Does anyone use git-notes?

I would like to write a post on how to use them, but I'm a bit confused as to how people actually use them on a day to day basis. I appears to me in trying to work out a flow for them that the lack of an ability to merge them makes them very difficult to use for anything practical. Can someone share with me how they use them and what the cycle is?

Thanks, Scott

  ----- Original Message -----
From: Scott Chacon
Date: 8/25/2010 12:15 PM

I do. I use them to automatically store a matching Perforce revision number for a given Git commit. Only the server assigns them due to the merging issues. I also change the ref, so they read as p4notes.

It irks me, though, that they show in the gitk --all view.  ARGH!

Josh

Scott Chacon wrote:

I run a cronjob that downloads mail from gmane and fetches git.git, then attempts to correlate mail with commits. The results are at

  git://repo.or.cz/git/trast.git notes/terse notes/full

I have previously posted about this at e.g.

  http://mid.gmane.org/200902091508.11460.trast@student.ethz.ch
  http://mid.gmane.org/201003042321.36240.trast@student.ethz.ch

though I should point out that it no longer tries to apply the patches as outlined in the first post; this was simply way too expensive.

I also use it when heavily editing a patch series to tack little reminders on my fixup patches. I have

  $ git config alias.fixup
  !sh -c 'r=$1; git commit -m"fixup! $(git log -1 --pretty=%s $r)"' -

so I usually do

  # hack
  git add -p
  git fixup HEAD~3
  git notes add -m 'fix foo'

which combined with notes.rewriteRef=refs/notes/* (and notes.rewriteMode left at the default of 'concatenate') means that I end up having a little list of reminders what changed since the last revisions. I then simply clear them out before starting a new round.

Scott Chacon <schacon@gmail.com> writes:

I have a custom post-applypatch hook (see my "todo" branch) enabled and then use the amlook script (again, see my "todo" branch) script to inspect the corresponding notes tree.

After applying a received patch e-mail, the result is annotated with enough information to identify which message it came from by the hook. When I am scouring the mailing list backlog to see if I have missed something interesting (sometimes I find interest in a topic that I thought totally uninteresting earlier), the amlook script is used to see if the patch has been already applied.

Looking through millions of notes this way won't scale (iow, amlook is a bad example of using this feature, as the notes tree is used as a database index in the wrong direction), by the way, so you may probably not want to use it as a good example.

Scott Chacon <schacon@gmail.com> writes:

I have a custom post-applypatch hook (see my "todo" branch) enabled and then use the amlook script (again, see my "todo" branch) to inspect the notes the hook left.

After applying a received patch e-mail, the result is annotated with enough information to identify which message it came from by the hook (I probably should add more information to the notes, but never got around to it). When I am scouring the mailing list backlog to see if I have missed something interesting (sometimes I find interest in a topic that I thought totally uninteresting earlier), the amlook script is used to see if the patch has already been applied, and if so to which branch.

Looking through millions of notes this way won't scale (iow, amlook is a bad example of using this feature, as the notes tree is used as a database index in the wrong direction), by the way, so you may probably not want to use it as a good example.

On Wed, Aug 25, 2010 at 20:15, Scott Chacon <schacon@gmail.com> wrote:

I use Notes in a very simplistic way. For example, with an emulator I'm writing I have two or three ways of using git-notes: 1: I commit fixes and changes with a commit message describing the fix. Later I may find out that the fix made a couple more programs of the set I have run in the emulator, so I add a note to the commit saying 'This change makes X and Y work'. 2: I add test cases to the instruction validator I wrote for the emulator, if the test currently fails I say so in the commit message. At some later point I may commit a fix that lets the test pass, then I go back to the original commit message (which said the test currently failed) and add a note saying 'Fixed by [commit] nnnnnnn'. 3: Occasionally I just use notes as a reminder. If a commit adds a new feature I may later (after using the feature) add a 'todo'-note about what improvements I should add.

One use case I have thought of for the company case is to add notes to a release commit with information that wasn't or couldn't be added at the time of the commit, e.g. a list of problem reports closed by the release. But I haven't actually figured out how I can get Notes transferred over a pull yet.. although I haven't tried since 1.7.0.

-Tor

Scott Chacon venit, vidit, dixit 25.08.2010 20:15:

I use it to store the patch comments (the part that goes after the "---", before the diff) along with my commits so that I have them in my git.git tree, not only in my mbox. This makes it also easier to redo a patch and keep the comment when regenerating the patch e-mail with format-patch. (I don't have a good solution for the cover letter yet.)

Michael

On Thu, 2010-08-26 at 10:09 +0200, Michael J Gruber wrote:

Not to stray off-topic, but I use tags for the cover letter eg: refs/tags/patches/short-description/v[N]

I hadn't thought of using git-notes for the per-patch comments, here I was with the opposite problem (A straighforward way to keep track of cover-letters, but no good way to keep track of the per-patch comments)

On Thu, Aug 26, 2010 at 4:15 AM, Scott Chacon <schacon@gmail.com> wrote:

I use them to store test execution output.

By post-processing, then diffing such notes, I can work out what was regressed or fixed between two commits.

jon.