spkg-configure.m4 for python: tzdata, pytest, vcversioner

[dimpase]

also, remove leftovers of setuptools_scm_git_archive, left in the tree by an oversight.

This is continuation of #36332

[mkoeppe]

Note that #36129 is going to remove vcversioner, so perhaps this part can be omitted

[orlitzky]

tzdata can be skipped in most cases because it provides timezone data that's already installed on every system; the system python will be configured to use the system timezone data. In Gentoo the tzdata package literally does nothing.

To check, we would need to run some zoneinfo command to see if it fails. I once started to write an autoconf macro for that but it's probably overkill here. Something like python -s -c 'import zoneinfo; print(zoneinfo.ZoneInfo("America/New_York"));'could work? (The -s is to simulate our runtime poisoning of PYTHONUSERBASE.)

[orlitzky]

"python" should probably be $PYTHON_FOR_VENV in that command, but I don't remember what the difference is any more

[orlitzky]

Maybe we don't have to think about tzdata at all. It's used only by pytz_deprecation_shim, which is used only by tzlocal, which is used only by rpy2. Both pytz_deprecation_shim and tzlocal are already disabled with --disable-r. I think we should add tzdata to the list of R packages in configure.ac and forget about it until there's enough consensus on sage-devel to delete R.

[mkoeppe]

I think we should add tzdata to the list of R packages in configure.ac

I agree

[orlitzky]

Rebase on top of #36129 with no vcversioner and with tzdata disabled by --disable-r?

I think it's about time we switch the default to --disable-r too. The last meaningful change to the R interface was,

commit 2c069c9c1224b2f9f26b9ae282e6ac22991dfa91
Author: Karl-Dieter Crisman <kcrisman@gmail.com>
Date:   Mon Apr 19 21:26:35 2010 -0400

    Trac 7665 - updates R interface to fully enable R graphics in nb and command line

[mkoeppe]

Rebase on top of #36129 with no vcversioner and with tzdata disabled by --disable-r?

Yes please

I think it's about time we switch the default to --disable-r too. The last meaningful change to the R interface was,

Hm? No, we have working rpy2, which is installed whenever there's a compatible system R.

[orlitzky]

I think it's about time we switch the default to --disable-r too. The last meaningful change to the R interface was,

Hm? No, we have working rpy2, which is installed whenever there's a compatible system R.

And when there's not, we build one for no reason.

What I should have said though is "switch R to optional" rather than disabled. Then if people want to use R, they install R.

[mkoeppe]

I think it's about time we switch the default to --disable-r too. The last meaningful change to the R interface was,

Hm? No, we have working rpy2, which is installed whenever there's a compatible system R.

And when there's not, we build one for no reason.

No, we don't. R is a dummy package.

[orlitzky]

No, we don't. R is a dummy package.

Of course it is. I'm going to take a nap.

[orlitzky]

In addition to adding tzdata to --disable-r, would it help to copy this REQUIRED-CHECK from rpy2?

SAGE_SPKG_CONFIGURE([rpy2], [
    SAGE_PYTHON_PACKAGE_CHECK([rpy2])
], [dnl REQUIRED-CHECK
    AC_REQUIRE([SAGE_SPKG_CONFIGURE_R])
    dnl rpy2 is only needed when there is a usable system R
    AS_VAR_IF([sage_spkg_install_r], [yes], [dnl
        AS_VAR_IF([sage_use_system_r], [installed], [dnl
            dnl Legacy SPKG installation of r
            AS_VAR_SET([SPKG_REQUIRE], [yes])
        ], [dnl No system package, no legacy SPKG installation
            AS_VAR_SET([SPKG_REQUIRE], [no])
        ])
    ])
])

It looks like tzlocal and pytz_deprecation_shim could benefit too, for the people who don't always pass --disable-r.

[mkoeppe]

Should rebase and remove vcversioner now

[github-actions]

Documentation preview for this PR (built with commit 9959f98; changes) is ready! 🎉