On Fri, Dec 18, 2009 at 02:39:04PM +0100, Johan Sørensen wrote: > On Sat, Dec 5, 2009 at 2:45 PM, Oswald Buddenhagen <[email protected]> wrote: > > 3) this is subject to connectivity problems. while the post-receive > > thing could kinda work if complete atomic ref snapshots are queued, it > > is a non-starter for a pre-receive hook (which needs to be called in > > real-time). > > To be honest, I'm quite hesitant to adding any pre-receive hooks that > does any kind of non-trivial work. See, if pushing is slow due to that > hook people won't complain about the hook, they'll complain, to me, > that Gitorious is slow. > well, yes. verifying that the hooks are Fast Enough (TM) would be part of your responsibility when integrating them.
> So I'd think hard and long about whether that hook really is essential > at that particular point in your workflow? > in svn, we have the various checks in the post-commit hook. this a) leads to a messy history (as all fixes are inherently after-the-fact) and b) reduces the probability that an issue is fixed in a timely manner (if at all - "what's in, is in"). so maybe it is not essential, but it is the right thing to do. an alternative to pre-receive hooks are pre-commit hooks, but then everyone has to install them, and that's a pretty safe way to make most people not use them (it works without them, after all). _______________________________________________ Kde-scm-interest mailing list [email protected] https://mail.kde.org/mailman/listinfo/kde-scm-interest
