I still not sure about this myself, but I believe the magic is as follows: The diff file has the format:
diff -r b0689af42a71 indra/cmake/LLAddBuildTest.cmake --- a/indra/cmake/LLAddBuildTest.cmake Wed Dec 15 22:44:21 2010 +0100 +++ b/indra/cmake/LLAddBuildTest.cmake Sun Dec 19 16:06:01 2010 +0100 etc, where the "b0689af42a71" must be the parent already in viewer-development that you based your diff on. Took me a while to figure that out and then I thought 'screw this check' and wrote a script that just forges that ID, using the following magic: UPSTREAMID="$(hg log -r 'max(not outgoing())' | grep '^changeset' | sed -e 's/.*://')" If you have a clean clone with a single patch, then that is the parent of your patch, and you can generate the diff with 'hg export'. Now,.. if you upload a patch on top of another patch (for the same RB issue) it asks you for an optional parent diff. You leave that empty when the new diff replaces the old one, but if it's an incremental diff, you just put there whatever diff needs to be applied before applying this new upload. Hope that helps, Aleric
_______________________________________________ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges
