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

    witty-crayon-22786

    11/02/2015, 8:16 PM
    @enough-analyst-54434: looks like @bright-dinner-81701 is aware and will be working on it soon
  • e

    enough-analyst-54434

    11/02/2015, 8:16 PM
    K - I'm trying to repro and if I can I may send up a quick xfail
  • e

    enough-analyst-54434

    11/02/2015, 8:26 PM
    Hmm, flaky. I'm guessing no resolution solution hides a fetch failure on a valid solution as a no-solution
  • b

    bright-dinner-81701

    11/02/2015, 8:28 PM
    @enough-analyst-54434: yeah ill look at it today… its weird, from local machines things dont fail like that
  • h

    happy-kitchen-89482

    11/02/2015, 8:28 PM
    Related: When writing new task code for the new engine, do not use the currently common idiom of grabbing option values in 
    __init__()
    and putting them in member variables. Instead, access them directly as needed. Option resolution is expensive, so let's only do that when needed.
  • b

    bright-dinner-81701

    11/02/2015, 8:28 PM
    i dont know if travis ci is throttling or if their network is just not great
  • b

    bright-dinner-81701

    11/02/2015, 8:28 PM
    i think travis ci might access the outside world through a proxy
  • b

    bright-dinner-81701

    11/02/2015, 8:28 PM
    and maybe that cant handle the load
  • e

    enough-analyst-54434

    11/02/2015, 8:29 PM
    It runs in aws
  • e

    enough-analyst-54434

    11/02/2015, 8:29 PM
    Not sure what they pay for
  • e

    enough-analyst-54434

    11/02/2015, 8:29 PM
    I'll file an issue with the 2 flakes I captured which both present differently - I grabbed the full logs
  • b

    bright-dinner-81701

    11/02/2015, 8:29 PM
    im not sure how to fix it tho
  • b

    bright-dinner-81701

    11/02/2015, 8:29 PM
    i mean, if any npm install can fail
  • e

    enough-analyst-54434

    11/02/2015, 8:29 PM
    Well, cache
  • e

    enough-analyst-54434

    11/02/2015, 8:30 PM
    We do this for jvm resolves
  • e

    enough-analyst-54434

    11/02/2015, 8:30 PM
    Its slow, but at least the data transfer to-from cache is free and all aws internal
  • e

    enough-analyst-54434

    11/02/2015, 8:30 PM
    We're only subject to flake on changes to examples or intg tests
  • b

    bright-dinner-81701

    11/02/2015, 8:31 PM
    well we cant cache the remote targets individually … you mean cache the resolved example targets?
  • b

    bright-dinner-81701

    11/02/2015, 8:31 PM
    i thought it did that automatically
  • e

    enough-analyst-54434

    11/02/2015, 8:31 PM
    Travis caches nothing you don't ask for
  • b

    bright-dinner-81701

    11/02/2015, 8:32 PM
    ok
  • e

    enough-analyst-54434

    11/02/2015, 8:32 PM
  • e

    enough-analyst-54434

    11/02/2015, 8:32 PM
    The ~/.ivy2 keeps us from fetching from maven-central, an old source of flake
  • e

    enough-analyst-54434

    11/02/2015, 8:32 PM
    We need similar for npm
  • b

    bright-dinner-81701

    11/02/2015, 8:33 PM
    ah ok
  • e

    enough-analyst-54434

    11/02/2015, 8:33 PM
    Its a crappy solution - granted - the ~/ivy2 is ~300MB just like npm
  • e

    enough-analyst-54434

    11/02/2015, 8:33 PM
    But its stable
  • b

    bright-dinner-81701

    11/02/2015, 8:33 PM
    ok yeah it might just be adding $HOME/npm to that list
  • b

    bright-dinner-81701

    11/02/2015, 8:34 PM
    well, maybe that plus telling pants to instruct NPM to use $HOME/.npm/pants instead of $HOME/.npm
  • e

    enough-analyst-54434

    11/02/2015, 8:34 PM
    On ci we get a blank slate
1...658659660...2420Latest