On 2024.12.05 21:09, Bence Ferdinandy wrote: > > On Thu Dec 05, 2024 at 20:50, Josh Steadmon <steadmon@xxxxxxxxxx> wrote: > > On 2024.12.05 10:58, Josh Steadmon wrote: > >> On 2024.11.22 13:28, Bence Ferdinandy wrote: > >> > When cloning a repository remote/HEAD is created, but when the user > >> > creates a repository with git init, and later adds a remote, remote/HEAD > >> > is only created if the user explicitly runs a variant of "remote > >> > set-head". Attempt to set remote/HEAD during fetch, if the user does not > >> > have it already set. Silently ignore any errors. > >> > > >> > Signed-off-by: Bence Ferdinandy <bence@xxxxxxxxxxxxxx> > >> > >> At $DAYJOB, we noticed that this breaks `git fetch --tags`, although I > >> haven't had a chance to figure out what causes the error just yet. > >> > >> I was able to bisect down to this commit using Jonathan Tan's > >> reproduction script: > >> > >> rm -rf test_tag_1 test_tag_2 > >> GIT=~/git/bin-wrappers/git > >> mkdir test_tag_1 && cd test_tag_1 > >> REMOTE=$(pwd) > >> $GIT init . > >> touch foo.txt > >> $GIT add foo.txt > >> $GIT commit foo.txt -m "commit one" > >> $GIT tag foo > >> cd .. > >> mkdir test_tag_2 && cd test_tag_2 > >> $GIT init . > >> echo fetch --tags > >> $GIT fetch --tags "file://$REMOTE" > >> echo regular fetch > >> $GIT fetch "file://$REMOTE" 'refs/tags/*:refs/tags/*' > >> $GIT --version > >> > >> > >> Prior to this change, the first `$GIT fetch --tags "file://$REMOTE"` > >> fetches the `foo` tag; with this change, it does not. > > > > > > FWIW, moving this: > > > > @@ -1643,6 +1703,8 @@ static int do_fetch(struct transport *transport, > > "refs/tags/"); > > } > > > > + strvec_push(&transport_ls_refs_options.ref_prefixes, "HEAD"); > > + > > > > to just above the prior `if` block fixes our issue and doesn't break any > > tests. However, I'm not sure yet why the order of ref_prefixes should > > matter here. > > Thanks for looking into this! I think the issue is with > > $GIT fetch --tags "file://$REMOTE" > > instead of adding a proper remote. Tbh, I've never seen the above syntax before, so first I just ran your script, which reproduced the issue for me, but then I modified it to use a proper remote which works as expected: > > rm -rf test_tag_1 test_tag_2 > GIT=~/git/bin-wrappers/git > mkdir test_tag_1 && cd test_tag_1 > REMOTE=$(pwd) > $GIT init . > touch foo.txt > $GIT add foo.txt > $GIT commit foo.txt -m "commit one" > $GIT tag foo > cd .. > mkdir test_tag_2 && cd test_tag_2 > $GIT init . > $GIT remote add origin $REMOTE > echo fetch --tags > $GIT fetch origin --tags > echo regular fetch > $GIT fetch origin 'refs/tags/*:refs/tags/*' > $GIT --version > > So I'm assuming this is why also the tests never caught this, since probably > all of them are using `git remote add`. > Yeah, I think the issue is that we check the `--tags` flag first, but only add the ref_prefixes entry if it's not empty already. Then after that we unconditionally add HEAD. So that's why moving your strvec_push earlier fixes it. I'll send a fix + test patch in just a minute.
