[ On Thursday, February 17, 2000 at 16:18:42 (-0700), Tobias Weingartner wrote: ]
> Subject: Re: CVS File Locking
>
> Aiee, therein is the rub. (I make a horrible scott, but I try) You'd have
> to define "merge", for the above to make sense. I can merge any three files.
> Simply concatenate them, or XOR them all together. The output may not make
> much sense, but they are "merged" in some sense of the word.
>
> I'll agree with you though if you say that there may not be any sensible
> method to automatically merge certain file-formats and keep the syntactic
> and semantic format of the file "correct"...
As you correct yourself in the second paragraph, you have to define the
goals of *automatic* merging for the purposes of CVS.
You also have to define methods of regression testing that can be
applied (outside of CVS, of course) to ensure that the results of such
merges are correct (or at least acceptable). CVS doesn't require that
you do regression testing after ever merge (as Aegis does), but any form
of concurrent development using merging requires you to do so in order
to be truly effective.
Strict serialisation of the development process doesn't really eliminate
this need either -- it just purports to make it easier to achieve (one
change at a time).
For files that are unmergable automatically it's simply a matter of
manually resolving conflicts, just as one would have to do anyway when
either the automated merge detects conflicts, or when the regression
tests prove that the merge failed despite the lack of conflicts.
--
Greg A. Woods
+1 416 218-0098 VE3TCP <[EMAIL PROTECTED]> <robohack!woods>
Planix, Inc. <[EMAIL PROTECTED]>; Secrets of the Weird <[EMAIL PROTECTED]>
