https://pantsbuild.org/ logo
Powered by
#development
  • w

    witty-crayon-22786

    06/16/2018, 12:53 AM
    the "before/after/around" bit is also called "aspect oriented" programming, iirc
    👌 1
  • a

    aloof-angle-91616

    06/16/2018, 12:54 AM
    wow i had heard the term and never looked it up
  • a

    aloof-angle-91616

    06/16/2018, 12:54 AM
    wow this is a wikipedia article i should have read sooner thanks a lot
  • a

    aloof-angle-91616

    06/16/2018, 12:59 AM
    ...and also thanks for the issue link which i should have checked out a lot earlier
    👍 1
  • a

    aloof-angle-91616

    06/16/2018, 1:09 AM
    are optional products considered super egregious i'm assuming the idea is that they're not
  • a

    aloof-angle-91616

    06/16/2018, 5:34 AM
    posted the PR, no clue how CI will do (although testing "works on my machine") but this is just to get it up https://github.com/pantsbuild/pants/pull/5970
  • a

    average-vr-56795

    06/18/2018, 1:17 PM
    @witty-crayon-22786 As far as I can tell, 
    tests/python/pants_test/task
    is testing file-based invalidation using the v1 file invalidation model which is no longer in use. Can those tests be deleted?
  • a

    average-vr-56795

    06/18/2018, 1:37 PM
    Oh waaaaaaa…. Hey @witty-crayon-22786, what’s the plan for killing… Whichever of 
    TestBase
    and 
    BaseTest
    is the bad one?
  • a

    average-vr-56795

    06/18/2018, 1:37 PM
  • a

    average-vr-56795

    06/18/2018, 2:02 PM
    Ooh, I think that one turned out to be easy 🙂 https://github.com/pantsbuild/pants/pull/5973
  • w

    witty-crayon-22786

    06/18/2018, 6:25 PM
    @rough-minister-58256: re no-fork for https://github.com/pantsbuild/pants/issues/4769 : yes, there are definitely hurdles.
  • w

    witty-crayon-22786

    06/18/2018, 6:26 PM
    but the benefits are massive. forking the way we do now is, imo, not a sustainable thing
  • w

    witty-crayon-22786

    06/18/2018, 6:26 PM
    and once Subsystems are migrated in, there will be a place for people to locate state (edit: not what i meant. they still won't be a good place to put state. but they will be a principaled cross-rule place to collect things)
  • r

    rough-minister-58256

    06/18/2018, 6:27 PM
    yeah, I agree it’d be good to do away w/ the fork-wo-exec pattern if we can do that safely
  • w

    witty-crayon-22786

    06/18/2018, 6:28 PM
    and i don't want to connect "forking for any reason" with "forking for an individual run". if anything, not forking for individual pants runs makes running 
    @rules
    multi-process easier
  • r

    rough-minister-58256

    06/18/2018, 6:28 PM
    and while I’m more than happy to work to knock out said hurdles, a forkless model presents a fundamental challenge: code running deeply in pants (and in private plugins we can’t see) can and is likely to cause gc issues in the daemon in that model
  • w

    witty-crayon-22786

    06/18/2018, 6:29 PM
    sure. and we'll need to improve our sandboxing significantly.
  • r

    rough-minister-58256

    06/18/2018, 6:29 PM
    so I just want to raise those issues early and make sure they’re understood and that we’ve considered possible alternatives etc
  • r

    rough-minister-58256

    06/18/2018, 6:29 PM
    yeah, for sure
  • r

    rough-minister-58256

    06/18/2018, 6:30 PM
    I know we’ve discussed sandboxing via maybe e.g. skylark for rule execution
  • r

    rough-minister-58256

    06/18/2018, 6:30 PM
    but do we have any ideas on other modes of sandboxing that you think weould help out here?
  • r

    rough-minister-58256

    06/18/2018, 6:31 PM
    i.e. 
    Task
    sandboxing?
  • w

    witty-crayon-22786

    06/18/2018, 6:31 PM
    that is the primary one i know of at the moment.
  • r

    rough-minister-58256

    06/18/2018, 6:31 PM
    since nofork means 
    Task
    execution which is the wild west
  • w

    witty-crayon-22786

    06/18/2018, 6:31 PM
    @rough-minister-58256: sorry, what do you mean
  • w

    witty-crayon-22786

    06/18/2018, 6:32 PM
    if you mean v1 Tasks, then no... we would still fork for them
  • r

    rough-minister-58256

    06/18/2018, 6:32 PM
    oh
  • w

    witty-crayon-22786

    06/18/2018, 6:32 PM
    the proposal is a new entry point.
  • w

    witty-crayon-22786

    06/18/2018, 6:33 PM
    or, potentially, a model of frontend rules that runs before we fork
  • w

    witty-crayon-22786

    06/18/2018, 6:33 PM
    (and then if there is nothing to fork for, we don't)
1...538539540...906Latest