#general
- a
ancient-stone-50795
03/07/2016, 9:08 PMwe == Square - g
gentle-potato-32845
03/07/2016, 9:11 PMoh boy. i think i did not know about this - g
gentle-potato-32845
03/07/2016, 9:11 PMi am Learning A Lot - g
gentle-potato-32845
03/07/2016, 9:18 PMim not sure i understand how this is meant to work - g
gentle-potato-32845
03/07/2016, 9:21 PMoh i think i get it - g
gentle-potato-32845
03/07/2016, 9:21 PMwoah this is crazy - g
gentle-potato-32845
03/07/2016, 9:21 PMlol - a
ancient-stone-50795
03/07/2016, 9:21 PMWelcome to Ivy! - g
gentle-potato-32845
03/07/2016, 9:22 PMso the idea is that you use managed_jar_dependencies to pin a version of library B that is used by library A that is a direct dependency of my project…? - g
gentle-potato-32845
03/07/2016, 9:23 PMand then managed_jar_libraries when i might need to depend on library B directly in some other target and i dont want to define a another jar_library for it - g
gentle-potato-32845
03/07/2016, 9:23 PMright? - w
witty-crayon-22786
03/07/2016, 9:33 PMmanaged_jar_libraries are optional... twitter isn't using them (yet) - w
witty-crayon-22786
03/07/2016, 9:35 PM@jolly-chef-92794: are there docs for managed dependencies usage? not finding them - a
ancient-stone-50795
03/07/2016, 10:15 PMThe only thing I see is in the BUILD dictionary and some examples under testprojects - a
ancient-stone-50795
03/07/2016, 10:15 PMtestprojects/src/java/org/pantsbuild/testproject/depman/ - g
gentle-potato-32845
03/07/2016, 10:21 PMoh hm. i may have misunderstood how using the :: syntax works - g
gentle-potato-32845
03/07/2016, 10:21 PMor something - g
gentle-potato-32845
03/07/2016, 10:22 PMso. i am in this dependency hell where things cannot agree on one version of guava to use - g
gentle-potato-32845
03/07/2016, 10:23 PMi managed to get BUILD files that depend on different versions, but when i run the tests using a command that uses the :: syntax to pick both them up - g
gentle-potato-32845
03/07/2016, 10:23 PMonly one version is used across all tests…? - g
gentle-potato-32845
03/07/2016, 10:24 PMi am having trouble describing what i am doing and i think because i might be doing something very wrong... - b
bored-art-40741
03/07/2016, 10:25 PM
just means "select all targets under this directory":: - b
bored-art-40741
03/07/2016, 10:26 PMAnd more precisely we mean "select those targets as the 'target roots'" - b
bored-art-40741
03/07/2016, 10:26 PMMost actual operations, including resolves, run over the closure of the target roots - b
bored-art-40741
03/07/2016, 10:27 PMAs a result, a single ivy resolve is running over the full closure of whatever
target roots, and a single resolve produces a single classpath--ivy uses whatever eviction strategy is configured to pick versionsfoo/bar:: - b
bored-art-40741
03/07/2016, 10:28 PMBy default I think it's newer version wins - g
gentle-potato-32845
03/07/2016, 10:28 PMyeah - g
gentle-potato-32845
03/07/2016, 10:28 PMi think i am doing something Bad then. because i basically end up with tests that pass when running a specific target but when running certain groups of targets :< - b
bored-art-40741
03/07/2016, 10:29 PMWhat you are experiencing here is the ivy resolution hell that motivated Foursquare to abandon Ivy altogether and enforce a globally consistent classpath. That's not feasible for most large codebases. You could do something more tricky with shading to split up the different versions into different packages - b
bored-art-40741
03/07/2016, 10:29 PMTwitter definitely has lots of experience with this problem too, and they do not have 4sq's globally consistent classpath