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

    witty-crayon-22786

    12/01/2017, 8:10 PM
    sorry, was halfway through a review
  • w

    witty-crayon-22786

    12/01/2017, 8:10 PM
    will post it and then continue
  • w

    witty-crayon-22786

    12/01/2017, 8:11 PM
    @average-vr-56795: yes, that absolutely seems like a bug in their implementation.
  • a

    average-vr-56795

    12/01/2017, 8:12 PM
    Cool, I'll file it with them 🙂 Thanks!
  • w

    witty-crayon-22786

    12/01/2017, 8:14 PM
    @average-vr-56795: but... um, i haven't used Streams before either.
  • w

    witty-crayon-22786

    12/01/2017, 8:14 PM
    my understanding is that the Stream should completely encapsulate its own lifetime.
  • a

    average-vr-56795

    12/01/2017, 8:14 PM
    Yeah, that's what I read them
  • a

    average-vr-56795

    12/01/2017, 10:46 PM
    @witty-crayon-22786 https://github.com/pingcap/grpc-rs/issues/123 in case you want to chime in or follow along
  • a

    average-vr-56795

    12/01/2017, 10:46 PM
    And with that, I disappear for the weekend 🙂
  • w

    witty-crayon-22786

    12/01/2017, 10:48 PM
    @average-vr-56795: enjoy!
  • r

    rough-minister-58256

    12/02/2017, 12:48 AM
    welp
  • r

    rough-minister-58256

    12/02/2017, 12:49 AM
    this repl output bug, afaict, is a GIL-based race
  • w

    witty-crayon-22786

    12/02/2017, 12:49 AM
    argh. sorry.
  • w

    witty-crayon-22786

    12/02/2017, 12:50 AM
    ...if it makes you feel any better, i spent most of the week whacking on bash scripts. i am thoroughly chastised.
  • r

    rough-minister-58256

    12/02/2017, 12:51 AM
    haha
  • w

    witty-crayon-22786

    12/02/2017, 12:51 AM
    Copy code
    while read path_tuple
do
  file_path = $(echo "$path_tuple" | awk -F'\t' '{print $1}')
  url_path = $(echo "$path_tuple" | awk -F'\t' '{print $2}')
  ...
done
  • r

    rough-minister-58256

    12/02/2017, 12:54 AM
    I just mean like, my write lock approach isn’t working because afaict unless the main thread does something to release the GIL it ends up racing to write an exit chunk before either stdio writer chunks can send their chunks.
  • r

    rough-minister-58256

    12/02/2017, 12:54 AM
    considering switching back to file mocks for stdout/err
  • r

    rough-minister-58256

    12/02/2017, 12:55 AM
    @rough-minister-58256 uploaded a file:

    imageâ–¾

  • w

    witty-crayon-22786

    12/02/2017, 12:56 AM
    @rough-minister-58256: how would the forked process write to those?
  • r

    rough-minister-58256

    12/02/2017, 12:57 AM
    ah, yeah
  • r

    rough-minister-58256

    12/02/2017, 12:57 AM
    welp
  • w

    witty-crayon-22786

    12/02/2017, 12:58 AM
    @rough-minister-58256: it's possible that N threads to manage N pipes starts simpler and then gets more complicated than having a select loop =/
  • w

    witty-crayon-22786

    12/02/2017, 12:58 AM
    ...one select loop
  • w

    witty-crayon-22786

    12/02/2017, 1:00 AM
    @rough-minister-58256: another possibility would be to have the main thread consume chunks off a queue to put on the socket
  • r

    rough-minister-58256

    12/02/2017, 1:00 AM
    yeah, that was my next approach
  • w

    witty-crayon-22786

    12/02/2017, 1:00 AM
    ...but i think straight to the socket should be fine, as long as all of the threads are joined before exiting...?
  • r

    rough-minister-58256

    12/02/2017, 1:01 AM
    no, joining doesn’t help
  • r

    rough-minister-58256

    12/02/2017, 1:01 AM
    by the time the stdio writer threads wake up and fire, the main threads already writtten an exit chunk and broken the other sides iter_chunk loop
  • r

    rough-minister-58256

    12/02/2017, 1:02 AM
    the exiter would need to know about them and join them before sending the exit chunk
1...379380381...906Latest