#development
Title
w
witty-crayon-22786
11/14/2021, 4:53 AMa problem that i expect we’ll see a lot of with multiple resolves (both on the JVM and on Python), and proposal(s): https://docs.google.com/document/d/1BSws879Ii_bM0nu4ZMNyvz6YRtBvaFNj3MXpsvwGdi0/edit# … comments appreciated.
comment access using a pants-devel@ email address
cc @hundreds-father-404, @happy-kitchen-89482, @enough-analyst-54434
e
enough-analyst-54434
11/14/2021, 5:03 AMI guess I don't get why interior nodes compatible with multiple resolves don't just have that information used purely to fail a subgraph. IOW all resolve picking must be driven from a binary / test root and that driver determines the inference for the subgraph. If an interior node has an incompatible set of compatible resolves, that just means fail fast.
w
witty-crayon-22786
11/14/2021, 5:05 AMinference occurs before that
so the problem statement deals with the point at which a target is determining its dependencies
since we fail due to ambiguity currently, target definitions get much, much more verbose than in the unconflicted case.
e
enough-analyst-54434
11/14/2021, 5:06 AMThat's just how it works today, right? Could do local inference, validate resolve, then 3rd party.
w
witty-crayon-22786
11/14/2021, 5:07 AMnot sure i understand that, but yea: maybe. if you want to Suggest a third proposal at the bottom it would be welcome
i’ve added an example of what i mean
e
enough-analyst-54434
11/14/2021, 5:07 AMWill do.
It sounded way more complex than needed so stopped short.
Ok, added.