pantsbuild #general

@bored-art-40741: patch https://rbcommons.com/s/twitter/r/2827/ ... i think we might want to not strongly hold analysis at all, but i need to look in sbt to see whether it does (which would make that moot)

freezing-motorcycle-14222

09/15/2015, 11:18 PM

@witty-crayon-22786: I am aware of @mammoth-pharmacist-54215's work. I was just helping by eliminating some obvious bugs I found myself while trying to apply

pylint

and documenting how they arose

witty-crayon-22786

09/15/2015, 11:18 PM

@freezing-motorcycle-14222: the model of "turning on one entire check at once" is a bit more tractable (edit: tractable/efficient)

freezing-motorcycle-14222

09/15/2015, 11:19 PM

Well, I originally wanted to turn on

pylint

in its entirety for certain subdirectories that were

pylint

clean, but it seems like

pycheck

doesn't support that?

witty-crayon-22786

09/15/2015, 11:20 PM

@freezing-motorcycle-14222: pants implements a blacklist for targets/files

witty-crayon-22786

09/15/2015, 11:20 PM

...i forget which.

freezing-motorcycle-14222

09/15/2015, 11:20 PM

Well, then I'll just submit one more local fix and then move on to larger project-wide fixes, then

freezing-motorcycle-14222

09/15/2015, 11:23 PM

Also, in the process of this I noticed that the

scalastyle_integration

test is not included in the global

:integration

target. Is there any reason for that or should I just enable it?

freezing-motorcycle-14222

09/15/2015, 11:48 PM

It seems like

pants

does not support

required

args? Is the rationale behind that they should be encoded as positional arguments instead?

bored-art-40741

09/15/2015, 11:55 PM

Our arg parsing is its own beast on top of argparse

bored-art-40741

09/15/2015, 11:55 PM

pants itself has no required args, and in general if a task truly requires an arg, it's expected to do validation when it's actually executed

bored-art-40741

09/15/2015, 11:55 PM

This is largely because a task has no guarantee, even after it's initialized and several of its lifecycle methods are run, that it is actually going to be run

bored-art-40741

09/15/2015, 11:56 PM

It only knows when

execute()

is hit, at which point more argument validation happens.

bored-art-40741

09/15/2015, 11:56 PM

It's plausible that we could provide some tooling to help ease this validation, but it's very difficult to fit the notion of required arguments into the pants model in a way that's totally opaque to the task author

witty-crayon-22786

09/15/2015, 11:57 PM

@freezing-motorcycle-14222: what type of argument is this?

freezing-motorcycle-14222

09/15/2015, 11:58 PM

Oh, this came up in the context of dep-usage. I noticed that the output file is mandatory, but the validation step comes a bit late

freezing-motorcycle-14222

09/15/2015, 11:58 PM

I was just trying to figure out a way to move the validation up earlier so that it would fail fast before trying out all the tasks

happy-kitchen-89482

09/15/2015, 11:59 PM

we use argparse as an implementation detail, not all its features are supported, by design.

freezing-motorcycle-14222

09/15/2015, 11:59 PM

Yeah, I noticed in the parser.py file that you intentionally don't support

required

so I figured it was by choice

rough-minister-58256

09/16/2015, 12:01 AM

https://github.com/pantsbuild/pants/blob/master/src/python/pants/backend/jvm/tasks/jar_publish.py#L562 <— anyone know of a good reason (or other examples, if they exist) for doing interactive user input in a task?

rough-minister-58256

09/16/2015, 12:02 AM

seems like maybe we should excise this?

bored-art-40741

09/16/2015, 12:04 AM

jar publish is especially dangerous

bored-art-40741

09/16/2015, 12:05 AM

I don't think there's anything fundamentally wrong with user input in a task

rough-minister-58256

09/16/2015, 12:07 AM

rough-minister-58256

09/16/2015, 12:08 AM

tho it does seem like allow user input in tasks would strongly affect the idea of a task running remotely (i.e. post engine-refactor)

rough-minister-58256

09/16/2015, 12:09 AM

s/allow/allowing