Strategies for push/merge problem?

[Hans Meine]

Am Dienstag, 15. Juli 2008 14:31:30 schrieb Adrian Buehlmann:
> Are you talking about deliberately committing states of the tree
> that fail the testsuite? Committing such states is bad.

I don't think so (anymore).  With a DVCS, I think committing (as opposed to 
pushing) such changes is good, since it allows you to tag your single steps 
with a commit message and increases the readability of "hg diff" (since it 
shows only related, small changes).

AFAICS, this is one of the advantages of a DVCS, that it encourages you to 
commit often, since nobody will suffer from such commits, as long as you 
don't publish the changes yet.

> For example, if you do refactorings, passing the testsuite is
> the invariant.

Of course, you're right that small steps that do *not* break things are 
preferable. ;-)  For things like "hg bisect", it could be a feature if one 
could tag changesets as "intentionally broken" / "internal commit" to be 
skipped.

-- 
Ciao, /  /
     /--/
    /  / ANS