https://pantsbuild.org/ logo
Join Slack
Powered by
Experiencing an issue with pylint while running `....
# development
w
Experiencing an issue with pylint while running 
./pants --no-v1 lint
in Pants 1.30.2rc0. Stack in thread
Untitled
Untitled
haven't seen this before. in previous 1.X versions i'd successfully got pylint to run on the v2 engine but never got a clean run
a
could you try with 
--no-process-execution-cleanup-local-dirs
and attempt to execute the command line yourself to repro?
w
hey sorry just saw this, @aloof-angle-91616 does this mean i should cd in to the process exceution dir left behind and try to run 
pylint
form my local environment or through the 
pylint_runner.pex
there's three pex files in there so im not sure what the right incantation would be
a
ah yes in pants v2 we were able to add a 
run.sh
script so you don't have to remember the right command line
if you run pants with 
-ldebug
it should print out the command line and environment variables used
w
ok ill do that thanks
a
lmk if that's clear
w
so far it is! I might have questions depending on the debug log output
a
it is definitely weird that this is happening, especially since you mentioned that it was successfully being invoked before
w
to be fair, last time i tried lintin on the v2 engine was months ago
a
ah! that is helpful
w
or at least more than a month ago
and several pants versions back as well
a
and could you post the command line and env vars here? i'm suspecting the PEX_PATH we're using is wrong, or we're not writing the entry_point to the pylint runner pex
i would also like to know the output of 
unzip -p pylint_runner.pex PEX-INFO | jq
w
ok might have to use a different selector than 
src::
because the output is MASSIVE
a
ah oops that could be our fault too. i have been tryiing to tell people we should make it easier for users to pipe certain tagged output into certain places instead of monolithic log levels but nobody else is seeing it for some reason
w
ok here's the command about to try to repro manually
Untitled
oh oyu know what there is a run.sh in here 🙈
a
oh awesome yes!
w
yes able to reproduce, uploading the run.sh here
__run.sh
__run.sh
a
i would also be interested in knowing the output of 
unzip -l pylint_runner.pex | grep -F .deps/pylint
(and whether there's a 
.deps/pylint/__main__.py
)
w
ok one seq, getting the PEX_INFO now
PEX INFO
Untitled
the 
unzip | grep .deps
line returns nothing
a
oh sorry then i guess you will want to try it on 
pylint.pex
w
there is no .deps folder in that pex
right ok
a
since that's in the 
pex_path
in PEX-INFO
w
here's the pylint tree
pex_listing.txt
a
so 
.deps/pylint-2.6.0-py3-none-any.whl/pylint/__main__.py
is in there
ridiculous
what happens if you try 
PEX_MODULE=pylint.__main__ PEX_VERBOSE=9 ./pylint.pex
?
the next thing i'm thinking is that the interpreter selection might be screwy so i would want to know what the 
interpreter_constraints
field in the PEX-INFO of 
pylint.pex
has
w
well that gives an import error due to not having the plugins in the pythong path
a
ah yes
perhaps add 
PEX_PATH=requirements.pex
?
or
oh
i mean 
PEX_INHERIT_PATH=fallback PYTHONPATH="$(pwd)/.__plugins"
feel free to post an issue about this, no rush, i know you have other things to do
w
yeah took a lunch break... I'll play around with these commands and try to gather the information
a
thank you, this is great
the zip was super helpful
w
you got it
w
@wonderful-iron-54019: is this an upgrade blocker for you all? i was going to cut 
1.30.2rc1
with the cache fix we discussed yesterday, but if this blocks we should maybe look into it first.
w
we're currently running pylint outside of pants so no
we have a path to v2 runing vanilla 
pylint
and then trying to move that to pants
w
ok. so the big question is whether it still happens on 
2.0.x
then i think. and you’re not in a position to confirm that probably
w
i mean i could attempt that locally
w
we have users using pylint on 
2.0.x
, so the question would be whether it is something environment specific, or whether it is a bug fixed between 
1.30.x
and head
ok, i’m going to cut 
1.30.2rc1
, and move triaging #10886 into 2.0 so that we make sure to take a look at whether it repros on 
2.0.x
before release.
w
cool thanks
w
thanks a lot!
and thanks @aloof-angle-91616 for suggesting taring the execution directory.
a
thank you for providing the 
2.0.x
context. still getting familiar with what our priorities are
w
w
nice! thanks
Open in Slack
PreviousNext
Powered by