Stefan Sperling wrote:
> 1) Why not just abort on exit code != 0?
Because the current behavior of proceeding with the merge, leaving the
conflicts is more useful in the common case.
There are two aspects of the current behavior I have issue with:
a) Why, when running "--accept launch", it doesn't try running the merge
script on subsequent conflicting files. While this would make sense for
SVN_ERR_CL_NO_EXTERNAL_MERGE_TOOL, just because the merge tool wasn't
able to resolve conflicts for one particular file doesn't mean it
wouldn't be able to resolve conflicts for subsequent files in the same
merge.
b) Why it chooses merged instead of postpone for the (first) conflicted
file the merge tool is unable to resolve. This appears to me to be a bug.
> 2) Could you extend your patch to document the new behaviour somewhere?
> I don't know the appropriate place to put this information. But I
> suppose you know where merge tool authors would be looking for this
> kind of information?
>
Existing documentation on the merge tool is scarce. The only place I've
found information useful for writing a merge tool is a rather oblique
reference in the "Using External Differencing and Merge Tools" section
of the Subversion book. Even that omits the SVN_MERGE environment variable.
My understanding is that updates to the book are done through a separate
mechanism.
------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=995253
Received on 2008-12-29 22:01:23 CET