perforce

[prev] [thread] [next] [lurker] [Date index for 2003/12/30]

From: Nicholas Clark
Subject: perforce
Date: 20:11 on 30 Dec 2003
So perforce is wonderful and perfect and always gets it right and everything
that CVS and subversion etc are not?

Bollocks.

I've hit this one three times now.

You have development going on in one branch, and you're integrating changes
in the repository across to your stable branch. Perforce is good at handling
branching, so this should all work. Perforce has numbered changesets, and
you can take several changesets from one branch and integrate them as a
single changeset in your stable branch. This approach makes sense when it
took several tweaks in the development branch to get something stable. And
nearly all the time it just works, and you believe what it says

But perforce SILENTLY FAILS to work when you attempt to integrate two
changes, where the first adds a file, and the second edits that file.
The edit isn't integrated. Why on earth not? This is actually a depressingly
common scenario - someone adds a file as part of a change, they discover
some problem they didn't expect, so next thing they do is edit their new
file. And the person doing the integration waits until all the dust settles
down, then attempts to integrate all these "changes" together. And normally
it works. But if the add/edit combination is present then Perforce fails.
Silently.


Bah.

And if you still think Perforce is perfect, why do they write a whole
FAQ about "how to back out a change"

  http://www.perforce.com/perforce/technotes/note014.html

rather than just fixing their software to make it possible?

Nicholas Clark

Generated at 19:01 on 27 Nov 2006 by mariachi 0.52