build/pkgs: Split out spkg liblzma from xz

[mkoeppe]

(from #30559)

$ grep xz build/pkgs/*/dependencies
build/pkgs/cmake/dependencies:curl zlib bzip2 xz
build/pkgs/gcc/dependencies:| $(MP_LIBRARY) mpfr mpc zlib xz
build/pkgs/gdb/dependencies:mpfr zlib ncurses $(PYTHON) xz
build/pkgs/gfortran/dependencies:| $(MP_LIBRARY) mpfr mpc zlib xz
build/pkgs/p_group_cohomology/dependencies:$(PYTHON)... | $(PYTHON_TOOLCHAIN) matplotlib gap xz $(SAGERUNTIME) ipywidgets
build/pkgs/python3/dependencies:zlib readline sqlite libpng bzip2 xz libffi
build/pkgs/r/dependencies:$(BLAS) gfortran iconv readline bzip2 xz pcre curl | pkgconf

In this list, cmake, r, python3 all depend on liblzma.
The others are merely build-time dependencies on the xz binary (for unpacking the tarball in sage_bootstrap.uncompress.tar_file - when sage-bootstrap-python is < 3.3 - see https://docs.python.org/3/library/tarfile.html) and have nothing to do with the shared library.

The new liblzma package is a script package without installation steps; it delegates the actual installation to the xz spkg. This follows the same approach as the libnauty/nauty packages.

Depends on #31409
Depends on #30350

CC: @dimpase @jhpalmieri

Component: packages: standard

Author: Matthias Koeppe

Branch/Commit: 1fb1fa9

Reviewer: John Palmieri

Issue created by migration from https://trac.sagemath.org/ticket/30948

[jhpalmieri]

comment:2

Python should be able handle uncompressing .tar.xz files, so maybe we should modify the SageTarXZFile class in sage_bootstrap.uncompress.tar_file: only use it if Python can't already handle .xz files.

[jhpalmieri]

comment:3

We could test

hasattr(tarfile.TarFile, 'xzopen')

for example.

[jhpalmieri]

comment:4

I guess this shouldn't be necessary: sage_bootstrap.uncompress.action.open_archive should check to see if tarfile.is_tarfile(FILE) is True first, and it should be for xz files with a supported version of Python. So I still don't know what happened in Volker's case: using an old Python?

[mkoeppe]

comment:5

Setting new milestone based on a cursory review of ticket status, priority, and last modification date.

[mkoeppe]

comment:6

If r is coming from the system or --disable-r (#31409) is used, then we can drop the SAGE_SPKG_DEPCHECK for xz in python3/spkg-configure.m4
because then there is no package depending on liblzma that will run in the same process as python3.
(cmake depends on liblzma but it does not run in the same process.)

[mkoeppe]

Dependencies: #31409

[mkoeppe]

Branch: u/mkoeppe/build_pkgs__split_out_spkg_liblzma_from_xz

[mkoeppe]

Author: Matthias Koeppe

[mkoeppe]

Commit: 546df52

[mkoeppe]

New commits:

c504bb9	build/pkgs: Split out spkg liblzma from xz
f83aac5	build/pkgs/{cmake,r}/dependencies: Depend on liblzma, not xz
9e1e643	build/pkgs/python3/dependencies: Add liblzma
fe36a64	build/pkgs/r/spkg-configure.m4: Change DEPCHECK from xz to liblzma
212b03e	build/pkgs/python3/spkg-configure.m4: Change DEPCHECK from xz to liblzma
546df52	build/pkgs/xz/spkg-configure.m4: Set sage_require_xz=no if sage-bootstrap-python is new enough

[mkoeppe]

comment:10

The changes discussed in comment:2 to comment:4 still need to be implemented

[sagetrac-git]

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

2b4b142	build/pkgs: Split out spkg liblzma from xz
5137ae9	build/pkgs/{cmake,r}/dependencies: Depend on liblzma, not xz
462fd33	build/pkgs/python3/dependencies: Add liblzma
ba6fdcd	build/pkgs/r/spkg-configure.m4: Change DEPCHECK from xz to liblzma
bba3a93	build/pkgs/python3/spkg-configure.m4: Change DEPCHECK from xz to liblzma
bf500be	build/pkgs/xz/spkg-configure.m4: Set sage_require_xz=no if sage-bootstrap-python is new enough

[sagetrac-git]

Changed commit from 546df52 to bf500be

[sagetrac-git]

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

bc01006

build/pkgs/xz: Update to 5.2.5, add upstream_url

[sagetrac-git]

Changed commit from bf500be to bc01006

[sagetrac-git]

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

009256e

build/pkgs/liblzma/dependencies: Update

[sagetrac-git]

Changed commit from bc01006 to 009256e

[jhpalmieri]

comment:15

Should there be a symlink build/pkgs/liblzma/package-version.txt -> ../xz/package-version.txt? Or is it okay to omit the version for lzma?

[sagetrac-git]

Changed commit from 009256e to de28c72

[sagetrac-git]

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

de28c72

build/pkgs/liblzma: Add symlink for package-version.txt

[mkoeppe]

comment:17

It's ok to omit it (it is using "none" in this case), but I agree that it's probably nicer with a version

[jhpalmieri]

comment:18

Thank you, looks good.

[jhpalmieri]

Reviewer: John Palmieri

[mkoeppe]

comment:19

Thanks!

[vbraun]

comment:20

Merge conflict

[sagetrac-git]

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

8ba1d79	trac 30350: remove the ATLAS package
69db839	trac 30350: remove "--with-blas=" option to ./configure
272424a	trac 30350: remove remaining bits in build/pkgs/atlas/distros
1fb1fa9	Merge #30350

[sagetrac-git]

Changed commit from de28c72 to 1fb1fa9

[mkoeppe]

Changed dependencies from #31409 to #31409, #30350

[vbraun]

Changed branch from u/mkoeppe/build_pkgs__split_out_spkg_liblzma_from_xz to 1fb1fa9