hundreds-father-404
07/16/2020, 9:26 PM!
? (https://github.com/pantsbuild/pants/pull/10385).
I think the motivating cases are:
1) Our dep inference logic has a flaw. This is an escape hatch for dep inference causing false positives.
2) Even if it’s not a false positive, a user for some reason doesn’t like having a dependency included, e.g. because it causes a cycle.
#2 is what I’m a little concerned about. Things like MyPy won’t work if you leave off a dependency. But I suppose this falls into “This is an advanced escape hatch 99% of users won’t need; trust the user when they say they don’t want something”?witty-crayon-22786
07/16/2020, 10:37 PMwitty-crayon-22786
07/16/2020, 10:38 PMhundreds-father-404
07/16/2020, 10:40 PM[':foo', '!//:foo']
. This is really only useful for dep inference.
Right now, we haven’t known of dep inference causing any false positives because it is overly conservative. So this is really for the second reason, but that second reason is better solved via cycle tolerancehappy-kitchen-89482
07/16/2020, 10:52 PMhundreds-father-404
07/16/2020, 10:52 PMhundreds-father-404
07/16/2020, 10:53 PMwitty-crayon-22786
07/16/2020, 10:55 PMwitty-crayon-22786
07/16/2020, 10:57 PMwitty-crayon-22786
07/16/2020, 10:57 PMhundreds-father-404
07/16/2020, 10:57 PMif PROD:
import really_expensive_thing
witty-crayon-22786
07/16/2020, 10:57 PMwitty-crayon-22786
07/16/2020, 10:58 PMwitty-crayon-22786
07/16/2020, 10:58 PMhundreds-father-404
07/16/2020, 10:59 PMwitty-crayon-22786
07/16/2020, 10:59 PMPEX_EXTRA_SYS_PATH
, for example.hundreds-father-404
07/16/2020, 11:01 PM