Sounds like this is just what the doctor ordered! I have had this problem, and
wished for a solution exactly like the one you are describing.
Alan Thompson
To: CVS Mailing List <[EMAIL PROTECTED]>
cc: [EMAIL PROTECTED] (bcc: Alan Thompson/Orincon)
Subject: patch for merging long-term branches
We've come up with a patch for rcscmds.c (attached) that makes merges
to/from longterm branches work very nicely for us. We're not sure of
all the ramifications for every possible use case of cvs, but it
seems reasonable that this should be the desired behavior for merges.
When doing a three-way merge versus a file in your working directory,
cvs checks out the two versions to compare against and creates
temporary copies in the working directory. Our patch addresses this
checkout. If the kflags on the file are -kb, then cvs behaves as it
normally would. Also, if explicit kflags are specified on the
command-line, then those are respected. However, if no kflags are
specified on the command-line, and the file is not marked as -kb,
then the checkout (of the two revisions against which the diff is to
be done) uses -kk. This results in comparisons against unexpanded
keywords, which eliminates spurious conflicts based on keyword
expansion alone.
I'd like to suggest incorporation of this patch into cvs, but I'm
curious whether anybody has reasons that it should not be
incorporated.
(Note: I don't take credit for this patch myself; it was actually
created by Jay Sachs, chief architect in charge of core technology
here at iClick.)
rcscmds.patch
--
+-------------------------------------------------------------------+
| Dave Makower <[EMAIL PROTECTED]> |
| Manager of Portal Architecture & Development |
+--------------------------------+----------------------------------+
| iClick, Inc. | (914) 872-8030 |
| 120 Bloomingdale Road | (914) 872-8100 fax |
| 3rd Floor | (914) 872-8000 main |
| White Plains, NY 10605 | http://www.iclick.com/ |
+--------------------------------+----------------------------------+
