https://pantsbuild.org/ logo
Powered by
#general
  • b

    bored-art-40741

    03/31/2016, 6:28 PM
    And I don't particularly like it
  • b

    bored-art-40741

    03/31/2016, 6:29 PM
    Basically, you append 
    .message
    to the end of the file path you're using for the lock, and you use that as the scratch space to store a message
  • b

    bored-art-40741

    03/31/2016, 6:29 PM
    The problem is that if the lock is, for example, in your workspace and you haven't gitignored it, you might accidentally check it in
  • b

    bored-art-40741

    03/31/2016, 6:29 PM
    How worried are people about that?
  • w

    witty-crayon-22786

    03/31/2016, 6:29 PM
    not worried about that
  • w

    witty-crayon-22786

    03/31/2016, 6:30 PM
    but there really isn't any way to put it in the same file?
  • b

    bored-art-40741

    03/31/2016, 6:30 PM
    There is not
  • b

    bored-art-40741

    03/31/2016, 6:30 PM
    Opening the file within the process drops the lock
  • r

    rough-minister-58256

    03/31/2016, 6:30 PM
    @bored-art-40741: does that mean the OwnerPrintingPidLock thing is fundamentally broken?
  • r

    rough-minister-58256

    03/31/2016, 6:30 PM
    because we write to the lockfile?
  • b

    bored-art-40741

    03/31/2016, 6:31 PM
    That lock is fundamentally broken because Pids aren't unique identifiers, and the presence of the pid in ps isn't a strong enough guarantee that the lock holder is actually still running
  • b

    bored-art-40741

    03/31/2016, 6:31 PM
    As a result, the lock can and does leak
  • r

    rough-minister-58256

    03/31/2016, 6:31 PM
    but the lock itself is sound aside from that concern?
  • b

    bored-art-40741

    03/31/2016, 6:31 PM
    ...no?
  • b

    bored-art-40741

    03/31/2016, 6:31 PM
    I mean, that concern is massive
  • b

    bored-art-40741

    03/31/2016, 6:31 PM
    The lock leaks, you can't be sure whether it's actually being held
  • b

    bored-art-40741

    03/31/2016, 6:32 PM
    It is therefore broken
  • w

    witty-crayon-22786

    03/31/2016, 6:32 PM
    @bored-art-40741: does the fact that OwnerPrintingPidLock stores a message have anything to do with its brokenness?
  • b

    bored-art-40741

    03/31/2016, 6:32 PM
    Anyway, a regular file lock works great, and we can get the message trick, we just need to use an additional file
  • b

    bored-art-40741

    03/31/2016, 6:33 PM
    @witty-crayon-22786: No, also note it doesn't store a message
  • b

    bored-art-40741

    03/31/2016, 6:33 PM
    It grabs the message based on psutils
  • w

    witty-crayon-22786

    03/31/2016, 6:33 PM
    ah... gotcha.
  • b

    bored-art-40741

    03/31/2016, 6:33 PM
    Well I take that back
  • r

    rough-minister-58256

    03/31/2016, 6:33 PM
    but it stores a pid
  • b

    bored-art-40741

    03/31/2016, 6:33 PM
    It stores a pid
  • b

    bored-art-40741

    03/31/2016, 6:33 PM
    Yeah
  • w

    witty-crayon-22786

    03/31/2016, 6:33 PM
    so
  • r

    rough-minister-58256

    03/31/2016, 6:33 PM
    if the lock is functional - in spite of potential for leaking on staleness - why cant you reuse its approach to writing while preserving the flock
  • b

    bored-art-40741

    03/31/2016, 6:33 PM
    Because it writes a pid into the flocked file
  • b

    bored-art-40741

    03/31/2016, 6:34 PM
    You can't write into the flocked file while the lock is held
1...111211131114...2420Latest