https://pantsbuild.org/ logo
Powered by
#general
  • f

    fast-megabyte-33579

    03/29/2017, 9:49 PM
    would you tie it in to that?
  • r

    rough-minister-58256

    03/29/2017, 9:49 PM
    yeah that’d be the idea
  • f

    fast-megabyte-33579

    03/29/2017, 9:49 PM
    👏 that sort of thing would make me v happy
  • r

    rough-minister-58256

    03/29/2017, 9:49 PM
    When arc is integrated with a lint toolkit, it enables the arc lint command and runs lint on changes during arc diff.
  • r

    rough-minister-58256

    03/29/2017, 9:49 PM
    etc
  • f

    fast-megabyte-33579

    03/29/2017, 9:50 PM
    currently we don't use any of the scalastyle stuff in compile (because our code fails a lot), if we moved it to a separate lint step that'd be fine I think
  • r

    rough-minister-58256

    03/29/2017, 9:51 PM
    yeah. theoretically in CI you owuldn’t even need to combine test and lint at the goal registration level - you could just run them as discrete steps in a build to potentially fail earlier than bombing with linter errors after you’ve run the 45 minutes worth of tests.
  • h

    happy-kitchen-89482

    03/29/2017, 11:12 PM
  • h

    happy-kitchen-89482

    03/29/2017, 11:13 PM
    Finally addresses the blocker that prevents us from upgrading pytest
  • r

    rough-minister-58256

    03/29/2017, 11:13 PM
    niiiiice
  • h

    happy-kitchen-89482

    03/29/2017, 11:22 PM
    There were a few hoops to jump through
  • h

    happy-kitchen-89482

    03/29/2017, 11:23 PM
    “parsing junit xml” was the least of it
  • w

    witty-crayon-22786

    03/30/2017, 12:28 AM
    counterpoint to the "take all checks out of `compile`" movement:
    Copy code
    Invalidated 238 targets.
F401:ERROR   tests/python/pants_test/base_test.py:021 'BuildConfiguration' imported but unused
     |from pants.build_graph.build_configuration import BuildConfiguration

F401:ERROR   tests/python/pants_test/base_test.py:022 'BuildFileAddressMapper' imported but unused
     |from pants.build_graph.build_file_address_mapper import BuildFileAddressMapper

F401:ERROR   tests/python/pants_test/base_test.py:024 'BuildFileParser' imported but unused
     |from pants.build_graph.build_file_parser import BuildFileParser

F401:ERROR   tests/python/pants_test/base_test.py:025 'MutableBuildGraph' imported but unused
     |from pants.build_graph.mutable_build_graph import MutableBuildGraph

F821:ERROR   tests/python/pants_test/base_test.py:268-274 undefined name 'EngineInitializer'
     |    graph_helper = EngineInitializer.setup_legacy_graph(
     |        self.pants_ignore_patterns,
     |        self.pants_workdir,
     |        build_ignore_patterns=self.build_ignore_patterns,
     |        exclude_target_regexps=None,
     |        subproject_roots=None,
     |      )

F821:ERROR   tests/python/pants_test/base_test.py:277 undefined name 'LiteralTargetRoots'
     |        LiteralTargetRoots([]),


FAILURE: 6 Python Style issues found. For import order related issues, please try `./pants fmt.isort <targets>`


17:27:14 00:17   [complete]
               FAILURE
  • w

    witty-crayon-22786

    03/30/2017, 12:28 AM
    nice to see all of those at once.
  • w

    witty-crayon-22786

    03/30/2017, 12:29 AM
    some checks are semantic checks... others are style. i haven't really seen an attempt to split them, but would like to.
  • r

    rough-minister-58256

    03/30/2017, 12:31 AM
    ez enough to 
    ./pants lint compile
    as standard workflow tho, no?
  • w

    witty-crayon-22786

    03/30/2017, 12:31 AM
    sure, but that doesn't split them
  • w

    witty-crayon-22786

    03/30/2017, 12:31 AM
    cake, and the eating of it
  • r

    rough-minister-58256

    03/30/2017, 12:34 AM
    imo, modular goal composition is better suited to support heterogeneous working styles - monolithic 
    compile
    forces a one-size-fits-all model
  • w

    witty-crayon-22786

    03/30/2017, 12:34 AM
    only two hard problems in computer science
  • w

    witty-crayon-22786

    03/30/2017, 12:35 AM
    can we call "checking for undefined symbols" 
    compile
    ? probably. but not unused imports.
  • f

    fast-nail-55400

    03/30/2017, 12:37 AM
    or "check" like rust does for 
    cargo check
    ?
  • w

    witty-crayon-22786

    03/30/2017, 12:38 AM
    yea
  • r

    rough-minister-58256

    03/30/2017, 12:41 AM
    I get the whole “composite compiler” thing, but it’s just not how 100% of people (e.g. python devs) are used to operating - so why force it on em.
  • w

    witty-crayon-22786

    03/30/2017, 12:44 AM
    for the same reason we ever change anything. to improve things.
  • r

    rough-minister-58256

    03/30/2017, 12:51 AM
    “improvement” feels quite subjective in this case
  • w

    witty-crayon-22786

    03/30/2017, 12:52 AM
    is having all of the undefined symbols reported better than having one reported?
  • r

    rough-minister-58256

    03/30/2017, 12:53 AM
    yes, because they affect the ability to execute - the ‘unused symbols’ do not and can be dealt with separately as pre-commit cleanup.
  • w

    witty-crayon-22786

    03/30/2017, 12:54 AM
    yea, so that's what i was saying about semantic vs style checks.
  • r

    rough-minister-58256

    03/30/2017, 12:56 AM
    maybe the mode change needed here would simply be: style errors on 
    compile
    -> warning, style errors on 
    lint
    -> error.
1...168616871688...2420Latest