Fix autoreconf 2.71 missing some files

[ChrisJefferson]

The release notes for autoconf 2.70 ( https://lwn.net/Articles/839395/ ) mention it may be necessary to run autoreconf --install. This should probably be added to each package, but for now we can add it to build_packages.sh -- this may be too much of a sledge-hammer, let's first see if this fixes problems at all.

See #4374.

[wilfwilson]

Sorry that CI problems are intruding on your holiday 🙁 But thanks for helping @ChrisJefferson.

Sadly the job has failed again, at least so far. If you or anyone else wants to look into this further, then adding this step:

    - uses: actions/checkout@v2
    - name: Setup tmate session
      if: ${{ failure() }}
      uses: mxschmitt/action-tmate@v3

to the workflow (from https://github.com/marketplace/actions/debugging-with-tmate) might be useful, for SSHing into a session where it doesn't compile.

[ChrisJefferson]

I (think) that's working now. The main problem is summed up by the paragraph below (quoted from the release notes). In order to get autoreconf -i to run on mac, I also had to install automake (reading the homebrew pages, while some bits of autoconf, including autoreconf, require automake, they don't automatically install automake if you install autoconf.

 As a consequence of improved support for cross compilation (see below),
  more macros now use the auxiliary scripts ‘config.sub’ and ‘config.guess’.
  If you use any of the affected macros, these scripts must be available
  when your configure script is run, even if you have no intention of
  ever cross-compiling your program.

  autoreconf will issue an error if any auxiliary scripts are needed but
  cannot be found.  (It is not currently possible to make autoconf
  itself issue this error.)

  ‘autoreconf --install’ will add ‘config.sub’, ‘config.guess’, and
  ‘install-sh’ to your source tree if they are needed.  If you are
  using Automake, scripts added to your tree by ‘autoreconf --install’
  will automatically be included in the tarball produced by ‘make dist’;
  otherwise, you will need to arrange for them to be distributed
  yourself.

  See the “Input” section of the manual for more detail, including
  where to get the auxiliary scripts that may be needed by autoconf macros.

[ChrisJefferson]

(there is still a whole bunch of warnings about deprecated things, that might well want fixing some time, but at least right now don't seem to be blocking us running CI, or building packages.)

[wilfwilson]

Brilliant

[fingolfin]

I'll submit an alternate PR.

But we still need to also deal with this for packages. I actually started looking into this yesterday, when I run into this independently for some GAP packages. I should have notified you folks to avoid duplication of effort, sorry.

[fingolfin]

No no no! We do not want to pull in automake

[fingolfin]

We can try to simply update config.sub and config.guess, that's something we have to do occasionally anyway (see issue #1478). There are other changes to be made in configure.ac. I'll look into that. But all of this must keep working with autoconf 2.69, which is what almost everyone has. In a quick test of mine, it does so.

[fingolfin]

I think the issue was actually specific to the io package, which actually had a buggy configure.ac that happened to work in autoconf 2.69, and now 2.70 detected it. To be precise: I set it up to not check for a "canonical host type", but then used autoconf macros which actually require that; in autoconf 2.69 this sailed through (and possibly resulted in buggy behavior, though this unlikely on any modern day OS). In autoconf 2.70 it detects this requirement and computes the canonical host type, which require config.guess and config.sub ... hence the issue

[fingolfin]

I really don't think this is required (nor appropriate), but I am open to have my mind changed about this :-)