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

    witty-crayon-22786

    01/31/2017, 9:05 PM
    the config that pants itself uses in CI might be helpful https://github.com/pantsbuild/pants/blob/master/pants.travis-ci.ini
  • h

    high-advantage-47550

    01/31/2017, 9:06 PM
    ah, thanks
  • w

    witty-crayon-22786

    01/31/2017, 9:06 PM
    twitter's is much more complicated, but we also use a combination of jenkins and other stuff
  • w

    witty-crayon-22786

    01/31/2017, 9:06 PM
    a summary is: we turn off nailgun for almost everything except zinc
  • w

    witty-crayon-22786

    01/31/2017, 9:07 PM
    we set up remote buildcaching
  • w

    witty-crayon-22786

    01/31/2017, 9:08 PM
    we completely disable incremental builds (which will cause more to be written to the cache, but can make CI slower)
  • w

    witty-crayon-22786

    01/31/2017, 9:08 PM
    ...and that's basically it.
  • h

    high-advantage-47550

    01/31/2017, 9:08 PM
    that helps, thank you!
  • w

    witty-crayon-22786

    01/31/2017, 9:10 PM
    Copy code
    $ ./pants help-advanced compile.zinc | grep -A3 'zinc-incremental'

--[no-]compile-zinc-incremental (default: True)
    When set, zinc will use sub-target incremental compilation, which
    dramatically improves compile performance while changing large targets. When
    unset, changed targets will be compiled with an empty output directory, as
--
--[no-]compile-zinc-incremental-caching (default: False)
    When set, the results of incremental compiles will be written to the cache.
    This is unset by default, because it is generally a good precaution to cache
    only clean/cold builds.
  • w

    witty-crayon-22786

    01/31/2017, 9:11 PM
    the default for caching incremental is false, because sbt/zinc are full of bugs in this area.
  • w

    witty-crayon-22786

    01/31/2017, 9:11 PM
    so turning off incrementalism allows more to be written to the cache.
  • h

    high-advantage-47550

    01/31/2017, 9:14 PM
    we were seeing a 
    ./pants compile //::
    taking about 1 hour for ~20 targets (which was way slower than sbt) — to be fair, this was when one of the 
    -util
    -esque projects changed, so most things had to be recompiled
  • h

    high-advantage-47550

    01/31/2017, 9:14 PM
    but seems like it should be faster, ill mess around some more
  • b

    billowy-minister-38128

    01/31/2017, 9:41 PM
    reviving a somewhat stale topic here, i’m trying to work through the proper way to use 
    test-changed
    on a CI machine. the specific issue i’m confronted with is the behavior of pull requests in TeamCity - and this is more a TC question than a pants question: what is the best way to get the required changed-diffspec? (since i have a trigger on 
    +:refs/pull/*/head
    it seems difficult to retrieve.
  • w

    witty-crayon-22786

    01/31/2017, 9:48 PM
    the right way to use it is: don't use it 😃
  • w

    witty-crayon-22786

    01/31/2017, 9:49 PM
    Copy code
    for i in `./pants changed --include-dependees | xargs ./pants minimize`; do ./pants test target; done
  • w

    witty-crayon-22786

    01/31/2017, 9:49 PM
    (basically)
  • w

    witty-crayon-22786

    01/31/2017, 9:50 PM
    to your diffspec question: if you branch is based on master, then the diffspec would be with master
  • w

    witty-crayon-22786

    01/31/2017, 9:50 PM
    in which case, 
    ./pants changed --parent=master --include-dependees
    should be sufficient.
  • b

    billowy-minister-38128

    01/31/2017, 9:56 PM
    cool, i wold prefer to just be able to use branch name. this would probably allow me to capture the branch picked up by 
    +:refs/pull/*/head
    and pass as a param to my build script… looking into that. thank you.
  • b

    billowy-minister-38128

    01/31/2017, 9:58 PM
    i can pick it out with 
    teamcity.build.vcs.branch.<VCS root ID>
  • b

    billowy-minister-38128

    01/31/2017, 10:01 PM
    and yes, i don’t actually use test-changed, i have some very similar to your for loop.
  • w

    witty-crayon-22786

    01/31/2017, 10:02 PM
    one last thing: we're going to deprecate 
    ./pants changed --parent=master
    soon in favor of 
    ./pants --changed-parent=master list
  • w

    witty-crayon-22786

    01/31/2017, 10:03 PM
    and 
    ./pants test-changed --parent=master
    in favor of 
    ./pants --changed-parent=master test
  • w

    witty-crayon-22786

    01/31/2017, 10:03 PM
    (basically: all goals now** support 
    --changed
    args, so no more need for specific goals. thanks @rough-minister-58256 )
    bowtie 1
  • b

    billowy-minister-38128

    01/31/2017, 10:08 PM
    nice.
  • r

    rough-minister-58256

    01/31/2017, 10:40 PM
    with the added travis capacity, would it makes sense to kill https://github.com/pantsbuild-osx now?
  • f

    faint-grass-11531

    01/31/2017, 11:00 PM
    hi everyone, @microscopic-intern-4465 and I are trying to port a project to Pants at Pinterest. We're running into this gnarly error on `./pants binary`:
    Copy code
    Exception message: Could not satisfy all requirements for auth-client==0.2.3:
    auth-client==0.2.3
    why is it that auth-client==0.2.3 does not satisfy the requirement that auth-client==0.2.3? would appreciate any help we can get!! we've been staring at this for a while.
  • f

    faint-grass-11531

    01/31/2017, 11:01 PM
    We've seen this error before when we depend on, say, boto, when another package we depend on also depends on boto. In that situation the solution was to remove the boto dependency from our BUILD file. However in this case none of the other packages explicitly require auth-client, so we're stumped
  • r

    rough-minister-58256

    01/31/2017, 11:02 PM
    that’s a resolver error.. so somewhere - direct or transitively - in your python dependencies this is being depended on and can’t be found in your index/find-links configuration
1...161616171618...2420Latest