Hi Kristis, On Mon, Jul 12, 2010 at 08:40, Kristis Makris <[email protected]> wrote:
> On Sun, 2010-07-11 at 23:30 +0300, Yavor Nikolov wrote: > > Hi Kristis, > > > > I share your point of view - I'd prefer to have either all changes > > successful or reject everything (including svn commit). But I'm not > > sure that would be possible to implement. > > It may by difficult to implement, but not impossible. > > Something similar is happening already with integration_add_tag. It is > issued only after integration_tag_exists reports that the tag does not > exist yet. > Just doing some checks before the actual change in Bugtracker doesn't guarantee Bugtracker will succeed committing the real change. The proper way to implement such thing seems to be some kind of distributed transaction with SCM and Bugtracker involved; using something like Two-phase commit or something like that... > > Additional functionality which may serve as partial workaround to > > above issue: > > 1) keep record of unsuccessful bugtracker operations > > 2) Make it possible (implement some utility) to facilitate to fix > > failed requests to bugtracker and attempt to apply them again. > > That's a possible course of action, but I don't like it. If there's a > problem committing, then committing should be rejected. Leting the > commit through with the expectation that some other utility will fix > something at a later time sounds like trouble. > Yes. I meant we're already in trouble. In that respect I meant it may be good at least being able to somehow fix the things after something has gone wrong (with some risk to mess them up more...). Getting things fine at first place is should be of course the preferred use case. /In fact such a thing may be useful in some other scenarios... E.g. - sometimes people forget to fill-in the proper bug id-s on SCM commit, or just mistype them. So sometimes I just want to generate a record in Bugzilla for a particular svn revision/. Regards, Yavor
_______________________________________________ scmbug-users mailing list [email protected] http://lists.mkgnu.net/cgi-bin/mailman/listinfo/scmbug-users
