git@vger.kernel.org mailing list mirror (one of many)
 help / color / mirror / code / Atom feed
From: Philippe Blain <levraiphilippeblain@gmail.com>
To: Elijah Newren <newren@gmail.com>
Cc: "Git mailing list" <git@vger.kernel.org>,
	"Eric Sunshine" <sunshine@sunshineco.com>,
	"Jeff King" <peff@peff.net>,
	"Nguyễn Thái Ngọc Duy" <pclouds@gmail.com>
Subject: Re: [BUG?] Weird interaction between `git -C`, aliases and worktrees
Date: Wed, 14 Oct 2020 18:35:39 -0400	[thread overview]
Message-ID: <88D2A179-806F-4321-828A-359692E646B1@gmail.com> (raw)
In-Reply-To: <CABPp-BF2Stb_UfNEqYOBPjvm09oO8=ST685kZ0+U2PtPRz5_mg@mail.gmail.com>

Hi Elijah,

> Le 14 oct. 2020 à 18:02, Elijah Newren <newren@gmail.com> a écrit :
> 
> On Wed, Oct 14, 2020 at 8:27 AM Philippe Blain
> <levraiphilippeblain@gmail.com> wrote:
>> 
>> Hello all,
>> 
>> Up to recently I had an alias 'st' for `git status`, i.e. `git config alias.st status`.
>> This works well in the main working tree as well as secondary worktrees (`git worktree add ...`),
>> and also shows paths relative to the current directory if I'm not at the root of the repo,
>> just like `git status`.
>> 
>> Now I wanted to change my alias so that it ran `git status` and then an additional command, i.e.
>> `git config alias.st '!git status && date'` (for demonstration purposes).
>> This works correctly in the main worktree and secondary worktrees, but since aliases
>> run from the root of the repo, the paths in the output of 'git st' are not relative
>> to the current directory.
>> 
>> So my next attempt was `git config alias.st '!git -C "${GIT_PREFIX}" status && date'`,
>> so that paths would be relative in the output. However, this works in the main worktree
>> but fails badly in the secondary worktree:
>> 
>> ```
>> rm -rf test &&
>> rm -rf test2 &&
>> git init test &&
>> cd test &&
>> mkdir folder &&
>> date>>folder/file &&
>> git add folder/file &&
>> git ci -m "commit" &&
>> git config alias.st '!git -C "${GIT_PREFIX}" status && date' &&
>> date >>folder/file &&
>> echo '=== `git st` in main worktree at root ===' &&
>> git st &&
>> cd folder &&
>> echo '=== `git st` in main worktree in folder ===' &&
>> git st &&
>> git worktree add ../test2 &&
>> cd ../test2 &&
>> date >>folder/file &&
>> echo '=== `git st` in secondary worktree at root ===' &&
>> git st &&
>> cd folder &&
>> echo '=== `git st` in secondary worktree in folder ===' &&
>> git st
>> ```
>> 
>> The last commands ouputs:
>> 
>> ```
>> === `git st` in secondary worktree in folder ===
>> On branch test2
>> Changes not staged for commit:
>>  (use "git add/rm <file>..." to update what will be committed)
>>  (use "git restore <file>..." to discard changes in working directory)
>>        deleted:    folder/file
>> 
>> Untracked files:
>>  (use "git add <file>..." to include in what will be committed)
>>        file
>> 
>> no changes added to commit (use "git add" and/or "git commit -a")
>> ```
>> 
>> So something is wrong in the automatic worktree path detection...
>> If I replace the alias with
>> 
>>    git config alias.st '!git -C "${GIT_PREFIX}" --work-tree=$PWD status && date'
>> 
>> then it works correctly, but I have a feeling this should not be necessary...
>> 
>> I've CC-ed Eric (because of his work on `git worktree`) and Elijah (because
>> I remember he worked on `dir.c` so maybe this is related to that code, but I'm
>> not sure).
>> 
>> Cheers,
>> Philippe
> 
> I don't think there's anything dir.c-specific here (thank
> goodness...).  I extended your aliases slightly to echo the command
> they would run (so I could see the value of $GIT_PREFIX) as well as to
> print all the GIT_* environment variables, and found that GIT_DIR was
> set as well as GIT_PREFIX, so you can duplicate the funny results with
> 
> GIT_DIR=$WHEREVER git -C folder status
> 
> or even
> 
> git --git-dir=$WHEREVER -C folder status
> 
> In fact, you don't need any special worktrees setup; using these
> commands will trigger the funny status output.  Looking at the
> documentation for core.worktree, I see
> 
> "If --git-dir or GIT_DIR is specified but none of --work-tree,
> GIT_WORK_TREE and core.worktree is specified, the current working
> directory is regarded as the top level of your working tree."

Yes, this is in line with git(1) [1] :

"Specifying the location of the ".git" directory using this option (or GIT_DIR environment variable)
 turns off the repository discovery that tries to find a directory with ".git" subdirectory
(which is how the repository and the top-level of the working tree are discovered), 
and tells Git that you are at the top level of the working tree. 
If you are not at the top-level directory of the working tree,
 you should tell Git where the top-level of the working tree is, 
with the --work-tree=<path> option (or GIT_WORK_TREE environment variable)"

> and indeed, I find that if I run from the toplevel:
> 
> git --work-tree=folder/ status
> 
> then I see the same output that you are complaining about.  So, it
> seems that "the current working directory is regarded as the top level
> of your working tree" is implemented to take effect AFTER the "-C
> folder" first changes the current working directory.
> 
> I haven't tracked down where in the code this happens, but I suspect
> that this is what is happening and is the culprit behind the behavior
> you are seeing.  If I am right, it doesn't answer whether this is a
> bug, of course -- it could be that this ordering is intentional and
> desired, or it could be that this ordering is not wanted.  Anyway,
> does this help you track it down?


I found two threads [2] and [3] which are manifestations of similar (or the same ?) problems...

Maybe git-config(1) could document which variables will 
be set in aliases that shell out, but even that might be tricky because it
seems the answer is "it depends" (according to [2])...

I also found 57ea7123c8 (git.c: make sure we do not 
leak GIT_* to alias scripts, 2015-12-20), in topic 'nd/clear-gitenv-upon-use-of-alias',
which tries to clear out GIT_* variables from aliases environment,
which I guess is defeated by the fact that I run `git -C` in my alias...

Anyway thanks for your answer.

[1] https://git-scm.com/docs/git#Documentation/git.txt---git-dirltpathgt
[2] https://lore.kernel.org/git/20200130102933.GE840531@coredump.intra.peff.net/
[3] https://lore.kernel.org/git/20200228190218.GC1408759@coredump.intra.peff.net/

Philippe.


  reply	other threads:[~2020-10-15  2:14 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-10-14 15:27 [BUG?] Weird interaction between `git -C`, aliases and worktrees Philippe Blain
2020-10-14 22:02 ` Elijah Newren
2020-10-14 22:35   ` Philippe Blain [this message]
2020-10-16  0:23     ` Jeff King

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=88D2A179-806F-4321-828A-359692E646B1@gmail.com \
    --to=levraiphilippeblain@gmail.com \
    --cc=git@vger.kernel.org \
    --cc=newren@gmail.com \
    --cc=pclouds@gmail.com \
    --cc=peff@peff.net \
    --cc=sunshine@sunshineco.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).