#general
- g
gorgeous-fish-11040
10/22/2015, 6:54 PM@witty-crayon-22786: maybe me 😉 - w
witty-crayon-22786
10/22/2015, 6:54 PM😉 - w
witty-crayon-22786
10/22/2015, 7:31 PM@bored-art-40741, @happy-kitchen-89482 : i'm going to look at implementing the 'optional direct deps only' mode for jvm compile - h
happy-kitchen-89482
10/22/2015, 7:31 PMmeaning? - w
witty-crayon-22786
10/22/2015, 7:31 PMsorry, talked with patrick about this earlier - w
witty-crayon-22786
10/22/2015, 7:32 PMoptionally only providing the direct deps to a target at build time - w
witty-crayon-22786
10/22/2015, 7:32 PMjava compilation should work with only direct, for example - h
happy-kitchen-89482
10/22/2015, 7:32 PMAh - h
happy-kitchen-89482
10/22/2015, 7:32 PMScala will not - w
witty-crayon-22786
10/22/2015, 7:33 PMscala might work with "direct"... possibly more than is explicitly listed in the file, but still not the complete transitive path - h
happy-kitchen-89482
10/22/2015, 7:33 PMI mean, we could hypothetically require users to explicitly specify their implicit deps, but that would be onerous - w
witty-crayon-22786
10/22/2015, 7:33 PMyea. hence, optional - h
happy-kitchen-89482
10/22/2015, 7:33 PMtype inference could bring in some type you know nothing about - w
witty-crayon-22786
10/22/2015, 7:33 PMyes, but stably. - h
happy-kitchen-89482
10/22/2015, 7:33 PMand shouldn't need to know about - w
witty-crayon-22786
10/22/2015, 7:34 PMin any case: i'm wondering whether this is bigger than jvm_compile - w
witty-crayon-22786
10/22/2015, 7:34 PMis this something that should be included native in target invalidation? - w
witty-crayon-22786
10/22/2015, 7:34 PMnatively* - h
happy-kitchen-89482
10/22/2015, 7:35 PMmeaning that your deps deps don't invalidate you? - w
witty-crayon-22786
10/22/2015, 7:35 PMif you've turned it on for your target, correct. - w
witty-crayon-22786
10/22/2015, 7:36 PMshould maybe
have a different mode?self.invalidated(invalidate_dependents) - h
happy-kitchen-89482
10/22/2015, 7:36 PMwhat's the need for this? - h
happy-kitchen-89482
10/22/2015, 7:36 PMWhat do we gain in return for the added complexity and the subtle invalidation bugs? - w
witty-crayon-22786
10/22/2015, 7:36 PM2500 module classpaths at compile time make for horrific IO - w
witty-crayon-22786
10/22/2015, 7:36 PMnot scalable. - w
witty-crayon-22786
10/22/2015, 7:37 PMeven with jars - w
witty-crayon-22786
10/22/2015, 7:38 PMbut since this is an experiment in practicality anyway, perhaps waiting to push it down into the
framework would make senseinvalidation - h
happy-kitchen-89482
10/22/2015, 7:39 PMsure, but that's separable from the invalidation logic - h
happy-kitchen-89482
10/22/2015, 7:39 PMwith
you can have onlyA->B->C
onB
compile time classpath, but still invalidateA
(and, obviouslyA
) ifB
is modified.C - h
happy-kitchen-89482
10/22/2015, 7:40 PMThe classpath thing seems order-of-magnitude easier to do