Re: Patch command execution

Karl Fogel wrote:

>Comments? Suggestions? Rotten fruit? :-)
>
A little of each.

Personally I find CVS's conflict markers the most horrible idea in the
world. It may not seem so to people who equate "version control" with "C
source files", which it seems some people do. But there are other
formats and languages out there (think of Python's multiline
docstrings). I also think the current behaviour where we forbid a commit
if a .rej file is present is a huge plus. It prevents the user from
accidentally committing a conflicted file, while "resolving" the
conflict can be as easy as removing the .rej file.

This is obviously a case where we'll never get consensus on the colour
of the bikeshed. Therefore it must be configurable. I'd go with Zack's
suggestion that we always expect to do a 3-way merge. The prerequisite
for that is that we have all three (original, local and reporitory)
versions of the file available, /which we already do/. Those three files
can passed on to a client-side merger.

I'd suggest our command line client has two merge methods built in: one
that creates .rej files (using the method Zack outluned), and one that
creates conflict markers. Adding other merge tools can be left for later.

-- 
Brane Čibej   <brane_at_xbc.nu>   http://www.xbc.nu/brane/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org