Re: [PATCH] setlocalversion: Add workaround for "git describe" performance issue

From: Rasmus Villemoes <ravi@prevas.dk>
Date: Nov 1, 2024, 10:23 UTC
Message-ID: <8734kbjlrq.fsf@prevas.dk>
In-Reply-To: <20241031122456.GB593548@coredump.intra.peff.net>

On Thu, Oct 31 2024, Jeff King <peff@peff.net> wrote:

Show 19 quoted lines

> On Thu, Oct 31, 2024 at 07:42:10AM -0400, Jeff King wrote:
>
>> That works, but I have a feeling that figured out what the heck is going
>> on with gave_up_on might produce a more elegant solution.
>
> OK, I think I might have made some sense of this.
>
> In finish_depth_computation(), we traverse down "list" forever, passing
> flags up to our parents, until we find a commit that is marked with the
> same "within" flag as our candidate. And then if everything left has
> that same "within" flag set, we can bail.
>
> So I _think_ the point is to basically count up what we'd get from this
> traversal:
>
>   $tag..$commit
>
> where "$tag" is the candidate tag we found, and "$commit" is what we're
> trying to describe (so imagine "git describe --match=$tag $commit").

Yeah, so this is really just what the setlocalversion script wants to know. For a few diffent possible values of $tag (in most cases just 1), we ask: Is $tag an annotated tag? Is it an ancestor of HEAD? And if so, how many commits are in $tag..HEAD.

Perhaps we could on the kernel side replace the "git describe --match" calls with a helper, something like this (needs a lot of polishing):

===
# Produce output similar to what "git describe --match=$tag 2>
# /dev/null" would.  It doesn't have to match exactly as the caller is
# only interested in whether $tag == HEAD, and if not, the number
# between the tag and the short sha1.
describe()
{
    # Is $tag an annotated tag? Could/should probably be written using
    # some plumbing instead of git describe, but with --exact-match,
    # we avoid the walk-to-the-start-of-history behaviour, so fine for
    # this demo.
    git describe --exact-match --match=$tag $tag >/dev/null 2>/dev/null || return 1

    # Can it be used to describe HEAD, i.e. is it an ancestor of HEAD?
    git merge-base --is-ancestor $tag HEAD || return 1

    # Find the number that "git describe" would append.
    count=$(git rev-list --count $tag..HEAD)
    if [ $count -eq 0 ] ; then
        echo "$tag"
    else
        echo "$tag-$count-$head"
    fi
}
===

But if we go this route, we should probably rework the logic somewhat. There's no point getting the count ourselves, stuffing that into a string, and then splitting that string with awk to %05d format the count.

I also don't know if either the --is-ancestor or the rev-list count could end up doing the same walk-all-commits we're trying to avoid.

Rasmus