Excuse the late jumping in, but ...
On March 22, 2004 06:50 pm, Dimitrie O. Paun wrote:
As for the issue of nighly build vs. randomly initiated build: we can argue till the cows come home. It seems that we have different points of view. If you want to have CVS-triggered builds for WRT, all the power to you, they makes sense in that (controlled) environment. For the nightly builds however, I would really like having a build that's kicked off at a precise moment in time (and as we discussed, 10GMT seems a good choice).
Is it not possible, somehow, for Alexandre to indicate when he has finished a run of commits? I know he endeavours to keep CVS stable by committing under the "Do No Harm" principle, but obviously the chances of hitting a snag are greater during a run of commits than once they are done. Certainly the *usefulness* of the testing is greater once the run is over, no matter what time(s) of day he may choose to do such processing. Presumably Alexandre has some mechanism of his own to incrementally test things during the wine-patches --> wine-cvs processing. Presumably also, there is some point in that mechanism of his where he might like having a large green button to the effect of; "thou alt solid locally, release thyself unto the distributed test network whereapon I shall know thy fidelity on the morrow". Or words to the effect ...
Please try to understand that as a developer on Wine, having this extra variability adds _nothing_ but headache. It makes it harder to think about the problem, to talk about it, to script for it, etc. I don't need it, and I don't want it.
Exactly, and again Alexandre is not only another (the?) wine developer, but also the one pulling the CVS strings - surely there must be some painless way to trigger these "snapshot moments" from his end. That is, after timezones have been argued to hell and back, really the only place where the right moment for a useful test can be accurately determined (and without unduly delay).
Cheers, Geoff