git.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Junio C Hamano <gitster@pobox.com>
To: Greg KH <greg@kroah.com>
Cc: git@vger.kernel.org
Subject: Re: renaming a file into a directory causes a pull error on older repos
Date: Mon, 17 Mar 2008 17:16:41 -0700	[thread overview]
Message-ID: <7vlk4ganti.fsf@gitster.siamese.dyndns.org> (raw)
In-Reply-To: <20080316043138.GA7942@kroah.com> (Greg KH's message of "Sat, 15 Mar 2008 21:31:38 -0700")

Greg KH <greg@kroah.com> writes:

> The problem:
>   If you turn a file in a repository into a directory, and place files
>   in that dir and commit it, any other person who had that repo cloned
>   somewhere else will get an error when they try to pull and update
>   their version.
>
> The error for me is:
> 	fatal: Entry 'stats/results-18-22.txt' would be overwritten by merge. Cannot merge.
> 	Merge with strategy recursive failed.
>
> I had turned the file "stats" into a directory.

So...

 - originally "stats" was a file.

 - then one branch removes it and creates stats/results-18-22.txt file.

 - another branch keeps working elsewhere in the tree but has not touched
   the "stats" file.

Now, the above error message complains about stats/results-18-22.txt being
overwritten, so I presume that:

 - You have checked out the branch that has stats/results-18.22.txt;

 - You are merging the other branch that still had stats as a file into
   that checked out branch with stats/results-18.22.txt file.

Are these presumptions correct?

Now, merge-recursive may be riddled with bugs in directory-file conflict
detection area.  The way it detects conflicts is quite bogus --- it builds
a list of files and directories in ancestor, our side and the other side,
and anything that changes directoryness is marked as conflict, when the
right thing to do is to complain only if the checking out of the result
needs to have a file and a directory at the same place.

But I do not think the above error message is from merge-recursive proper.
"Entry X would be overwritten by merge. Cannot merge." is an error message
the 3-way read-tree (driven from merge-recursive) issues when you have
local changes to file X that will go away as the result of the merge, to
prevent us from losing your local changes to the file.  Didn't you have
changes to that file when you did the merge?

I have spotted an unrelated bug in git-merge-one-file.sh that would have
caused something similar symptom when you had used "resolve" strategy, by
the way (unfortunately I do not think it applies to merge-recursive).

  reply	other threads:[~2008-03-18  0:17 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2008-03-16  4:31 renaming a file into a directory causes a pull error on older repos Greg KH
2008-03-18  0:16 ` Junio C Hamano [this message]
2008-03-18  0:48   ` [PATCH] git-merge-one-file: fix longstanding stupid thinko Junio C Hamano
2008-03-19  1:51   ` renaming a file into a directory causes a pull error on older repos Greg KH
2008-03-19  6:31     ` Junio C Hamano

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

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=7vlk4ganti.fsf@gitster.siamese.dyndns.org \
    --to=gitster@pobox.com \
    --cc=git@vger.kernel.org \
    --cc=greg@kroah.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.
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 NNTP newsgroup(s).