Hello, I’m trying to get 👖 to run with a separate configuration file (with the ultimate goal of running in CodeBuild w/ build caching). I’ve set the following in a “local.ini” file—

[GLOBAL]
pants_workdir: /Users/ddcaaronm/Desktop/pants/.pants.d
plugin_cache_dir: /Users/ddcaaronm/Desktop/pants/plugin_cache/

[cache]
read_from: ['/Users/ddcaaronm/Desktop/pants/.cache']
write_to: ['/Users/ddcaaronm/Desktop/pants/.cache']

Most concerningly, compilation fails with the following:

[1/6] Compiling 1 scala-js source in 1 target (src/scala/com/coxautoinc/mediastrategy/strategize/domain:domain-js).
09:18:46 00:24       [compile]

                   compile(src/scala/com/coxautoinc/mediastrategy/strategize/domain:domain-js) failed: /Users/ddcaaronm/git/strategy-recommendations is not a directory containing /Users/ddcaaronm/Desktop/pants/.pants.d/resolve/scala-js-compile/a53b6931478d/src.scala.com.coxautoinc.mediastrategy.strategize.domain.domain-js/current/z.analysis
FAILURE: Compilation failure: Failed jobs: compile(src/scala/com/coxautoinc/mediastrategy/strategize/domain:domain-js)

Is this caused by a configuration issue? Secondly, when I look at my config w/

./pants --pants-config-files=local.ini options cache

many directories don’t appear to inherit from the settings specified in

local.ini

. Would I need to specificy all of these options? I expected these to be picked up from the

cache.read_from/write_to

config.

cache.read_from = ['/Users/ddcaaronm/Desktop/pants/.cache'] (from CONFIG in local.ini)
...
cache.write_to = ['/Users/ddcaaronm/Desktop/pants/.cache'] (from CONFIG in local.ini)
cache.analysis.read_from = ['/Users/ddcaaronm/git/strategy-recommendations/.cache'] (from HARDCODED)
cache.analysis.write_to = ['/Users/ddcaaronm/git/strategy-recommendations/.cache'] (from HARDCODED)
cache.analysis.zinc.read_from = ['/Users/ddcaaronm/git/strategy-recommendations/.cache'] (from HARDCODED)
cache.analysis.zinc.write_to = ['/Users/ddcaaronm/git/strategy-recommendations/.cache'] (from HARDCODED)
cache.bash-completion.read_from = ['/Users/ddcaaronm/git/strategy-recommendations/.cache'] (from HARDCODED)
cache.bash-completion.write_to = ['/Users/ddcaaronm/git/strategy-recommendations/.cache'] (from HARDCODED)
cache.bench.read_from = ['/Users/ddcaaronm/git/strategy-recommendations/.cache'] (from HARDCODED)
cache.bench.write_to = ['/Users/ddcaaronm/git/strategy-recommendations/.cache'] (from HARDCODED)
cache.binary.read_from = ['/Users/ddcaaronm/git/strategy-recommendations/.cache'] (from HARDCODED)
cache.binary.write_to = ['/Users/ddcaaronm/git/strategy-recommendations/.cache'] (from HARDCODED)

is there a command to say, "build all the python targets where the compatibility requirement is this"? context is one of our targets is only py3.7, and we want to separate the ci command for py3.7 from the rest

I'm trying to write a code generating task. This core logic require to run javac with a specfic jar in the classpath and a given set of arguments. So far I've written this:

class QueryDslGen(SimpleCodegenTask, JavacCompile):
    def __init__(self, *args, **kwargs):
        super(QueryDslGen, self).__init__(*args, **kwargs)

    def synthetic_target_type(self, target):
        return JavaLibrary

    def is_gentarget(self, target):
        return isinstance(target, QueryDslLibrary)

    def execute_codegen(self, target, target_workdir):
        if not isinstance(target, QueryDslLibrary):
            raise TaskError('Invalid target type "{class_type}" (expected QueryDslLibrary)'
                            .format(class_type=type(target).__name__))

        result = ???

        if result != 0:
            raise TaskError('twirl-gen ... exited non-zero ({code})'.format(code=result))

    @classmethod
    def get_args_default(cls, bootstrap_option_values):
        return super(JavacCompile, cls).get_args_default(cls, bootstrap_option_values) + ('-proc:only', '-processor', 'com.querydsl.apt.jpa.JPAAnnotationProcessor', '-s', target_workdir)

obviously I cannot use the

get_args_default

to pass the argument since I need the

target_workdir

value passed to

execute_codegen

. I suppose I need to use

self.context.options

somehow but I'm not sure how. Furthermore I'm a complete python newbie so I'm not sure how python will work with regards to the multiple inheritance and the diamond I've got here

any direction where I can find some guidance ?

