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

    happy-kitchen-89482

    12/10/2018, 7:02 PM
    That configd process has been up since 11/15, but the pantsd pidfile was written on 12/2. So 60 was never its pid.
  • h

    happy-kitchen-89482

    12/10/2018, 7:03 PM
    When I reboot my machine, 
    configd
    receives pid 60 again, so the problem recurs.
  • h

    happy-kitchen-89482

    12/10/2018, 7:04 PM
    So somehow the pantsd pidfile contains the wrong pid, in a somewhat dangerous way.
  • h

    happy-kitchen-89482

    12/10/2018, 7:04 PM
    Could it have been truncated somehow? Maybe written incompletely?
  • h

    happy-kitchen-89482

    12/10/2018, 7:06 PM
    Problem solved by nuking the pidfile, but that’s not ideal
  • w

    witty-crayon-22786

    12/10/2018, 7:11 PM
    wow. have definitely not seen that.
  • w

    witty-crayon-22786

    12/10/2018, 7:11 PM
    how old might that pidfile have been?
  • h

    happy-kitchen-89482

    12/10/2018, 7:33 PM
    See dates above
  • h

    happy-kitchen-89482

    12/10/2018, 7:33 PM
    Written a week ago
  • w

    witty-crayon-22786

    12/10/2018, 7:57 PM
    ah, missed that, apologies. no, have definitely not experienced that. a ticket would not hurt.
  • w

    witty-crayon-22786

    12/10/2018, 7:58 PM
    i don't think that the format of that file has changed
  • f

    fancy-queen-20734

    12/11/2018, 1:24 AM
    - (moved channel)
  • a

    average-vr-56795

    12/11/2018, 7:19 PM
    Observation of the day: conditional code which calls coroutines is awkward to unit test...
    Copy code
    @rule(Foo, [Select(Bar)]
def foo(bar):
  if True:
    b = yield Get(Baz, Bar, bar)
  else:
    b = yield Get(Bash, Bar, bar)
  yield b.Foo
    In 
    run_rule
    , I need paths from 
    Bar
    to both 
    Baz
    and 
    Bash
    in my mock rules dict, not just the one that is used, because of the pre-parsing of `Get`s
  • a

    aloof-angle-91616

    12/11/2018, 7:21 PM
    yep! i made an apparently hard to understand comment about this in the native backend because i wanted to do a 
    yield Get(...)
    in a lambda with 
    Platform#resolve_platform_specific()
    a
    • 2
    • 2
  • a

    average-vr-56795

    12/11/2018, 7:21 PM
    We could maybe defer the checking until runtime or something, in tests
    ✨ 1
  • a

    aloof-angle-91616

    12/11/2018, 7:21 PM
    that's a very interesting idea
  • w

    witty-crayon-22786

    12/11/2018, 7:25 PM
    i'd be fine with that
  • a

    aloof-angle-91616

    12/11/2018, 7:25 PM
    what if we could make the runtime checking part configurable (hm, i guess this isn't useful outside of tests now that i think of it)
  • a

    average-vr-56795

    12/11/2018, 7:39 PM
    @aloof-angle-91616 Any thoughts on how we could ergonomically typecheck datatype tuple contents?
    a
    • 2
    • 19
  • a

    average-vr-56795

    12/13/2018, 10:30 AM
    @aloof-angle-91616 I was looking at your dependency cycle because I was curious…
  • a

    average-vr-56795

    12/13/2018, 10:30 AM
    graph.000.svg
    graph.000.svg
  • a

    average-vr-56795

    12/13/2018, 10:31 AM
    That’s the graph I ended up with, with a cycle in 
    transitive_hydrated_target
    on 
    testprojects/tests/python/pants/dummies:target_with_source_dep
    - I have no idea why
  • a

    aloof-angle-91616

    12/13/2018, 7:38 PM
    oh i just assumed it was a concurrency issue screwing up our graph
  • a

    aloof-angle-91616

    12/13/2018, 7:38 PM
    i can see at least one literal loop
  • a

    aloof-angle-91616

    12/13/2018, 7:39 PM
    if 
    dependencies
    of a 
    TransitiveHydratedTarget
    contains the 
    root
    target then i can see order-dependence causing this, which would mean i'd only need to change one line of code to see it stop. hm.
  • a

    aloof-angle-91616

    12/13/2018, 7:43 PM
    that would have been too easy
  • w

    witty-crayon-22786

    12/13/2018, 8:59 PM
  • a

    average-vr-56795

    12/14/2018, 12:09 PM
    Grrrrrr why isn’t more of 
    std::time::Duration
    stabilised? 😞
    a
    • 2
    • 8
  • a

    average-vr-56795

    12/14/2018, 5:48 PM
    Using functions as parameters or return values is so fiddly in Rust 😞
  • a

    average-vr-56795

    12/14/2018, 6:23 PM
    @witty-crayon-22786 Ok, I give up, I’m asking for help… How can I make a closure which is an 
    FnMut
    not an 
    FnOnce
    , which needs to clone in some 
    Arc
    from its surrounding context?
1...599600601...906Latest