[cvsnt] Re: vendor branch and import problem

[Oliver Giesen]

Tony Hoyle wrote:

>>This was when you accidentally mix up vendor branch and release tag on 
>>import. This does not seem to be what is happening here, though...
> 
> No, but the cause of it was the lack of checking for an existing
> vendor tag. 

No, the cause was the lack of checking whether the given release tag 
already existed (doesn't matter if it's as a branch or regular tag).


 > Import still isn't particularly error checked but it's
> harder to stuff things up now (I just realised the release tags aren't
> checked either).

This alone should have sufficed AFAICT. Of course, then there's the 
other issue with Import apparently being hardcoded to always use the 
1.1.1 branch regardless of the given vendor branch name... Therefore you 
/must/ check the given vendor branch name indeed - and only allow it if 
it either *matches* the name of the existing branch specified by the -b 
option (implying -b1.1.1) or if there is no branch off the 1.1 revision 
at all yet! A better and more consistent (and less complex) solution IMO 
would be if using a non-existing vendor branch tag actually caused the 
creation of a new branch (I'd suggest using odd numbers as the third 
figure, e.g. 1.1.3, 1.1.5, etc.).


Sorry for dragging this on, but I really do think that disallowing 
reusing of vendor branch names was an unnecessary change at best. In the 
light of Phil Richards' recent observations it even appears to me that 
it is actually making matters worse than before...

Cheers,

-- 
Oliver
----	------------------
JID:	ogiesen at jabber.org
ICQ:	18777742	(http://wwp.icq.com/18777742)