I'm really used to statically type language hence I'm a bit lost when I see an argument passed around and want to know its type. For exemple with the context, I managed to find a reference here https://www.pantsbuild.org/dev_tasks.html but in a more general way how do you procede ? you just run and print ?

Hi I'm still stuck on my QueryDslGen task. I'm having trouble to compute the classpath of a target. Here is what I've achieved so far:

-.py

In execute_codegen, line 130 the target_classpath does not contain what I want, i.e it contains scala std libs but not the target dependencies like the hibernate library in the following setup :

querydsl_library(
    name='meta-model',
    dependencies=[
        'libs/hibernate/src/main/scala',
    ],
    sources=rglobs('*.java'),
)

I'm still confused with how pants exactly works

is there like a dev presentation of pants internals ?

(I'm using 1.9.0 btw)

our scaladoc fails with gc overhead limit, in that case if we want to supply different jvm args for scaladoc we would put a clause like

[doc.scaladoc]

and have a dic of

jvm_options

. does that sound correct?

quick question: how do I force pants to use python3 when I do a

./pants run <python target>

?

if that's possible to begin with that is

the target needs to have a compatibility clause

https://www.pantsbuild.org/python_readme.html search for compatibility

hi sorry to ask again but I'm really stuck. I'm still working on my code generation pants plugin that uses javac annotation processor. I managed to go one step further. i.e. I have my target

a

which declare

b1

and

b2

as dependencies (where

b1

andb

b2

are ScalaLibrary with source in my mono repo). I managed to duplicate the logic that compute the directory where class files of

b1

and

b2

will be put but I cannot find how the classpath for the compilation of

b1

and

b2

is computed. This raise a more general question on how pants works. When I look at JavacCompile task, it executes itself on a collection of target it finds in its context. Hence, when I say

./pants compile path/to/my/library

it does not direcly translate to the call of some

def compile(target)

function. I wonder how much meta data is gathered on targets before calling the desired task.

perhaps a quick question to be sure that I don't go in a wall. Somebody told me on this chanel that code gen needs to be run before compile. This statement is only valid for one target, right ? If a code gen target

a

depends on java_library target

b

, code gen of

a

is run after compilation of

b

?

i'm hacking on some ubertooth one chips to broadcast max quality aptx hd audio and it looks like the existing C/C++ support with the

python_dist()

integration will allow me to put all the disparate libraries together with zero effort out of the box. that or i may turn up some bugs, either way it's hype

so I just tried to publish an artifact using

./pants publish pants-plugins/scala:: --no-publish-jar-prompt

and I have the following message :

14:53:02 01:11   [publish]
14:53:02 01:11     [jar]
                   Skipping check for a clean None branch in test mode.Skipping publish of foo#bar;0.0.1 in test mode.

I do not found any mention of the "test mode" in the documentation and the only skip=true that I have is scaladoc:

./pants options --options-name=skip                
binary.dup.skip = False (from HARDCODED)
bundle.dup.skip = False (from HARDCODED)
detect-duplicates.skip = False (from HARDCODED)
doc.javadoc.skip = False (from HARDCODED)
doc.scaladoc.skip = True (from CONFIG in pants.ini)
fmt.skip = False (from HARDCODED)
lint.skip = False (from HARDCODED)
test.junit.skip = False (from HARDCODED)
test.pytest.skip = False (from HARDCODED)

Did I miss something ?

the answer is yes

I missed dryrun >_<

Hey, I’ve got a CI failure, not sure exactly where to start looking here…has anyone seen this before: https://travis-ci.org/pantsbuild/pants/jobs/454036165#L2627

looks like a flake: i restarted it.

sorry for the trouble!

hey, no worries! Thanks Stu!

So I upgraded my Mac to Mojave, and I’m getting an error with pantsbuild 1.9.0 I know there is something within the pants.ini that I can change to fix this issue. Where in the docs is this?

@nutritious-monkey-78516: https://github.com/pantsbuild/pants/issues/6548

Shoot I knew upgrading to Mojave was too early. I’m also having trouble building a PEX file. I get the following:

Failed to execute PEX file, missing macosx_10_14_x86_64-cp-27-cp27m compatible dependencies for:
                     pytest-cov
                     coverage

might just need a clean-all.