Huh... apparently i never sent this first one before sending the shell scriptlet. this was meant to go *before* the previous mail :)
On Mon, 2011-02-28 at 09:45 +0100, Guido Günther wrote: > We're not using stash at all at the momment and try to avoid branch > switching. In order to do so we use git-read-tree and git-write-tree > used at other places. okay, i'll take a look at doing something similar here then. > If I read the code correctly the stash is needed to make sure we have a > clean tree before switching to upstream to update the submodules? That was my impression, yeah. > Since the update of the submodules is a fast forward I wonder if we > can't update them without the switch? Minor correction: the submodules are by default detached heads, so it's more like a reset than a fast forward, afaik. I think the tricky part is that the submodules in the upstream branch might != the submodules in the debian branch, but i'll see if i can find a way to get that without branch switching. > Also we do we update the submodules at all before exporting them? Can't > we assume they're up to date and simply add their contents to the > tarball? We don't update the upstream branch either. afaik, the submodules in the upstream tree are statically defined hashes, so if the upstream tree isn't updated then the hashes aren't updated. but it's possible that the repo has been cloned and the submodules are not yet fetched/available. anyway i'll RTFM a little and see if i can come back with something new along the lines that you've suggested. sean
signature.asc
Description: This is a digitally signed message part
