git@vger.kernel.org mailing list mirror (one of many)
 help / color / mirror / code / Atom feed
From: Junio C Hamano <gitster@pobox.com>
To: Thomas Gummerer <t.gummerer@gmail.com>
Cc: git@vger.kernel.org, Marc Strapetz <marc.strapetz@syntevo.com>
Subject: Re: [PATCH v5 2/3] stash push: avoid printing errors
Date: Tue, 20 Mar 2018 09:54:21 -0700	[thread overview]
Message-ID: <xmqqzi32sn5u.fsf@gitster-ct.c.googlers.com> (raw)
In-Reply-To: <20180319232156.30916-3-t.gummerer@gmail.com> (Thomas Gummerer's message of "Mon, 19 Mar 2018 23:21:55 +0000")

Thomas Gummerer <t.gummerer@gmail.com> writes:

> ...
> Fix this by avoiding the 'git clean' if a pathspec is given, and use the
> pipeline that's used for pathspec mode to get rid of the untracked files
> as well.

That made me wonder if we can get rid of 'git clean' altogether by
pretending that we saw a pathspec '.' that match everything when no
pathspec is given---that way we only have to worry about a single
codepath.

But I guess doing it this way can minimize potential damage.  Those
who do not use pathspec when running "git stash" won't be affected
even if this change had some bugs ;-)

> diff --git a/git-stash.sh b/git-stash.sh
> index 4c92ec931f..5e06f96da5 100755
> --- a/git-stash.sh
> +++ b/git-stash.sh
> @@ -308,14 +308,16 @@ push_stash () {
>  	if test -z "$patch_mode"
>  	then
>  		test "$untracked" = "all" && CLEAN_X_OPTION=-x || CLEAN_X_OPTION=
> -		if test -n "$untracked"
> +		if test -n "$untracked" && test $# = 0
>  		then
>  			git clean --force --quiet -d $CLEAN_X_OPTION -- "$@"
>  		fi
>  
>  		if test $# != 0
>  		then
> -			git add -u -- "$@"
> +			test -z "$untracked" && UPDATE_OPTION="-u" || UPDATE_OPTION=
> +			test "$untracked" = "all" && FORCE_OPTION="--force" || FORCE_OPTION=
> +			git add $UPDATE_OPTION $FORCE_OPTION -- "$@"
>  			git diff-index -p --cached --binary HEAD -- "$@" |
>  			git apply --index -R
>  		else

Thanks, I'll take the change as-is.

I however wonder if we restructure the code to

	if test $# = 0
	then
		# no pathspec
		if test -n "$untracked"
		then
			git clean --force --quiet -d $CLEAN_OPTION -- "$@"
		fi
		git reset --hard -q
	else
		test -z "$untracked" && UPDATE=-u || UPDATE=
		test "$untracked" = all && FORCE=--force || FORCE=
		git add $UPDATE $FORCE-- "$@"
		git diff-index -p --cached --binary HEAD -- "$@" |
		git apply --index -R
	fi

it becomes easier to understand what is going on.

That way, once we have a plumbing command to help the else clause of
the above, i.e. "git checkout --index <tree-ish> -- <pathspec>"
[*1*], then we can lose the if/then/else and rewrite the whole "we
have created stash, so it's time to get rid of local modifications
to the paths that match the pathspec" code to:

	if test "$untracked"
	then
		git clean --force --quiet -d $CLEAN_OPTION -- "$@"
	fi
	git checkout --index HEAD -- "$@"

[Footnote]
cf. https://public-inbox.org/git/xmqq4loqplou.fsf@gitster.mtv.corp.google.com/

What we want in the case in the code in question when there is
pathspec is "match the index entries and the working tree files to
those that appear in a given tree for paths that match the given
pathspec".  This is close to "git checkout <tree-ish> -- <pathspec>"
but not quite.  Current "git checkout <tree-ish> -- <pathspec>" is
an overlay operation in that paths that match <pathspec> but do not
exist in <tree-ish> are *NOT* removed from the working tree.  We
obviously cannot change the behaviour of the command.

But we can add an option to ask for the new behaviour.  In general,
for an operation that affects the index and the working tree, we can
have "--cached" mode that affects only the index without touching
the working tree, and "--index" mode that affects both.

"git reset <tree-ish> -- <pathspec>", which is a UI mistake, is
better spelled "git checkout --cached <tree-ish> -- <pathspec>".  We
take paths that match <pathspec> from <tree-ish> and stuff into the
index, and remove entries from the index for paths that do not exist
in <tree-ish>.  And if we extend that to "--index" mode, that is
exactly what we want to happen.

  reply	other threads:[~2018-03-20 16:54 UTC|newest]

Thread overview: 33+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-03-03  9:44 git stash push -u always warns "pathspec '...' did not match any files" Marc Strapetz
2018-03-03 15:46 ` Thomas Gummerer
2018-03-04 10:44   ` Marc Strapetz
2018-03-09 22:18     ` Junio C Hamano
2018-03-10  9:18       ` Marc Strapetz
2018-03-10 11:12         ` Thomas Gummerer
2018-03-14 21:46           ` [PATCH v2 1/2] stash push: avoid printing errors Thomas Gummerer
2018-03-14 21:46             ` [PATCH v2 2/2] stash push -u: don't create empty stash Thomas Gummerer
2018-03-15 20:09               ` Junio C Hamano
2018-03-16 20:10                 ` Thomas Gummerer
2018-03-15  8:51             ` [PATCH v2 1/2] stash push: avoid printing errors Marc Strapetz
2018-03-16 20:12               ` Thomas Gummerer
2018-03-16 20:43             ` [PATCH v3 0/2] stash push -u -- <pathspec> fixes Thomas Gummerer
2018-03-16 20:43               ` [PATCH v3 1/2] stash push: avoid printing errors Thomas Gummerer
2018-03-16 21:31                 ` Junio C Hamano
2018-03-16 20:43               ` [PATCH v3 2/2] stash push -u: don't create empty stash Thomas Gummerer
2018-03-16 22:37               ` [PATCH v4 0/3] stash push -u -- <pathspec> fixes Junio C Hamano
2018-03-16 22:37                 ` [PATCH v4 1/3] stash: fix nonsense pipeline Junio C Hamano
2018-03-16 22:37                 ` [PATCH v4 2/3] stash push: avoid printing errors Junio C Hamano
2018-03-16 22:37                 ` [PATCH v4 3/3] stash push -u: don't create empty stash Junio C Hamano
2018-03-17 11:36                 ` [PATCH v4 0/3] stash push -u -- <pathspec> fixes Thomas Gummerer
2018-03-19 23:21                 ` [PATCH v5 " Thomas Gummerer
2018-03-19 23:21                   ` [PATCH v5 1/3] stash: fix nonsense pipeline Thomas Gummerer
2018-03-19 23:21                   ` [PATCH v5 2/3] stash push: avoid printing errors Thomas Gummerer
2018-03-20 16:54                     ` Junio C Hamano [this message]
2018-03-21 21:36                       ` Thomas Gummerer
2018-03-21 21:53                         ` [PATCH] stash: drop superfluos pathspec parameter (was: Re: [PATCH v5 2/3] stash push: avoid printing errors) Thomas Gummerer
2018-03-21 22:07                           ` [PATCH] stash: drop superfluos pathspec parameter Junio C Hamano
2018-03-21 21:56                         ` [PATCH v5 2/3] stash push: avoid printing errors Junio C Hamano
2018-03-19 23:21                   ` [PATCH v5 3/3] stash push -u: don't create empty stash Thomas Gummerer
2018-03-20 10:06                   ` [PATCH v5 0/3] stash push -u -- <pathspec> fixes Marc Strapetz
2018-03-19 15:44               ` [PATCH v3 0/2] " Marc Strapetz
2018-03-19 21:51                 ` Thomas Gummerer

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=xmqqzi32sn5u.fsf@gitster-ct.c.googlers.com \
    --to=gitster@pobox.com \
    --cc=git@vger.kernel.org \
    --cc=marc.strapetz@syntevo.com \
    --cc=t.gummerer@gmail.com \
    /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).