git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH v3 2/3] builtin/tag.c: add --trailer arg

From
John Passaro <john.a.passaro@gmail.com>
Date
Apr 30, 2024, 21:48 UTC
Message-ID
<CAJdN7Kggv-y8xRhdFanTyxa5wP-bG0omBPuETW6qijA67gn8oQ@mail.gmail.com>
In-Reply-To
<xmqqv83yrduw.fsf@gitster.g>
On Tue, Apr 30, 2024 at 12:53 PM Junio C Hamano <gitster@pobox.com> wrote:
>

Thanks for the feedback. Hoping for a couple points of clarification then I'll put in one more version of this patch series.

Show 21 quoted lines
> "John Passaro via GitGitGadget" <gitgitgadget@gmail.com> writes:
>
> > From: John Passaro <john.a.passaro@gmail.com>
> >
> > git-tag currently supports interpreting trailers from an annotated tag
> > message, using --list --format="%(trailers)". There is no ergonomic way
> > to add trailers to an annotated tag message.
>
> Well said.  Drop "currently", though.  The usual way to compose a
> log message of this project is to
>
>  - Give an observation on how the current system work in the present
>    tense (so no need to say "Currently X is Y", just "X is Y"), and
>    discuss what you perceive as a problem in it.
>
>  - Propose a solution (optional---often, problem description
>    trivially leads to an obvious solution in reader's minds).
>
>  - Give commands to the codebase to "become like so".
>
> in this order.

Understood. In the most recent version of this patch, I updated the message. However on second thought I think I'm gonna keep this on the next submission of this patch (without "currently" of course).

Show 28 quoted lines
>
> > In a previous patch, we refactored git-commit's implementation of its
> > --trailer arg to the trailer.h API. Let's use that new function to teach
> > git-tag the same --trailer argument, emulating as much of git-commit's
> > behavior as much as possible.
>
> Nicely described.
>
> > @@ -178,6 +179,19 @@ This option is only applicable when listing tags without annotation lines.
> >       Implies `-a` if none of `-a`, `-s`, or `-u <key-id>`
> >       is given.
> >
> > +--trailer <token>[(=|:)<value>]::
> > +     Specify a (<token>, <value>) pair that should be applied as a
> > +     trailer. (e.g. `git tag --trailer "Signed-off-by:T A Ger \
> > +     <tagger@example.com>" --trailer "Helped-by:C O Mitter \
> > +     <committer@example.com>"` will add the "Signed-off-by" trailer
> > +     and the "Helped-by" trailer to the tag message.)
> > +     The `trailer.*` configuration variables
> > +     (linkgit:git-interpret-trailers[1]) can be used to define if
> > +     a duplicated trailer is omitted, where in the run of trailers
> > +     each trailer would appear, and other details.
> > +     The trailers can be seen in `git tag --list` using
> > +     `--format="%(trailers)"` placeholder.
>
> I can see this was copied-and-pasted from git-commit, but I am not
> sure if the ones used in the example are good fit for tag objects.
> What does Helped-by even mean in the context of an annotated tag?

I can see that the git project itself doesn't typically add trailers to tags. If y'all were in that habit I imagine this feature would already be implemented :-) Nonetheless Signed-off-by or Approved-by is easy to imagine, for example in an environment where multiple sign-offs are required (i.e. not just the implicit sign-off of the tagger). So we could just leave that in and be done with it.

I have a couple more hypothetical trailers that are both plausible and somewhat generic; do any of them seem expressive enough to include in the docs?

* Tested-by: T E Ster <tester@example.com>
* Testing-assigned-to: T E Ster <tester@example.com>
* Scheduled-Deployment-Date: 2024-05-15 (or 1714500385 -05:00)
* Deployment-assigned-to: Oscar P Staff <ops@example.com>
* (for RC/alpha tags) Full-release-scheduled-for: 2024-06-05

There's also project-specific trailers. For example, on my team, we use "Deploy-Strategy: ..." to tell CICD what deployment routines to run. This is pretty specific to us but worth calling out. Maybe could translate to a documentation example with something like "<Project-specific-trailer>: foo"

