cvs vs svn reloaded

[Patrys :: Patryk Zawadzki]

Dnia 08-10-2005, sob o godzinie 20:57 +0200, Paweł Sakowski napisał(a):
> On Sat, 2005-10-08 at 20:41 +0200, Patrys :: Patryk Zawadzki wrote:
> > > No, but the good thing is that nothing can get silently overwritten or
> > > sth -- I get a nice commit conflict. A superuser resolving those
> > > manually should be enough.
> > 
> > Do we need superuser intervention in such cases?
> 
> We do. In that case SPECS/trunk/foo.spec and foo/trunk/foo.spec have
> different contents, which inevitable leads to having a mess. Which is
> why I detect such cases in pre-commit and refuse any other commit to
> foo.spec until the problem is resolved.

Then if some people commit using trunk and some using general SPECS,
such locks might become common. Maybe we should ignore changes made to
SPECS and always sync changes from proper packages? It's not that common
that someone needs to commit hundreds of spec files in one atomic
operation. Even then he could probably cd $svn_root and do a full svn ci
or a commit on a set of specified directories.

> Oh, by superuser I don't mean geteuid()==0, I mean somebody who is
> permitted to commits changes which aren't processed by
> 'mirror' (function trust_change defines who can do that).

That is clear to me, requiring uid 0 would be pretty awkward.

-- 
Patrys :: Patryk Zawadzki <patrys at pld-linux.org>
PLD Linux
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: To jest =?iso-8859-2?Q?cz=EA=B6=E6?= listu podpisana
	cyfrowo
Url : /mailman/pipermail/pld-devel-en/attachments/20051008/e7321e0a/attachment-0002.bin