ant-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Nicolas Lalevée <>
Subject Re: IVY-1455: effects of override tag escaping definition scope
Date Thu, 26 Dec 2013 23:03:06 GMT

Le 24 déc. 2013 à 01:12, Charles Duffy <> a écrit :

> Howdy, all --
> I have a reproducer for an issue similar to IVY-1333 ("impossible to get
> artifacts when data has not been loaded") which demonstrates some
> surprising behavior, and would appreciate feedback from folks more familiar
> with the ResolveEngine's internals:
> An <override> tag in a transitive dependency is impacting the overall
> resolution, even when no child of that dependency matches the pattern
> described by the tag in question.
> The tree looks as follows:
> reproducer/top-level
> |- conflict/conflict==1 [conf="with-direct-dep->default"]
> |- reproducer/phase-one==1 [conf="master->*"]
> |  |- conflict/conflict==2
> |  |- reproducer/phase-two==1
> |  |  |- empty-module/empty-module==1
> |  |  |- [override:conflict/conflict==1]
> Because phase-two has no children with a dependency on conflict/conflict
> (its only child being empty-module), it is entirely surprising to me that
> its override tag has any effect whatsoever. (However, removing the
> dependency on empty-module *also* circumvents the bug).
> The top-level reproducer has two confs descended from master,
> "with-direct-dep" and "without-direct-dep". In both, conflict/conflict is
> resolved to version 2, but in the latter, conflict/conflict==1 is not
> loaded at all. Only when both of these confs are invoked within a single
> ivy:resolve task does the issue in question trigger.
> Could I ask for some pointers in the right direction to get a better handle
> on what might be going on here?
> The reproducer in question is attached to IVY-1455 as a zip file.

That seems to be a hell of a test case, you did a wonderful job getting that reproducible.

But I don't know much that "override" feature. I would need to actively debug it to understand
what the bug is about. So I wouldn't be much of a help to you now, maybe later (completely
unpredictable later).


To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message