CVS - both branch and revision

[clown]

This

cvs up -dP -r [branch] -D [timestamp] [simple_dir]

doesn't work for me absolutely. In CVS FAQ i saw some (undocumented?) trick for such a task:

cvs up -dP -r [branch]:[timestamp] [simple_dir]

commented by audience as "dreaming" because it doesn't work either.

The 'alwaysUseLatest = True' does a little help since it affects only when there is no autodetected source revision:

        # if revision is None, use the latest sources (-rHEAD)
        revision = s.revision
        if not revision and not self.alwaysUseLatest:
            revision = self.computeSourceRevision(s.changes)

But I can hardly imagine such circumstances.

I made a quick workaround for this CVS bug but cannot be sure is this a best way.

--- /usr/share/pyshared/buildbot/slave/commands.py~     2008-07-25 03:36:25.000000000 +0400
+++ /usr/share/pyshared/buildbot/slave/commands.py      2009-02-19 17:51:06.000000000 +0300
@@ -1472,7 +1472,7 @@
         command = [self.vcexe, '-z3'] + self.global_options + ['update', '-dP']
         if self.branch:
             command += ['-r', self.branch]
-        if self.revision:
+        elif self.revision:
             command += ['-D', self.revision]
         c = ShellCommand(self.builder, command, d,
                          sendRC=False, timeout=self.timeout)
@@ -1490,7 +1490,7 @@
                    [verb, '-d', self.srcdir])
         if self.branch:
             command += ['-r', self.branch]
-        if self.revision:
+        elif self.revision:
             command += ['-D', self.revision]
         command += [self.cvsmodule]
         c = ShellCommand(self.builder, command, d,

[clown]

Replying to clown:

This

cvs up -dP -r [branch] -D [timestamp] [simple_dir]

doesn't work for me absolutely.

I made quick workaround in master.cfg:

class MyCVS( CVS ):
	def startVC( self, branch, revision, patch ):
		CVS.startVC( self, branch, None, patch )

[dustin]

So the issue is that CVS can't do both a revision and branch at the same time? If this is the case, we shouldn't silently drop one or the other (as both patches do), but fail in this circumstance.

[clown]

Replying to dustin:

So the issue is that CVS can't do both a revision and branch at the same time? If this is the case, we shouldn't silently drop one or the other (as both patches do), but fail in this circumstance.

This cannot be adequate if somebody uses a standard commit-AnyBranchScheduler?-builder chain: any CVS commit to a non-default branch would lead to a failed CVS step in build.

[dustin]

So maybe the CVS Step needs to be constructed with an argument that says to pay attention to the branch *or* the revision.

The alternative -- always checking out the latest code (by ignoring the revision) might be OK, I suppose, as long as it's documented. I would want to poll the mailing list on that one first.

[dustin]

clown - please do ask on the mailing list to see how this should be solved.

[dustin]

moving out to the next release

[dustin]

Let's see if we can tackle this in 0.8.1

[dustin]

I haven't seen any action here, and no complaints from CVS users. Is this still a problem? Reopen if so.