Extend PURGE_TARGETS for unreproducible ruby files

Description:

Currently every PKGBUILD implements its own cleanup mess for unreproducible files. This could be standardised in PURGE_TARGETS so it works out of the box.

added priority3-normal severity4-low statusunconfirmed labels

assigned to @freswa, @foxboron, @anthraxx, @toolybird, and @gromit

For the record the current cleanup code ruby packages use to remove unreproducible files is:

  # remove unrepreducible files
  local _gemdir="$(gem env gemdir)"

  rm --force --recursive --verbose \
    "tmp_install/${_gemdir}/cache/" \
    "tmp_install/${_gemdir}/gems/${_name}-${pkgver}/vendor/" \
    "tmp_install/${_gemdir}/doc/${_name}-${pkgver}/ri/ext/"

  find "tmp_install/${_gemdir}/gems/" \
    -type f \
    \( \
      -iname "*.o" -o \
      -iname "*.c" -o \
      -iname "*.so" -o \
      -iname "*.time" -o \
      -iname "gem.build_complete" -o \
      -iname "Makefile" \
    \) \
    -delete

  find "tmp_install/${_gemdir}/extensions/" \
    -type f \
    \( \
      -iname "mkmf.log" -o \
      -iname "gem_make.out" \
    \) \
    -delete

Taken from https://wiki.archlinux.org/title/Ruby_package_guidelines

Not real familiar with the subject matter, but trying to translate the above into something suitable for PURGE_TARGETS might be a bit unwieldy..

Yeah I'm also not sure that would make sense.

IIRC @anthraxx and I already talked about this in the past and one idea was to add some kind of function that could be called in ruby PKGBUILD files to reduce the amount of lines needed from the above 27 to 1.

Can files be removed post packaging from ${pkgdir} instead of ${tmp_install}?

Yes they can. We just need to update the paths in the commands above.

added scopeenhancement statusconfirmed labels and removed statusunconfirmed label

unassigned @freswa, @toolybird, and @gromit

This is not a PURGE_TARGETS job.

I'd distribute a custom libmakepkg file /usr/share/makepkg/tidy/ruby.sh or with pacman/makepkg to automatically handle this if the clean-up can be done post package(). If the clean-up needs to happen during build(), then provide a function in /usr/share/makepkg/reproducible/ruby.sh.

The cleanup can easily happen also at the end of the package() function call.

Would ruby package maintainer then need to call a special function at the end of the package() function or will this be done automatically? If it gets done automatically how will the system detect that the package is a ruby package?

This is super easy then! Just add this file as usr/share/makepkg/tidy/ruby.sh to the pacman package and it the clean-up should be automatic.

Note: that file needs tested...

Just an FYI to the pacman packagers - I'm happy to provide such scripts to integrate into libmakepkg, but I don't think they should be hosted in the pacman repo. If needed, we can create another project to hold these scripts.

This is super easy then! Just add this file as usr/share/makepkg/tidy/ruby.sh to the pacman package and it the clean-up should be automatic.

Note: that file needs tested...

I also didn't test it yet but in general it looks good. So this function is always executed but only runs when it detects the gem command?

Just an FYI to the pacman packagers - I'm happy to provide such scripts to integrate into libmakepkg, but I don't think they should be hosted in the pacman repo. If needed, we can create another project to hold these scripts.

Right now on my system I already see 7 files in /usr/share/makepkg/tidy. Where are they coming from and could we add the ruby.sh to that place?

They come from pacman, and are much more generic in their utility. This will not be added there.

I also didn't test it yet but in general it looks good. So this function is always executed but only runs when it detects the gem command?

Correct.

I guess for now we could also add this in the pkgsources but that would be something to consider for the package maintainers.

How can I easily test this? Just add the file to the folder locally, remove the cleanup from the PKGBUILD and see if the output is the same?

Yes - it is a drop-in file. Just place in the relevant directory and it should be good to go.

@allan I tried to test your file, but it seems the tidy_ruby() function is never called. I added some debug output to the file on the root level and I see it, but when I add some output directly to the tidy_ruby() function (before the if) then I don't see this.

It is hard to test further and I'm at a loss how to test this further.

I also printed the tidy_remove variable after the assignment of the tidy_ruby variable and i see that it was added.

Any idea why the function is never called?

Work for me...

==> Tidying install...
  -> Removing empty directories...
  -> Removing libtool files...
  -> Purging unwanted files...
  -> Purging unreproducible ruby files...

Ok it seems to work now. My error was to thinking it work with pkgctl build but the file is never created there. When building directly with makepkg I see it being called and it seems to work.

I have a question: Where does ${_name} and ${pkgver} from? I can imagine the second one is omnipresent but the first seems like something that might not be there. Maybe we could just use * instead of ${_name}-${pkgver}.

Is there an easy way for me to test this inside of pkgctl build?

@allan any thoughts on the questions in my previous comment in this thread?

Check out the libmakepkg_dropins repo where that is fixed.

ruby.sh

A slightly updated version.

mentioned in merge request pacman/pacman!233 (closed)

The libmakepkg drop-in file can be sourced from https://gitlab.archlinux.org/pacman/libmakepkg-dropins

That looks nice. What is the planned/intended way to distribute these now to the package maintainers?

@allan what was your plan here?

mentioned in issue gitlab#14 (closed)

mentioned in issue #44