Git Rev News: Edition 64 (June 25th, 2020)

Welcome to the 64th edition of Git Rev News, a digest of all things Git. For our goals, the archives, the way we work, and how to contribute or to subscribe, see the Git Rev News page on git.github.io.

This edition covers what happened during the month of May 2020.

Discussions

Reviews

  • fetch: allow adding a filter after initial clone

    Xin Li sent a patch to allow using a partial clone filter when fetching even if the original clone didn’t use one.

    As the patch was lacking a proper commit message, Junio Hamano, the Git maintainer, asked Xin to add one. So Xin sent an explanation of what the commit does taking as an example a shallow clone, with git clone --depth=1 ..., that is transformed into a partial clone with git fetch --unshallow --filter=blob:none .... Xin wrote that being able to do so is easier for the user than having to manually set the core.repositoryFormatVersion and extensions.partialClone config options.

    Junio replied asking how we could be sure that the server actually supports the protocol extensions required for partial clone, as setting up the config options manually or automatically shouldn’t be done without verifying that.

    Brian Carlson also replied to Xin’s initial patch wondering if it was safe to set the core.repositoryFormatVersion to 1 as it would make Git fail if any already configured extensions.* config option, like extensions.tomatoSalad, was not recognized.

    Xin then sent a version 2 of his patch. This time there was a commit message in the patch and a cover letter along with the patch that explained the changes. The patch itself added tests and checked that the existing configuration had no unsupported extensions.* configuration before upgrading the core.repositoryFormatVersion config option to 1.

    Jonathan Nieder reviewed the patch suggesting a number of small changes and moving some content from the cover letter to the commit message. Xin then sent a version 3 of his patch applying Jonathan’s suggestions and without a cover letter.

    Junio in the meantime commented on the version 2 that it was not enough to check that the existing configuration had no unsupported extensions.* configuration variable, but that instead the code should check that there is no extensions.* configuration variable at all, as any such configuration variable could have an unsupported value. Jonathan then agreed with Junio.

    Xin replied to them that the current code doesn’t actually check repositoryFormatVersion, so setting extensions.partialclone is enough to make git fetch --filter=... work. He then asked what we want to do about this situation.

    Jonathan replied that this should be fixed even though it was a separate concern from the patch.

    Xin then sent a version 4 of his patch taking into account Junio’s comments. Junio reviewed it and suggested other improvements. He further elaborated on core.repositoryFormatVersion upgrade by suggesting the following criteria:

    (1) if upgrading from v0, there must be no extensions.* and

    (2) if upgrading from other versions, there must be no extensions.* we do not recognize.

    Jonathan also reviewed version 4 of Xin’s patch focusing on the changes made in tests.

    Xin sent a version 5 of the patch addressing the comments. He also replied to Jonathan’s review. Junio also replied to Jonathan’s review agreeing with him that some changes in the patch actually belong to another patch.

    Xin then sent a version 6 with 4 patches instead of 1. Junio reviewed the patches and found that only the last one could be controversial as it could perhaps break an existing set-up.

    The last patch indeed contained the changes about checking repositoryFormatVersion that Jonathan and Junio wanted to see separately. With that patch the extensions.* config options don’t take effect unless core.repositoryFormatVersion has been upgraded to 1 or higher.

    So for example if extensions.worktreeConfig is set to true but core.repositoryFormatVersion isn’t set, then with the patch Git will behave as if extensions.worktreeConfig wasn’t set.

    Junio thinks that this change is the right thing to do in the longer term though. He asked for comments about that but no-one answered. So the patch series has been merged to ‘pu’ and then ‘next’ and according to the last “What’s cooking in git.git” email from Junio, the plan is to merge to ‘master’.

Releases

Other News

Various

Light reading

Git tools and sites

  • Piranha by Uber is an Open Source tool for automated clean up of stale code caused by feature flags that are no longer required. It currently supports Java, Swift, and Objective-C.
  • git-fuzzy is a CLI interface to Git that relies heavily on the fzf general-purpose command-line fuzzy finder (version 0.21.0 or higher).

Credits

This edition of Git Rev News was curated by Christian Couder <christian.couder@gmail.com>, Jakub Narębski <jnareb@gmail.com>, Markus Jansen <mja@jansen-preisler.de> and Kaartic Sivaraam <kaartic.sivaraam@gmail.com>.