#general
- f
fancy-queen-20734
09/21/2015, 8:58 PMyeah - e
enough-analyst-54434
09/21/2015, 8:59 PMI ruly do think mirroring the local dep graph to N ivy files and resolving each leaf to roots is the only way to go - f
fancy-queen-20734
09/21/2015, 8:59 PMI could throw a run in there, and was planning on it, but it’s not really necessary to test for the problem - e
enough-analyst-54434
09/21/2015, 8:59 PMNope - I'm good on understanding this now, thanks - w
witty-crayon-22786
09/21/2015, 9:00 PM@enough-analyst-54434: hm... how does having a full ivy graph help in this case? - e
enough-analyst-54434
09/21/2015, 9:00 PMwell, you need to resolve each node in the local library graph - w
witty-crayon-22786
09/21/2015, 9:00 PM...because you compile inner nodes with what they would have resolved to? - e
enough-analyst-54434
09/21/2015, 9:00 PMthat requires an ivy file per each - e
enough-analyst-54434
09/21/2015, 9:00 PMsimply, resolve is per-target like compile - e
enough-analyst-54434
09/21/2015, 9:01 PMso each target needs an ivy - w
witty-crayon-22786
09/21/2015, 9:01 PMwe had a long discussion about this back when we were looking into isolation - w
witty-crayon-22786
09/21/2015, 9:01 PMand i think where we came out on it was that batched resolves are desirable - e
enough-analyst-54434
09/21/2015, 9:01 PMOnly for speed - e
enough-analyst-54434
09/21/2015, 9:01 PMnot correctness - w
witty-crayon-22786
09/21/2015, 9:01 PMyea - e
enough-analyst-54434
09/21/2015, 9:01 PMWhich is a funny stance! - w
witty-crayon-22786
09/21/2015, 9:01 PMwell, they're fine for correctness - w
witty-crayon-22786
09/21/2015, 9:01 PMif you hash things. - w
witty-crayon-22786
09/21/2015, 9:01 PMcf, patch - e
enough-analyst-54434
09/21/2015, 9:02 PMI don't think it can be, you have to guess the resolver's solutions - w
witty-crayon-22786
09/21/2015, 9:02 PMhonestly, i always thought that the fact that ivy was hashed across the entire set was enough to invalidate for this case - w
witty-crayon-22786
09/21/2015, 9:02 PM@enough-analyst-54434: no: just its outputs - w
witty-crayon-22786
09/21/2015, 9:02 PM...oh. i see. - w
witty-crayon-22786
09/21/2015, 9:02 PMer - e
enough-analyst-54434
09/21/2015, 9:02 PMI do think we have to resolve each target alone to be correct - w
witty-crayon-22786
09/21/2015, 9:03 PMthis isn't the only place where we hash the output of one task as the input to another task? - w
witty-crayon-22786
09/21/2015, 9:04 PM@enough-analyst-54434: correct in the sense that the compile time classpath matches the runtime classpath? - e
enough-analyst-54434
09/21/2015, 9:04 PMcorrect in the sense that today you can do a resolve - 1 resolve - e
enough-analyst-54434
09/21/2015, 9:04 PMand then as ClasspathProducts for the classpath of an interior node - e
enough-analyst-54434
09/21/2015, 9:05 PMThat classpath wil != the interior node resolved on its own