pantsbuild #general

Channels

welcome

github-notifications

announce

rough-minister-58256

09/08/2017, 11:27 PM

that might lead to unexpected user local behavior?

rough-minister-58256

09/08/2017, 11:27 PM

also, fmt should never be cached or run in CI?

rough-minister-58256

09/08/2017, 11:34 PM

afaik, the strategy for consuming lint + fmt was always predicated on the idea of an integration with

arc

rough-minister-58256

09/08/2017, 11:34 PM

so when you

arc diff

to post a PR, it’d run

arc lint

which would run

./pants lint

- and then if that failed

arc

would automatically run

./pants fmt

for you to attempt to fix the lint failure

rough-minister-58256

09/08/2017, 11:35 PM

which apparently is an inbuilt feature of

arc

rough-minister-58256

09/08/2017, 11:36 PM

so the idea between the split initially was to enable that workflow, where we’d expect most users to be running both

lint

and

fmt

at the time they actually go to post the review vs while they’re running tests.

rough-minister-58256

09/08/2017, 11:36 PM

(and ofc assuming a baseline passing

lint

state in master)

fancy-queen-20734

09/09/2017, 12:04 AM

the issue with fmt as it is now is that fmt will always run on the full closure of targets when run locally, unless fmt runs are put in buildcache

rough-minister-58256

09/09/2017, 12:30 AM

yeah, that’s #4838 tho afaict

rough-minister-58256

09/09/2017, 12:31 AM

then the

arc lint

impl would run e.g.

./pants --changed-parent=X --changed-include-dependees-transitive lint

(and

fmt

) to net transitivity for a given change

rough-minister-58256

09/09/2017, 12:32 AM

fmt runs can’t be put in buildcache afaik because they actively mutate source code in the build root

rough-minister-58256

09/09/2017, 12:32 AM

as opposed to having a cacheable product

rough-minister-58256

09/09/2017, 12:33 AM

unless we had a better design for that, but imo fmt is sort of a special beast

rough-minister-58256

09/09/2017, 12:37 AM

and not sure i see the linkage between transitivity and “unless fmt runs are put in buildcache”

fancy-queen-20734

09/09/2017, 12:57 AM

oh, right. So, if fmt tasks are transitive, they will always be run on each applicable target in the full closure. If you have a target with 2000 dependencies, that might make fmt quite slow

rough-minister-58256

09/09/2017, 12:59 AM

afaik most autofmtr runtimes I’ve seen are on the order of seconds, but in that workflow

fmt

wouldnt get invoked unless

lint

failed and that should be cacheable

rough-minister-58256

09/09/2017, 1:00 AM

maybe there’d be some way in the

arc

logic to only run

fmt

on non-transitive targets that were failing

lint

rough-minister-58256

09/09/2017, 1:01 AM

O(what failed) vs O(the entire transitive change)

average-vr-56795

09/11/2017, 7:49 AM

Is there a design doc for fmt somewhere? I'm a little surprised by how it behaves, but seeing the story behind how it came to be would probably inform my surprise away 🙂

wide-energy-11069

09/11/2017, 8:18 AM

if you are surprised it formats transitively, there’s plan to change that to target spec only

wide-dog-50095

09/11/2017, 6:30 PM

how do I generate ensime config into ./.ensime as visual code, and most others expect it to be placed

plain-sundown-25537

09/11/2017, 10:35 PM

Hello, may be this is more a Pex question than Pants, how should we override the Python interpreter for PEX execution ?

rough-minister-58256

09/11/2017, 10:39 PM

at build time or run time?

bulky-motorcycle-41302

09/11/2017, 10:56 PM

@plain-sundown-25537 I've found that I can use PEX_PYTHON to control which python interpreter is used for executing a pex, even one that was built with a different python

bulky-motorcycle-41302

09/11/2017, 10:57 PM

I'm looking for information on how to control the python interpreter used to execute tests.

bulky-motorcycle-41302

09/11/2017, 10:57 PM

I would like to run python tests using both a Python 2 and a Python 3 interpreter, and to get reports for each, preferrably separately

bulky-motorcycle-41302

09/11/2017, 10:58 PM

Were I using tox, I could use the tox.ini file to set up a matrix of versions to use in running test commands.

bulky-motorcycle-41302

09/11/2017, 10:58 PM

I'm looking for a similar functionality here, though I'd settle for being able to run tests in each interpreter as sequential commands

bulky-motorcycle-41302

09/11/2017, 11:02 PM

I've been digging through the docs, but am having trouble locating information on how to accomplish this task

rough-minister-58256

09/11/2017, 11:07 PM

hm, makes sense