Re: [PATCH 00/13] bisect: v2.30.0 "run" regressions + make it built-in

[Johannes Schindelin <Johannes.Schindelin@gmx.de>]

[-- Attachment #1: Type: text/plain, Size: 3915 bytes --]

Hi Ævar,

I see you Cc:ed me on this, but I have to admit that I am not motivated to
review this patch series because it seems to be designed to interfere with
`js/bisect-in-c` instead of being an honest effort to help me get that
patch series over the finish line.

I'd much rather see you assist me in the most efficient/minimal way
possible to get `js/bisect-in-c` into a shape that can be integrated.

Ciao,
Johannes

On Fri, 4 Nov 2022, Ævar Arnfjörð Bjarmason wrote:

> This fixes the regression Lukáš Doktor reported in [1], and also gets
> us the full way to a builtin/bisect.c and "git rm git-bisect.sh".
>
> Only 1-4/13 here are needed to fix the "git bisect run <cmd> [...]
> --log" regression Lukáš reported, but as Jeff points out we'd still
> conflate "--bisect-*" with the user arguments. That's fixed in 11/13
> here.
>
> The 1-4/13 here also fixes other but probably more minor "git bisect
> run" regressions in v2.30.0, we changed the output in a few ways
> without intending it. 4/13 gets us mostly back to v2.29.0 behavior,
> 5/13 keeps the best of it and the current output.
>
> I think for the v2.30.0 regressions we're better off with just
> something like 1-4/13 here for now, and possibly 5/13 too.
>
> But getting to the point of fixing the root cause of "--bisect-*"
> being conflated is going to take quite a bit of churn. In the
> side-thread Đoàn's diffstat is on the order of 1/2 of the size of this
> series, and this gives us built-in "bisect".
>
> The 6-13 here is something I had already for a couple of days, I was
> seeing if I could distill Johannes's [2] down to something much
> smaller, to just make a beeline towards a built-in bisect.
>
> Johannes's refactors the "term" passing in [3], and Đoàn ends up
> needing to do much the same in [4].
>
> Here in 9/13 I instead just extend the OPT_SUBCOMMAND() API so it's
> able to accept function callbacks with custom signatures, which
> eliminates the need for most of that refactoring. 11/13 then makes use
> of it.
>
> 1. https://lore.kernel.org/git/1cb1c033-0525-7e62-8c09-81019bf26060@redhat.com/
> 2. https://lore.kernel.org/git/pull.1132.v6.git.1661885419.gitgitgadget@gmail.com/
> 3. https://lore.kernel.org/git/92b3b116ef8f879192d9deb94d68b73e29d5dcd6.1661885419.git.gitgitgadget@gmail.com/
> 4. https://lore.kernel.org/git/081f3f7f9501012404fb9e59ab6d94f632180b53.1667561761.git.congdanhqx@gmail.com/
>
> Johannes Schindelin (3):
>   bisect--helper: remove dead --bisect-{next-check,autostart} code
>   bisect--helper: make `state` optional
>   Turn `git bisect` into a full built-in
>
> Ævar Arnfjörð Bjarmason (10):
>   bisect tests: test for v2.30.0 "bisect run" regressions
>   bisect: refactor bisect_run() to match CodingGuidelines
>   bisect: fix output regressions in v2.30.0
>   bisect run: fix "--log" eating regression in v2.30.0
>   bisect run: keep some of the post-v2.30.0 output
>   bisect test: test exit codes on bad usage
>   bisect--helper: emit usage for "git bisect"
>   bisect--helper: have all functions take state, argc, argv, prefix
>   parse-options API: don't restrict OPT_SUBCOMMAND() to one *_fn  type
>   bisect--helper: convert to OPT_SUBCOMMAND_CB()
>
>  Makefile                               |   3 +-
>  builtin.h                              |   2 +-
>  builtin/{bisect--helper.c => bisect.c} | 250 +++++++++++++------------
>  git-bisect.sh                          |  84 ---------
>  git.c                                  |   2 +-
>  parse-options.c                        |   9 +-
>  parse-options.h                        |  31 ++-
>  t/t6030-bisect-porcelain.sh            | 109 +++++++++++
>  8 files changed, 277 insertions(+), 213 deletions(-)
>  rename builtin/{bisect--helper.c => bisect.c} (86%)
>  delete mode 100755 git-bisect.sh
>
> --
> 2.38.0.1452.g710f45c7951
>
>