Re: [ANNOUNCE] Git v2.49.0-rc1

From: Todd Zullinger <tmz@pobox.com>
Date: Mar 4, 2025, 18:37 UTC
Message-ID: <Z8dIZmscTdi8dZAY@teonanacatl.net>
In-Reply-To: <xmqqjz94r8p0.fsf@gitster.g>

Junio C Hamano wrote:

> A release candidate Git v2.49.0-rc1 is now available for testing at
> the usual places.  It is comprised of 367 non-merge commits since
> v2.48.0, contributed by 68 people, 17 of which are new faces [*].

[...]

>  * Lazy-loading missing files in a blobless clone on demand is costly
>    as it tends to be one-blob-at-a-time.  "git backfill" is introduced
>    to help bulk-download necessary files beforehand.

In <Z8HW6petWuMRWSXf@teonanacatl.net>, I noted that one of the git backfill command's test fails on s390x:

expecting success of 5620.4 'do partial clone 2, backfill min batch size':
        git clone --no-checkout --filter=blob:none      \
                --single-branch --branch=main           \
                "file://$(pwd)/srv.bare" backfill2 &&
        GIT_TRACE2_EVENT="$(pwd)/batch-trace" git \
                -C backfill2 backfill --min-batch-size=20 &&
        # Batches were used
        test_trace2_data promisor fetch_count 20 <batch-trace >matches &&
        test_line_count = 2 matches &&
        test_trace2_data promisor fetch_count 8 <batch-trace &&
        # No more missing objects!
        git -C backfill2 rev-list --quiet --objects --missing=print HEAD >revs2 &&
        test_line_count = 0 revs2
+++ pwd
++ git clone --no-checkout --filter=blob:none --single-branch --branch=main 'file:///tmp/git-t.6AFX/trash directory.t5620-backfill/srv.bare' backfill2
ok 3 - do partial clone 1, backfill gets all objects
Cloning into 'backfill2'...
+++ pwd
++ GIT_TRACE2_EVENT='/tmp/git-t.6AFX/trash directory.t5620-backfill/batch-trace'
++ git -C backfill2 backfill --min-batch-size=20
++ test_trace2_data promisor fetch_count 20
++ grep -e '"category":"promisor","key":"fetch_count","value":"20"'
error: last command exited with $?=1
not ok 4 - do partial clone 2, backfill min batch size

Being a new command, perhaps it's not terrible if it doesn't work on a less popular architecture, but I don't want to let it slip past without anyone noticing.

As I mentioned the other day, I don't have shell access to the s390x host, but can run the test suite with additional debugging if needed. I have the test-results directory from the build as well.

If anyone wants to poke at it, the full build log (which contains the test-results directory as a base64-encoded, zst-compressed tar archive) can be found here for a few weeks:

    https://kojipkgs.fedoraproject.org//work/tasks/7999/129807999/build.log

-- 
Todd