Hi Kristis,

I'll try and answer all your questions, let me know if I forget any,
sorry if the answers are a bit brash, they are not intended that way :-)

> 
> This is a known issue, and apparrently the Subversion team likes it
that
> way. I don't agree, but ...
> 
> http://bugzilla.mkgnu.net/show_bug.cgi?id=385

I have a feeling that will never change!

>
> I don't understand. If the bug is locked, shouldn't activity_verify
> fail ? And if activity_verify succeeds, shouldn't it also acquire the
> lock so that activity_commit won't fail (and will release the lock) ?
> 

activity_verify just reads data from Test Director, so it doesn't matter
if a use has it locked, as locked items can be read but not written to.
(For obvious reasons such as running reports shouldn't fail just because
a user is updating a Bugs details).  This is why an explicit check is
required to see if it is locked.

The second point you raise is about making activity_verify acquire a
lock, and activity_commit release the lock.  This will leave a big hole
where bugs can remain and get stuck in a locked state.

The Scenario: If a user commits some changes, the activity_verify locks
the bug and gives the go-ahead. Subversion then starts doing the commit
and fails the commit due to a file clash that needs updating.  In this
scenario the post-commit hook will not be called so the Bug will never
be unlocked.

> There are two issues here:
> 
> - In the Testdirector backend there's an incomplete implementation of
> this locking business.
> 

I'm not sure what other way I can close this whole other than what I
have suggested, ideas are very welcome.

> - In Scmbug, there's no mechanism of reporting errors on
> activity_commit.

I have tried to think of different ways to solve this, but the thing I
keep coming back to is sending the E-mail.

> I understand that you are providing a patch for the second one. That
had
> to happen at some point. But I'd rather the first issue was resolved
> too. And it sounds like it could be resolved with no implementation
and
> maintenance of an additional integration_* function (e.g.
> integration_check_bug_lock)

I hope I've done a reasonable job above of explaining the problem, I
didn't want to hijack any of the other interfaces for this as it will
end up giving rubbish error codes back which will not really explain the
problem to the user.

> [SNIP CODESTRIKER]
> I don't like the idea of a ScmbugUtils.pm. That's what scmbug-common
was
> created for. If we need to shift code around so that you don't
maintain
> a ScmbugUtils.pm I'd rather we did that.
> 

I agree in an ideal world this would link into a Utilities file within
Scmbug.  I originally wrote it over 6 months ago and just wanted to get
something up and running to demo to people here. Maybe at a later date
when David has had a chance to integrate what is there (I believe he has
just started looking into doing so re-integration of patches) we will be
able to just swap it over.

> > Also we use this information for merges as we do a reasonable amount
> > of changes on branches and do specific bug fixes for customers that
> > need merging into the latest product line.  (Although the merges are
> > pretty much by hand at the moment)
> 
> You should have a look at the Merger tool. It's close to being
finished:
> all the necessary information are retrieved from a bug-tracker. We
just
> need to issue the appropriate svn commands to merge.

I am actually hoping that we can use that tool when it is up and running
:-)
My only thing at the moment is allowing users to run it on remote
machines rather than the machine running Scmbug, but I'm sure I could
wrap it up in a simple web page to do it for them :-)

Hope this helps

Thanks

Rob

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

Reply via email to