/flint/spkg-install.in: Use -D_GNU_SOURCE
829a7d0	mpc
d3ef3d4	correct flags of flint and arb for musl libc
bd34cd5	build/pkgs/giac/spkg-install.in: Add workaround for musl libc

[sagetrac-git]

Changed commit from 1fe1e85 to bd34cd5

[sagetrac-git]

Changed commit from bd34cd5 to f453452

[sagetrac-git]

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

f453452

Merge tag '9.7.beta4' into t/33083/system_package_information___tox_ini___gh_actions_for_alpine_linux

[mkoeppe]

Removed branch from issue description; replaced by PR #35285