info-cvs
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

should I be able to create a conflict with this setup?


From: Jacob Burckhardt
Subject: should I be able to create a conflict with this setup?
Date: Thu, 9 Nov 2000 15:53:51 -0800 (PST)

John Klassa writes:
>
> ...
>
> Oddly, I had one file with several conflicts in it, some
> large.  This made absolutely no sense to me.  The trunk and the dev
> branch are identical right after a merge, and the dev branch only has so
> much done to it before I go ahead and merge down to the trunk, so that
> it and the trunk are in sync again.
> 
> If no changes occur on the trunk at all, how can I possibly get a
> conflict when I merge from the dev branch down to the trunk?  Doesn't a
> conflict *require* that the file has changed in both places, in a way
> that conflicts?

Yes.  However, the diff3 code has a bug which causes it to incorrectly
say that there are conflicts even when one of the two parallel
branches has no changes.  CVS 1.11 contains the diff3 source code and
so this bug also occurs when using CVS 1.11.  Here is a reproducible
test case which shows this bug.

The test case contains two files called OLDFILE and YOURFILE.  I run
diff3 like this:

diff3 -m OLDFILE OLDFILE YOURFILE

Note that the diff documentation says that the way to run diff3 is:

     diff3 OPTIONS... MINE OLDER YOURS

Notice that the the MINE and OLDER arguments I pass, specify the same
file.  That means there are no changes going from OLDER to MINE.
Therefore there are no changes from OLDER to MINE which are in conflict 
with changes from OLDER to YOURS.  Therefore there should be no
conflicts at all.  But yet the above command does output conflicts.

To see how this bug also occurs with CVS, run the attached script
cvstest.csh.  Before I can get cvs to merge these two files, I have
to do a lot of set up which is why the test script is so complicated.

I submitted a bug report of this to address@hidden on
Sept. 12, 2000, but I have not received a reply.

Note that Peter Miller's fmerge utility correctly handles the above
case.  That utility is at
http://www.canb.auug.org.au/~millerp/fhist.html

Attachment: OLDFILE
Description: Binary data

Attachment: YOURFILE
Description: Binary data

Attachment: cvstest.csh
Description: Binary data


reply via email to

[Prev in Thread] Current Thread [Next in Thread]