This week has almost only been cleaning up the describe atom code and making it ready to be a proper patch series that can be sent out to the mailing list. So no new work has been put into either .mailmap or Hariom’s idea, which is kind of sad because I have been postponing it for a while now, especially Hariom’s idea but things should settle in this week.

Interesting stuff

A couple of interesting things popped up while I was cleaning up the “describe” atom code. The first one being that a totally unrelated test was failing on linux-sha256 (CI). The reason this test was failing, particularly on linux-sha256 was because of the change in the raw:size of certain objects due to new commits and tags that were added as part of the describe tests. This would re-arrange the refs when sorting and hence the test would fail.

So, I wrote another commit to fix this. This commit takes the approach of making all the describe related tests run on describe-repo (a repo exclusively for describe tests to run), instead of the usual main repo of t6300-for-each-ref.sh.

I originally thought to squash this commit onto the duplication commit, but Christian suggested that it deserves an explanation and I agree with him.

I sent these patches to the mailing list (this fix commit on top of the duplication commit). They can be read on this thread.

Another thing I came across was the output of the following command when run in the trash directory of t6300 right after the “Verify sorts with raw:size” test. 

$ git for-each-ref --format="%(refname) %(raw:size)" --sort=raw:size refs/heads/main refs/myblobs/ refs/mytrees/first


which gives

refs/myblobs/blob8 0
refs/myblobs/first 17
refs/myblobs/blob7 2
refs/myblobs/blob4 3
refs/heads/main 446
refs/myblobs/blob1 5
refs/myblobs/blob2 5
refs/myblobs/blob3 6
refs/myblobs/blob5 6
refs/myblobs/blob6 6
refs/mytrees/first 64


which is strange because the sorting seems to happens while only taking into account the first digit of raw:size. I don’t know if this is expected and if it was not, then I would need to dig a little more deeper for this.

According to the “What’s cooking” email sent on June 30, the change in %(describe:abbrev=...) test in t4205-log-pretty-formats.sh is on its way to master.

Christian sent out a reply to “What’s cooking” saying that the signature atom patches where ready to be merged (they had been on “Needs review” for a while now) since they had seen so many changes and the current version was looking good.

I have also made sure to change the %(describe:abbrev=...) test in the duplication of describe too.

‘til next time,
Kousik