Why not delay the circular dependency check untill the resolve is finished?
Maybe the module causing the circular dependency gets evicted so there might be
no problem at all?
Maarten
----- Original Message ----
From: Xavier Hanin (JIRA) <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED]
Sent: Friday, December 7, 2007 9:22:43 PM
Subject: [jira] Resolved: (IVY-663) latest compatible conflict manager fails
with circular dependencies and dynamic revision
I've just committed a fix for this, but it slightly changes the way
circular dependencies are handled, we may have to discuss that further.
The problem is that for the moment we consider we have a circular
dependency as soon as we find the same module id twice in the visit path. It
means this happens before any conflict resolution is done, therefore
letting no chance to a conflict manager to deal with the problem. I
haven't actually changed this yet, I've only delayed the moment when
isCircular is checked by the ResolveEngine to know if it has to continue the
graph visit. But I'm wondering if we shouldn't delay the whole circular
dependency check (including the call to CircularDependencyStrategy) to
the moment we have twice the same ModuleRevisionId in the path, and
not only ModuleId. We should still handle the case of the module
currently resolved specifically, but I think it makes sense to say that
#A;1->#B;1->#A;2 is not a circular dependency (this can only happen with a
conflict manager accepting multiple revisions of the same module).
Thoughts?
____________________________________________________________________________________
Looking for last minute shopping deals?
Find them fast with Yahoo! Search.
http://tools.search.yahoo.com/newsearch/category.php?category=shopping
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
