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

    rough-minister-58256

    11/02/2015, 11:53 PM
    that part doesn’t exist yet, but it could. right now the BG is the only thing optimized (cached/updated/reused)
  • r

    rough-minister-58256

    11/02/2015, 11:54 PM
    we could back that out to just running pants via pailgun
  • r

    rough-minister-58256

    11/02/2015, 11:54 PM
    and then put that back in post-engine
  • e

    enough-analyst-54434

    11/02/2015, 11:54 PM
    Can we get a fresh graph via pailgun w/o Sergei's stuff?
  • r

    rough-minister-58256

    11/02/2015, 11:55 PM
    yeah
  • r

    rough-minister-58256

    11/02/2015, 11:55 PM
    that’s what I’m suggesting re: ‘back that out’ - fresh graph via pailgun
  • e

    enough-analyst-54434

    11/02/2015, 11:55 PM
    but then are you left with same to slower?
  • e

    enough-analyst-54434

    11/02/2015, 11:56 PM
    I'm guessing no
  • e

    enough-analyst-54434

    11/02/2015, 11:56 PM
    ie: at least all backends are loaded
  • e

    enough-analyst-54434

    11/02/2015, 11:56 PM
    and you re-load graph per main run
  • r

    rough-minister-58256

    11/02/2015, 11:56 PM
    approximately. haven’t measured the startup cost reduction tho. and then the watchman stuff isn’t actually updating anything
  • r

    rough-minister-58256

    11/02/2015, 11:57 PM
    the backends still get loaded inline currently - e.g. in the proposed PantsRunner.run()
  • r

    rough-minister-58256

    11/02/2015, 11:57 PM
    but we can move more of that into the daemon framework
  • r

    rough-minister-58256

    11/02/2015, 11:57 PM
    and having watchman check for updates to where those live etc
  • e

    enough-analyst-54434

    11/02/2015, 11:57 PM
    Yeah - it seems to me a viable 1.0 thing would be to have pailgun calling a daemon that had all backends loaded 1x and re-loaded a graph every time
  • e

    enough-analyst-54434

    11/02/2015, 11:58 PM
    But I'm guessing based on minimal exposed RB's and my estimations of time and availability of expert humans
  • e

    enough-analyst-54434

    11/02/2015, 11:58 PM
    ... you
  • r

    rough-minister-58256

    11/02/2015, 11:58 PM
    sure
  • e

    enough-analyst-54434

    11/02/2015, 11:59 PM
    My main concern is shipping 1.0 pre new year and not rocking the boat too much, being pretty sure we're stable for a 1.0 long term branch and not hurting installability or perf when doing so
  • e

    enough-analyst-54434

    11/02/2015, 11:59 PM
    hurting stability rather!
  • r

    rough-minister-58256

    11/02/2015, 11:59 PM
    does 1.0 include new engine?
  • e

    enough-analyst-54434

    11/02/2015, 11:59 PM
    no
  • r

    rough-minister-58256

    11/03/2015, 12:00 AM
    oh
  • e

    enough-analyst-54434

    11/03/2015, 12:00 AM
    its just touch pants.ini and configure ~nothing and pants works
  • e

    enough-analyst-54434

    11/03/2015, 12:00 AM
    This is basically done save scala BUILD.tools and breaking up more backends as sdists
  • e

    enough-analyst-54434

    11/03/2015, 12:01 AM
    Delicate dance between Twitter schedules, pants schedules, etc I know
  • e

    enough-analyst-54434

    11/03/2015, 12:02 AM
    I'm going to be switching my focus to other projects in the next few weeks so I'd like to attack the 1.0 blockers to leave the path forward clear, ie cut the branch within the next few weeks if possible
  • r

    rough-minister-58256

    11/03/2015, 12:03 AM
    switching focus to other projects == no longer working on pants?!
  • e

    enough-analyst-54434

    11/03/2015, 12:03 AM
    not as much time spent as I have
  • e

    enough-analyst-54434

    11/03/2015, 12:03 AM
    Never stopping work
1...660661662...2420Latest