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

    average-vr-56795

    04/24/2018, 7:11 PM
    Let’s pick one or the other for now and iterate from there; will be easy to switch to the other behaviour if we decide it makes more sense 🙂
  • w

    witty-crayon-22786

    04/24/2018, 7:11 PM
    ...i'm also thinking that an integration test testing the compiler invoke case would be helpful
  • w

    witty-crayon-22786

    04/24/2018, 7:12 PM
    ok. then i would say fail by default. because i think it will almost always be an error. and when it isn't, we should make sure it's easy to remove a path from a Snapshot before merging it.
  • a

    average-vr-56795

    04/24/2018, 7:13 PM
    Sounds good 🙂
  • w

    witty-crayon-22786

    04/24/2018, 7:30 PM
    @average-vr-56795: the Digest at the root of a Snapshot will always be a Directory...?
  • w

    witty-crayon-22786

    04/24/2018, 7:31 PM
    or is that not guaranteed
  • a

    average-vr-56795

    04/24/2018, 8:59 PM
    @witty-crayon-22786 Is guaranteed
    👍 1
  • a

    aloof-angle-91616

    04/25/2018, 1:58 AM
    ...am i missing something obvious or is there not a way to ask a scheduler for a singleton rule? because requesting with an empty array of subjects gets me an empty array of products and requesting with a random value says no installed rules can satisfy 
    Select(MySingleton)
    for a root subject of type 
    Any
    (or 
    None
    )
  • w

    witty-crayon-22786

    04/25/2018, 3:24 PM
    that sounds right, yea.
  • w

    witty-crayon-22786

    04/25/2018, 3:41 PM
    ...what's the usecase?
  • a

    aloof-angle-91616

    04/25/2018, 4:12 PM
    good question
  • a

    aloof-angle-91616

    04/25/2018, 4:19 PM
    my attempts to refactor the native backend using rules was mostly successful
  • a

    aloof-angle-91616

    04/25/2018, 4:20 PM
    one of the datatypes involved was 
    Platform
    which contains the result of 
    normalize_os_name(get_os_name())
  • a

    aloof-angle-91616

    04/25/2018, 4:20 PM
    where i defined Platform I made it into a singleton rule
  • a

    aloof-angle-91616

    04/25/2018, 4:20 PM
    ok, i think i know what i should do, actually
  • a

    aloof-angle-91616

    04/25/2018, 4:21 PM
    basically i wanted to get a Platform and do something conditional on which platform i was on (which is done in a sanitary-ish way)
  • a

    aloof-angle-91616

    04/25/2018, 4:21 PM
    in this case it was "use the xcode cli tools to compile 
    python_dist()
    specifically"
  • a

    aloof-angle-91616

    04/25/2018, 4:21 PM
    and so i was trying to do a product_request() from within the BuildLocalPythonDistributions task
    w
    • 2
    • 10
  • a

    aloof-angle-91616

    04/25/2018, 4:23 PM
    but i think i can turn this into a separate 
    @rule
    which selects the native toolchain and platform and returns a wrapper type which is either a 
    Linker
    object as usual, or a 
    Linker
    object that uses the xcode tools. the wrapper type because then i can ensure that it's only for building `python_dist()`s
  • a

    aloof-angle-91616

    04/25/2018, 4:24 PM
    and btw, also immediately going to add testing for not-just-python-native-extensions but also native code using ctypes, since that is what our internal use case does (in an immediately subsequent pr)
  • a

    aloof-angle-91616

    04/25/2018, 4:25 PM
    btw thanks for the comments on the 
    typed_datatype()
    pr (or i guess just 
    datatype()
    now) @witty-crayon-22786, i know exactly how to make the changes and will probably be able to get to them today
  • s

    straight-waiter-86090

    04/25/2018, 6:24 PM
    @straight-waiter-86090 has left the channel
  • a

    aloof-angle-91616

    04/25/2018, 8:41 PM
    @witty-crayon-22786 just made https://github.com/pantsbuild/pants/pull/5747 which slims down the changes as much as possible. i made the native backend changes as small as possible, but i left them in because i could not figure out another way to actually test that these changes work outside of writing another whole new backend. i have another finished PR which i can put up as soon as this one is merged.
  • a

    aloof-angle-91616

    04/25/2018, 8:58 PM
    the other finished pr isn't time sensitive whatsoever though
  • a

    aloof-angle-91616

    04/26/2018, 8:39 AM
    @witty-crayon-22786 i fixed this pr to use 
    datatype()
    like you suggested, thanks
  • a

    aloof-angle-91616

    04/26/2018, 9:17 AM
    i had an error in the register rules in backends pr that i couldn't figure out on CI which also reproed locally, will look at this with fresh eyes tomorrow but this is vaguely mystifying as of now: https://github.com/pantsbuild/pants/pull/5747#issuecomment-384571496
  • a

    aloof-angle-91616

    04/26/2018, 9:21 AM
    even a tip on where i might be able to repro that failure without waiting however long for the release script to fail would be nice, i can def figure it out though
  • w

    witty-crayon-22786

    04/26/2018, 3:03 PM
    @aloof-angle-91616: i believe that the script manually specifies which backends to use maybe...?
  • w

    witty-crayon-22786

    04/26/2018, 3:04 PM
    would just grep for the "in a venv" bit.
  • a

    aloof-angle-91616

    04/26/2018, 5:34 PM
    ah amazing
1...478479480...906Latest