Reuse repositories when updating

[ruuda]

If I understand correctly, currently PKGBUILDer can only do a clone, but not a pull. For an update, an existing repository is not reused by doing a pull, but instead one has to use the -c flag to remove the old repository and clone again. It would be great if instead of

==> Cloning the git repository...
==> ERROR: Directory <package> already exists, please run with `-c` to remove it.
  -> skipping package <package>

it would simply do a pull to get the latest version.

[Kwpolska]

It certainly makes sense, I’ll get that done soon.

[Kwpolska]

Fixed on the develop branch, will make a release soon. New behavior:

• if no directory exists, clone as usual
• if directory exists and is a git repository, pull it and update
• if directory exists and is not a git repository, skip this package

[ruuda]

Awesome!

[AladW]

This is somewhat inconvenient since any diffs are lost after git pull is done. I guess you could git diff / git merge manually before running PKGBUILDer...

[Kwpolska]

Are you sure about that? git pull should not remove unstaged/uncommitted changes from your working copy.

[AladW]

Well, in particular I meant upstream changes, for example when someone else adopted an AUR package and made a bunch of changes to the PKGBUILD compared to the last checkout.

[Kwpolska]

I’m not sure I understand the problem.

[AladW]

When I have new targets I'd run pkgbuilder -G so that I can take a look at PKGBUILDs before building them.

When I have existing targets I'd run pkgbuilder -G again, but since a git merge was run (via git pull) I'd have no way to see what changes are "new" (for example, a new maintainer changed the source to something malicious) and have to review the whole bunch anew.