<@U06A03HV1> OK, so current behaviour in the cours...
# development
a
@witty-crayon-22786 OK, so current behaviour in the coursier work is to select a resolve per
coarsened_target
in java/scala checks. Were there any further changes needed?
f
Select a resolve for an entire compilation graph, right?
w
no, that sounds good. i’m sure that there will be further work around fixing up root selection (in particular: things like
test
should select a resolve per test target, etc), but that is fine for further changes (since it’s not actually changing in this PR).
f
So deploy_jar could be built from different resolves currently?
a
@fast-nail-55400 nope, selecting for the entire compilation graph doesn’t do the right thing here — it was trying to select a common resolve for
exampleapp
and
pants_java_parser
, which is not right
w
@fast-nail-55400: no: the
package
goal will just currently fail if it can’t find a single resolve. but that is not a change from master, and is something that we can fix in followups
a
No.
deploy_jar
works correctly (see
classpath.py
), it’s just
javac_check
and
scalac_check
that need to select per-
coarsened_target
w
(both
test
and
package
should be selecting a resolve per root)
f
pants_java_parser should not be part of the output of a deploy_jar so it would be a different resolve
ah this is for check, got it
👍 1
a
Right. That’s where the confusion came from.
w
🚢
@ancient-vegetable-10556: adjusting the behavior for
package
and
test
might make a good fast follow: need to request a different
Classpath
per root, if we aren’t already
…oh. we are. yay.
nevermind then 😃
a
Yup, you got that right 🙂