From mboxrd@z Thu Jan 1 00:00:00 1970 From: Robert Lipe To: Oleg Krivosheev Cc: egcs@cygnus.com Subject: Re: Direct CVS Access to egcs sources Date: Thu, 15 Jan 1998 16:30:00 -0000 Message-id: <19980115012422.08714@dgii.com> References: <11672.884670861@hurl.cygnus.com> X-SW-Source: 1998-01/msg00485.html I have nothing to say in the official egcs policy on this topic. I do have an understanding of how CVS can help glue a geographically challenged team together. Oleg Krivosheev wrote: > how you're going to bump snapshot number/date/location? > Just randomly from time to time? Since not everyone will have CVS access, (much of the world is still UUCP connected, for example) it's my understanding that snapshots will still be made. Even if everyone *did* have CVS access, not everyone will want to track things that closely. We do need some way to distinguish the CVS snapshots of any given moment in time. > And what about people doing hard work testing latest snapshots on different > computer/systems? Looks like it doesn't make any sense now? I must not understand the concern. It still make perfect sense for things to be tested on a wide variety of hardware. Now people have a choice. If they're CVS enabled (and internet connected) and can crank out an update/build/test each hour, they can find out *exactly* when something broke or began working. (If you're actually in that category, perhaps getting outside once in a while would do you some good. :-) If they're not, they can wait for the announcement of the snaps and grab them as always. Of course, if they are CVS enabled and just choose not to track the release of the moment, they can still use CVS to get to the snapshots via the tags that are applied as part of the snapshot process. Having CVS access also allows us to work out patch integration issues in a way that we know they'll apply cleanly to the master sources, making the lives of those with actual write access to the master repository easier. There is no more waiting for someone to submit a patch to the list for CVS users. When someone reports, "Yeah, I committed that last week in dwarf2out.c and reload.c", we can type 'cvs update dwarf2out.c reload.c' and inherit those fixes without nuking everything else and without the developer making an explicit patch. CVS users can also do our testing on different computers/systems when it's best for them instead of trying to synchronize it to the times when snaps are made available. I know the backup cycles on my machine and when my WAN is most idle. I know when I have the most clock cycles available. (hint: usually I'm in bed) Now I can target test cycles to those times instead of waiting for the next one of those times after Jeff's insomnia acts up and he happens to make a snapshot. ;-) Target-specific bugs can be implemented, detected, and corrected in a shorter time frame than a previous snapshot cycle. When I totally go off on a rant and do something Really Bad to my local copy of the sources, it's just a matter of 'cvs status' to see which files we've hosed and a matter of nuking them and a 'cvs update' to refresh them. We have lots of choices and options available to us now. If you want to stay with the snapshot/patch model, I understand that'll still be available. I welcome the options that immediate EGCS repository access gives me. If EGCS was previously the bazaar, we now have the option of a traditional bazaar/flea market or a "speed" auction with very fast turnaround cycles and the ability to bid when we're not even in the room, but more when it's convenient for us. I like it. -- Robert Lipe http://www.dgii.com/people/robertl robertl@dgii.com