From: Johannes Schindelin <Johannes.Schindelin@gmx.de>
To: phillip.wood@dunelm.org.uk
Cc: Jochen Sprickerhof <jochen@sprickerhof.de>,
Junio C Hamano <gitster@pobox.com>,
git@vger.kernel.org
Subject: Re: [PATCH] add -p: coalesce hunks before testing applicability
Date: Fri, 22 Mar 2019 15:06:25 +0100 (STD) [thread overview]
Message-ID: <nycvar.QRO.7.76.6.1903221453360.41@tvgsbejvaqbjf.bet> (raw)
In-Reply-To: <d6a8f77b-0a83-90ae-a7fb-a3954ac3b346@talktalk.net>
Hi,
On Thu, 13 Sep 2018, Phillip Wood wrote:
> On 03/09/2018 20:01, Jochen Sprickerhof wrote:
>
> > * Phillip Wood <phillip.wood@talktalk.net> [2018-08-30 14:47]:
> >> When $newhunk is created it is marked as dirty to prevent
> >> coalesce_overlapping_hunks() from coalescing it. This patch does not
> >> change that. What is happening is that by calling
> >> coalesce_overlapping_hunks() the hunks that are not currently selected
> >> are filtered out and any hunks that can be coalesced are (I think that
> >> in the test that starts passing with this patch the only change is the
> >> filtering as there's only a single hunk selected).
> >
> > Agreed here. It would be enough to include the first hunk in the test to
> > make it fail again. Still I would see the patch as going in the right
> > direction as we need something like coalesce_overlapping_hunks() to make
> > the hunks applicable after the edit.
>
> Yes in the long term we want to be able to coalesce edited hunks, but I
> think it is confusing to call coalesce_overlapping_hunks() at the moment
> as it will not coalesce the edited hunks.
Agreed. I actually have code to coalesce even edited hunks, but it is all
in C.
> >> This is a subtle change to the test for the applicability of an
> >> edited hunk. Previously when all the hunks were used to create the
> >> test patch we could be certain that if the test patch applied then if
> >> the user later selected any unselected hunk or deselected any
> >> selected hunk then that operation would succeed. I'm not sure that is
> >> true now (but I haven't thought about it for very long).
> >
> > I'm not sure here. If we use the same test from t3701, do s(plit),
> > y(es), e(dit), it would fail later on. Can you come up with an
> > example?
>
> I think that if you split a hunk, edit the first subhunk, transforming a
> trailing context line to a deletion then try if you try to stage the
> second subhunk it will fail. With your patch the edit will succeed as
> the second subhunk is skipped when testing the edited patch. Then when
> you try to stage the second subhunk it will fail as it's leading context
> will contradict the trailing lines of the edited subhunk. With the old
> method the edit failed but didn't store up trouble for the future.
Indeed, this is a problem I also stumbled over.
> >> We could restore the old test condition and coalesce the hunks by
> >> copying all the hunks and setting $hunk->{USE}=1 when creating the
> >> test patch if that turns out to be useful (it would be interesting to
> >> see if the test still passes with that change).
> >
> > We set USE=1 for $newhunk already, or where would you set it?
>
> To match the old test it needs to be set on the hunks we've skipped or
> haven't got to yet so they're all in the patch that's tested after
> editing a hunk.
The way I fixed this in the C code is by teaching the equivalent of the
`coalesce_overlapping_hunks()` function to simply ignore the equivalent of
`$hunk->{USE}`: the function signature takes an additional `use_all`
parameter, which will override the `use` field.
Furthermore, my C code actually does the coalescing as part of the
`reassemble_patch()` function, feeding the output directly into the
`stdin` of the `git apply` process (with, or without `--check`).
And crucially, my C code does not try to assemble a new `hunks` array, but
simply works in-place, reverting the changes if the hunk edits result in a
patch that does not apply. The Perl version probably does not need that
part, as it is pretty careless with memory (as Perl encourages to do).
See for yourself:
https://github.com/dscho/git/commit/6f8ac4809280f2cd018683ab5199b004ada2350e#diff-f58d2179be56b196b9f35c6d24799a8eR337
Ciao,
Dscho
P.S.: Yes, this is part of my work to complete Slavica's "`git add -i`
in C" project. There are quite a few loose ends to tie, but I can already
use it myself to call `git add -p`, which is what I care most about, as it
regularly takes more than one second to spin up Perl on Windows, which is
friggin' annoying, I tell ya.
next prev parent reply other threads:[~2019-03-22 14:06 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-08-28 8:58 [PATCH] add -p: coalesce hunks before testing applicability Jochen Sprickerhof
2018-08-28 18:07 ` Junio C Hamano
2018-08-30 13:47 ` Phillip Wood
2018-08-30 14:51 ` Junio C Hamano
2018-09-03 19:01 ` Jochen Sprickerhof
2018-09-13 10:20 ` Phillip Wood
2018-09-23 17:16 ` Jochen Sprickerhof
2019-03-22 14:06 ` Johannes Schindelin [this message]
2019-06-02 14:17 ` Phillip Wood
2019-06-03 13:40 ` Johannes Schindelin
2019-06-03 14:59 ` Phillip Wood
2019-06-04 13:32 ` Johannes Schindelin
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
List information: http://vger.kernel.org/majordomo-info.html
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=nycvar.QRO.7.76.6.1903221453360.41@tvgsbejvaqbjf.bet \
--to=johannes.schindelin@gmx.de \
--cc=git@vger.kernel.org \
--cc=gitster@pobox.com \
--cc=jochen@sprickerhof.de \
--cc=phillip.wood@dunelm.org.uk \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
Code repositories for project(s) associated with this public inbox
https://80x24.org/mirrors/git.git
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).