From: Martin Langhoff Date: Sun, 17 Jun 2007 21:27:17 GMT Subject: Re: [PATCH] cvsserver: fix legacy cvs client and branch rev issues Message-ID: <46a038f90706171427g43c4ccf2vdf1172a962481964@mail.gmail.com> In-Reply-To: <20070617103744.GE1828@planck.djpig.de> On 6/17/07, Frank Lichtenheld wrote: > On Sun, Jun 17, 2007 at 10:10:51AM +0100, Dirk Koopman wrote: > > Frank Lichtenheld wrote: > > >On Sat, Jun 16, 2007 at 07:50:06PM +0100, Dirk Koopman wrote: > > >Hmm, I don't see how you could have a problem with that since cvsserver > > >doesn't support branches and never generates any revision numbers in > > >that format? > > > > > >There is probably much more code out there in cvsserver that does assume > > >that revision is always a simple integer. > > Let me rephrase that (after actually looking through the code): > All of the revision handling code assumes that. Exactly. cvsserver emulates CVS on a single HEAD, that's why you use the headname as the 'module' parameter you pass to CVS when doing a checkout. ... > Hmm, so you did the cvs update in an old working copy of the original > CVS repository? Then CVS sent those version numbers from the CVS/Entries > file to the server, cvsserver certainly never generates numbers like > that. And I would be very suprised if you could do anything remotely > useful with abusing the old working copy this way... Agreed - that's not really supported. Now, I'd _love_ to have a bit of time to implement CVS-style branch support to cvsserver (so a check for valid version numbers that have more dots would be a good thing), but it's hard hard hard, specially because there are many ambiguities to resolve. It would enormously useful to have branch support together with support for a bit of "version skew" so that you can replace a real CVS server with cvsserver and have people continue using the old cvs checkouts -- because the file versions and branches match. As things stand, I want to say thanks to Frank for giving cvsserver some love :-) cheers, martin