On Thu, 2006-12-07 at 23:07 +0100, Anna / Marek Latuskiewicz wrote:
> >> - the feature around label_name - is there any specific reason that 
> >> scmbuga
> >> assumes the directory must be in root - ie. /branch /tags, not
> >> /project_name/branch & /project_name/tags
> >
> > If scmbug assumes, somewhere, the directory must be in root, it's
> > probably a limitation of some original design; and should be lifted. It
> > seems that SCM_Functions.pm.in:get_label_date suffers from this (used by
> > the VDD Generator). Have you found other places ?
> >
> > We should fix this, yes.
> OK - but it can be fixed in config files - All you have to do is to play 
> with regex:
> 
>  label_directories => [
>          'tags',
>          'branches'
>          ],
> 
> 
> 
> ex. you just need to add -all tags in the second level
> 
>          '[^/]*/tags
> The pros is that you have the complete control of where you are looking for 
> so there is no constraints to for examples 'tags' subdirectores deep within 
> the project.

This breaks automatic product name detection.


> 
> You need also to change naming rules - for ex.
> 
>         '^.+?_RELEASE_[0-9]+-[0-9]+-[0-9]+$',
> 
> 
> 
> to:
> 
>         '^.*?_?RELEASE_[0-9]+-[0-9]+-[0-9]+$',
> 
> 
> 
> as there is no need to name build within the directory.
> 
> Therefore we need to change default config templates only....
> 
> 
> 
> >
> > http://bugzilla.mkgnu.net/show_bug.cgi?id=893
> >
> > For example, we should require the user to supply in both the VDD
> > Generator and the Merger a label name like "project_name/tags/sometag"
> > instead of just "sometag". So while you are implementing the Merger,
> > assume the bigger label name, and I'll take care of fixing the VDD
> > generator.
> >
> > How does this sound ?
> OK - how about proposed solution in config file? Looks fine for me...

I believe it comes down to the fact that we want to have the capability
of automatically detecting "product_name1" and "product_name2" in paths
like:

product_name1/trunk
product_name1/branches
product_name2/trunk
product_name2/branches
/trunk/product_name1
/trunk/product_name2
/branches/product_name1
/branches/product_name2


> >> - what do you think of declining any updates in tags by scmbug?
> >
> > That's a great idea. It was recommended in the past:
> >
> > http://bugzilla.mkgnu.net/show_bug.cgi?id=859#c4
> >
> > When 859 is implemented you would be able to somehow mark tags as
> > "closed/frozen" branches in which new commits would be disallowed.
> 
> It proposes a stricter control I thought - the problem I see is that we 
> would duplicate some of the functionality of 'auth' file within subversion. 

I don't know how that file works. But cases (2),(3),(4) would still have
to be implemented.

> Don't you think? Do we have an infrastructure to analyze this file from 
> scmbug?

I doubt this file has the notion of "it's required to supply a bug id".
Or any other file in any other SCM system. It seems that we need to
implement this regardless.

> I reckoned to implement only something like this in the configuration :
> 
>  frozen_directories => [
>          'tags'
>          ],
> 
> + small changes in perl scripts

It could work. It could also work, as David pointed out, if (4) was a
special case of (3) in the link above. I feel it would be simpler as
David proposed, since we still have to do the work for (2),(3),(4).


_______________________________________________
scmbug-users mailing list
[email protected]
http://lists.mkgnu.net/cgi-bin/mailman/listinfo/scmbug-users

Reply via email to