Hello,
I created a patch for implementing this feature inside cygport.
I did several tests and it worked fine in my opinion.
I hope that you will find this useful.

Sincerely,

Carlo Bramini.

Il giorno dom 18 mag 2025 alle ore 20:46 Jon Turney
<jon.tur...@dronecode.org.uk> ha scritto:
>
> On 29/04/2025 16:42, Carlo B. via Cygwin-apps wrote:
> [...]
> >
> > I can see these possible improvements into the code:
> >
> > 1)
> > The __tar() function inside lib/pkg_pkg.cygpart could be simplified by
> > writing just this:
> >
> > TAR_COMPRESSION_OPT=$(__tar_compress_opt ${TAR_COMPRESSION_EXT})
> >
> > and removing duplicated code now moved inside __tar_compress_opt()
> >
> > 2)
> > For adding the selection of the type of compression, I suggest to add
> > a new identifier called GIT_COMPRESSION_EXT. This will be used in a
> > similar manner of existing TAR_COMPRESSION_EXT.
> >
> > Inside cygclass/git.cygclass, we can add this:
> >
> > if ! defined GIT_COMPRESSION_EXT
> > then
> >      GIT_COMPRESSION_EXT="bz2"
> > fi
> >
> > which forces the default extension to bz2 for compatibility (I don't
> > know if it is useful or not).
> > Next, git_tarball symbol can be modified like this:
> >
> > git_tarball="${GIT_MODULE}-${PV}.tar.${GIT_COMPRESSION_EXT}"
> >
> > and finally, the call to tar will be:
> >
> > tar $(__tar_compress_opt ${GIT_COMPRESSION_EXT})cf
> > ${top}/${git_tarball} --exclude-vcs ${GIT_MODULE}/${GIT_SUBDIR}
> >
> > 3)
> > The same logic can be replicated if somebody will want to add the same
> > feature for selecting the compressor for other revision control
> > system: subversion, bazaar, cvs, mercurial.
>
> Thanks very much for looking into this.
>
> I think I would take a patch which just changes the compression used for
> archives made from git to lzma.
>
> (I can't see any particular use the ability to force the previous
> behavior of bz2 compression, but the more elaborate changes you propose
> seem like the right approach for doing that)
>
> > What do you think? Actually, this is first solution that entered in my
> > head and perhaps it could not be the best one.
>
> Give yourself a pat on the back for demonstrating an awareness of
> cognitive biases!
>

Attachment: cygport.patch
Description: Binary data

Reply via email to