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

    witty-crayon-22786

    10/31/2017, 8:40 PM
    but what is in ~/.cache/pants is everything cargo has fetched
  • w

    witty-crayon-22786

    10/31/2017, 8:41 PM
    so, git clones, registry copies, installed tools
  • w

    witty-crayon-22786

    10/31/2017, 8:41 PM
    grpc/rustfmt are compiled and then installed there
  • w

    witty-crayon-22786

    10/31/2017, 8:41 PM
    and cached
  • r

    rough-minister-58256

    10/31/2017, 8:41 PM
    ic
  • w

    witty-crayon-22786

    10/31/2017, 8:42 PM
    but @wide-energy-11069: this is particularly relevant to you... you shouldn't actually be experiencing the grpc bit.
  • w

    witty-crayon-22786

    10/31/2017, 8:42 PM
    unless you're wiping your toolchain, or have never let it finish.
  • w

    witty-crayon-22786

    10/31/2017, 8:43 PM
    @rough-minister-58256: going to roll with this unless you have any other ideas: https://github.com/pantsbuild/pants/issues/3774#issuecomment-340900815
  • r

    rough-minister-58256

    10/31/2017, 8:49 PM
    hrm
  • r

    rough-minister-58256

    10/31/2017, 8:53 PM
    do you see the same behavior with e.g. 
    repl.py
    ?
  • w

    witty-crayon-22786

    10/31/2017, 8:54 PM
    ...good question.
  • r

    rough-minister-58256

    10/31/2017, 8:54 PM
    seems pretty obvious where the message is being emitted from
  • w

    witty-crayon-22786

    10/31/2017, 8:54 PM
    yes. but not how it is ending up on the socket.
  • w

    witty-crayon-22786

    10/31/2017, 8:54 PM
    will try.
  • r

    rough-minister-58256

    10/31/2017, 8:54 PM
    guessing there’s a stdout/stderr default to 
    None
    in the javaexecutor
  • w

    witty-crayon-22786

    10/31/2017, 8:55 PM
    which would not end up on the socket
  • w

    witty-crayon-22786

    10/31/2017, 8:55 PM
    not directly, certainly
  • r

    rough-minister-58256

    10/31/2017, 8:58 PM
    in the pantsd/pantsd-runner context, inheriting 
    sys.stdout
    would net a 
    NailgunStreamWriter
  • w

    witty-crayon-22786

    10/31/2017, 8:58 PM
    yea, it does.
  • w

    witty-crayon-22786

    10/31/2017, 8:59 PM
    i don't see this issue with the python repl... probably because it doesn't subprocess.Popen
  • w

    witty-crayon-22786

    10/31/2017, 9:15 PM
  • r

    rough-minister-58256

    11/01/2017, 2:56 AM
  • r

    rough-minister-58256

    11/01/2017, 2:56 AM
    the idea of building the native engine in the sdist may be easier than we think
  • r

    rough-minister-58256

    11/01/2017, 2:57 AM
  • r

    rough-minister-58256

    11/01/2017, 2:57 AM
    if we had a python native_engine.sh path it seems like it could be drop in
  • r

    rough-minister-58256

    11/01/2017, 2:58 AM
    then we’d call that path via ./pants during dev and from distribution to setup.py time our sdist would act as expected by python folks
  • r

    rough-minister-58256

    11/01/2017, 6:43 AM
    also, I had forgotten but 
    MODE=debug
    is good for a ~66% rust compilation speedup (but has a runtime slowdown)
  • w

    witty-crayon-22786

    11/01/2017, 4:13 PM
    yea
  • w

    witty-crayon-22786

    11/01/2017, 4:13 PM
    (at least 66%)
  • r

    rough-minister-58256

    11/01/2017, 5:42 PM
    btw, re: terminal support you may want to look into 
    os.openpty
    vs 
    os.pipe
1...371372373...906Latest