#general
- e
enough-analyst-54434
09/21/2015, 9:05 PMin all cases - w
witty-crayon-22786
09/21/2015, 9:06 PMsimilar to the existence of downstream language targets affecting the identity of a thrift_library? - w
witty-crayon-22786
09/21/2015, 9:06 PMor, its outputs at least - e
enough-analyst-54434
09/21/2015, 9:07 PMI guess - yes. Just that interior node may depend on guava 18 and when resolved in a larger group 19 wins, so when you ask ClasspathProducts for the interior node's cp - you get guava 19 - e
enough-analyst-54434
09/21/2015, 9:07 PMwhich is different than if you resolved it alone - w
witty-crayon-22786
09/21/2015, 9:07 PMbecause in this context, that's correct - e
enough-analyst-54434
09/21/2015, 9:07 PMThis could/should/better be fine - e
enough-analyst-54434
09/21/2015, 9:07 PMsince all the code runs together - w
witty-crayon-22786
09/21/2015, 9:08 PMotherwise we'd be compiling it against one thing, and then running it with something else - e
enough-analyst-54434
09/21/2015, 9:08 PMbut its different from maven - b
bored-art-40741
09/21/2015, 9:08 PMIt's not fine in many situations - e
enough-analyst-54434
09/21/2015, 9:08 PMAnd at least some of Nicks design doc was on squaring expectations with maven - b
bored-art-40741
09/21/2015, 9:08 PMWe ran into this (with guava specifically) - b
bored-art-40741
09/21/2015, 9:08 PMIt's only fine if you can truly pin versions at the specification level - w
witty-crayon-22786
09/21/2015, 9:08 PM@bored-art-40741: not really: see nick's patch - b
bored-art-40741
09/21/2015, 9:09 PMSorry, not really what? - w
witty-crayon-22786
09/21/2015, 9:09 PMit's fine if you recompile the inner node, and it has a different identity depending on context - w
witty-crayon-22786
09/21/2015, 9:09 PMit's not really necessary to pin everything - b
bored-art-40741
09/21/2015, 9:09 PMIt's necessary to be able to - w
witty-crayon-22786
09/21/2015, 9:09 PMbut this was a bug - b
bored-art-40741
09/21/2015, 9:09 PMAnd there are issues that aren't caught at compile time - w
witty-crayon-22786
09/21/2015, 9:09 PMyes, that is true - h
happy-kitchen-89482
09/21/2015, 9:10 PM@bored-art-40741: did you try looking at the timing stats in the pants server? that includes timings for background work. - b
bored-art-40741
09/21/2015, 9:10 PMI hadn't - b
bored-art-40741
09/21/2015, 9:10 PMI'm trying to do a laptop local repro now - b
bored-art-40741
09/21/2015, 9:10 PMI'm still downloading - b
bored-art-40741
09/21/2015, 9:10 PMI'm on Indiana internet, precisely capped at 250 KB/s - b
bored-art-40741
09/21/2015, 9:10 PMMy download is going on an hour now - b
bored-art-40741
09/21/2015, 9:11 PMWhen it finishes I will confirm the repro with local-only - b
bored-art-40741
09/21/2015, 9:12 PMI'm beginning to suspect that doubling the artifact size more than doubles the cost