Show 45 quoted lines
> > @@ -338,10 +344,22 @@ static void create_tag(const struct object_id *object, const char *object_ref,
> >               }
> >               close(fd);
> >
> > -             if (launch_editor(path, buf, NULL)) {
> > -                     fprintf(stderr,
> > -                     _("Please supply the message using either -m or -F option.\n"));
> > -                     exit(1);
> > +             if (trailer_args->nr && amend_file_with_trailers(path, trailer_args))
> > +                     die(_("unable to pass trailers to --trailers"));
> > +
> > +             if (should_edit) {
> > +                     if (launch_editor(path, buf, NULL)) {
> > +                             fprintf(stderr,
> > +                                     _("Please supply the message using either -m or -F option.\n"));
> > +                             exit(1);
> > +                     }
> > +             } else if (trailer_args->nr) {
>
> When both should_edit and trailer_args->nr are true, this block will
> not be entered.  We first do the "amend_file" thing, and then run an
> editor on it, and that is the end of the story in that case.
>
> When we do not have should_edit (e.g., -m "tag message" is given),
> we would have run "amend_file" thing on it to tweak the message,
> and we come in here.
>
> > +                     strbuf_reset(buf);
> > +                     if (strbuf_read_file(buf, path, 0) < 0) {
> > +                             fprintf(stderr,
> > +                                     _("Please supply the message using either -m or -F option.\n"));
> > +                             exit(1);
>
> Does this error message make sense here in this context?  The
> earlier one was introduced by 7198203a (editor.c: Libify
> launch_editor(), 2008-07-25)---after we fail to run the editor, as
> we somehow seem to be unable to run an editor, we suggest the user
> to give us a message in other ways.  But this one is different.  The
> user gave us in one of these other ways already instead of using an
> editor, but mucking with that with the "amend_file" thing somehow
> made it unreadable.  Shouldn't it be more like
>
>         die_errno(_("failed to read '%s'"), path);
>
> or something along that line?

I didn't realize that the first message is intended to augment more expressive failure messages previously printed in launch_editor(). Knowing that, your suggested message will point users in the right direction much more effectively. Also i guess die() probably preferable since unlike launch_editor(), which may signal non-exceptional failure, this error is more likely to be a bug.

However, in service of helping users find workarounds, shouldn't we tell them --trailer may be the culprit?

> Failed to read '%s'. Try again without --trailer (use -e or -F to add trailers manually).
Previous: Junio C HamanoNext: Junio C Hamano
Message 17 of 37 in “builtin/tag.c: add --trailer arg”
  1. builtin/tag.c: add --trailer argJohn Passaro via GitGitGadget, Apr 29, 2024
  2. Patrick SteinhardtApr 29, 2024
  3. John PassaroApr 29, 2024
  4. John PassaroApr 29, 2024
  5. Junio C HamanoApr 29, 2024
  6. Junio C HamanoApr 29, 2024
  7. John PassaroApr 29, 2024
  8. Junio C HamanoApr 29, 2024
  9. builtin/tag.c: add --trailer argJohn Passaro via GitGitGadget, Apr 29, 2024
  10. 0/3 builtin/tag.c: add --trailer optionJohn Passaro via GitGitGadget, Apr 29, 2024
  11. 1/3 builtin/commit.c: refactor --trailer logicJohn Passaro via GitGitGadget, Apr 29, 2024
  12. Patrick SteinhardtApr 30, 2024
  13. Junio C HamanoApr 30, 2024
  14. 2/3 builtin/tag.c: add --trailer argJohn Passaro via GitGitGadget, Apr 29, 2024
  15. Patrick SteinhardtApr 30, 2024
  16. Junio C HamanoApr 30, 2024
  17. John PassaroApr 30, 2024
  18. Junio C HamanoApr 30, 2024
  19. John PassaroMay 5, 2024
  20. 3/3 po: update git-tag translationsJohn Passaro via GitGitGadget, Apr 29, 2024
  21. Junio C HamanoApr 29, 2024
  22. John PassaroApr 29, 2024
  23. 0/3 builtin/tag.c: add --trailer optionJohn Passaro via GitGitGadget, Apr 30, 2024
  24. 1/3 builtin/commit.c: remove bespoke option callbackJohn Passaro via GitGitGadget, Apr 30, 2024
  25. Patrick SteinhardtMay 2, 2024
  26. 2/3 builtin/commit.c: refactor --trailer logicJohn Passaro via GitGitGadget, Apr 30, 2024
  27. Patrick SteinhardtMay 2, 2024
  28. 3/3 builtin/tag.c: add --trailer optionJohn Passaro via GitGitGadget, Apr 30, 2024
  29. Patrick SteinhardtMay 2, 2024
  30. 0/3 builtin/tag.c: add --trailer optionJohn Passaro via GitGitGadget, May 5, 2024
  31. 1/3 builtin/commit: use ARGV macro to collect trailersJohn Passaro via GitGitGadget, May 5, 2024
  32. John PassaroMay 7, 2024
  33. Junio C HamanoMay 7, 2024
  34. 2/3 builtin/commit: refactor --trailer logicJohn Passaro via GitGitGadget, May 5, 2024
  35. 3/3 builtin/tag: add --trailer optionJohn Passaro via GitGitGadget, May 5, 2024
  36. Patrick SteinhardtMay 6, 2024
  37. Junio C HamanoMay 6, 2024

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.