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

    witty-crayon-22786

    01/28/2016, 7:37 PM
    so it would have to be a pretty darn good version
  • b

    bored-art-40741

    01/28/2016, 7:37 PM
    Everyone will have to be sure to include this in their ivysettings.xml
  • b

    bored-art-40741

    01/28/2016, 7:37 PM
    Copy code
    <caches
    defaultCacheDir="${ivy.cache.repository.dir}"
    resolutionCacheDir="${ivy.cache.resolution.dir}"
    lockStrategy="artifact-lock-nio"
  />
  • b

    bored-art-40741

    01/28/2016, 7:38 PM
    Nothing more or less
  • b

    bored-art-40741

    01/28/2016, 7:38 PM
    Otherwise pants won't be able to find the generated ivy report
  • b

    bored-art-40741

    01/28/2016, 7:38 PM
    I wish I could just set these from the cmdline, but I can't
  • w

    witty-crayon-22786

    01/28/2016, 7:38 PM
    so you would ship a default, but if you wanted to override you'd have to ensure that you added that setting?
  • b

    bored-art-40741

    01/28/2016, 7:39 PM
    Yeah
  • w

    witty-crayon-22786

    01/28/2016, 7:39 PM
    the primary reason people override is to add more repos
  • b

    bored-art-40741

    01/28/2016, 7:39 PM
    We've talked about just pull the entirety of ivysettings.xml into config too, right
  • b

    bored-art-40741

    01/28/2016, 7:40 PM
    You could define everything in pants.ini and we'll just always generate the config for you on the fly
  • w

    witty-crayon-22786

    01/28/2016, 7:40 PM
    i think it gets much, much easier to do when we're not doing resolves in the critical path anymore
  • b

    bored-art-40741

    01/28/2016, 7:41 PM
    Another thought: ship a default one, but if a custom one is provided, programmatically inject this tag into the XML
  • b

    bored-art-40741

    01/28/2016, 7:41 PM
    It's trivial to parse the custom one and inject one element, then dump that file back out to a tmp
  • w

    witty-crayon-22786

    01/28/2016, 7:41 PM
    as long as there is a way to disable it, that might work.
  • w

    witty-crayon-22786

    01/28/2016, 7:41 PM
    alternatively, would it be possible to just fail noisily if it isn't present?
  • b

    bored-art-40741

    01/28/2016, 7:41 PM
    I can't see how we'd get away with disabling
  • w

    witty-crayon-22786

    01/28/2016, 7:42 PM
    if somebody has already included it, then they might disable the injection
  • f

    fancy-queen-20734

    01/28/2016, 7:42 PM
    @jolly-chef-92794: by test-jar, do you mean the ext is test-jar?
  • j

    jolly-chef-92794

    01/28/2016, 7:42 PM
    Just the type
  • f

    fancy-queen-20734

    01/28/2016, 7:42 PM
    ah, so the problem is that type isn’t part of the key
  • j

    jolly-chef-92794

    01/28/2016, 7:42 PM
    Though we map ext -> type
  • j

    jolly-chef-92794

    01/28/2016, 7:43 PM
    Kind of.. we currently use the “ext” for the type
  • j

    jolly-chef-92794

    01/28/2016, 7:43 PM
  • j

    jolly-chef-92794

    01/28/2016, 7:43 PM
    But that only applies if there is more than one artifact being fetched
  • j

    jolly-chef-92794

    01/28/2016, 7:44 PM
    Here we explicitly want the type to be ‘jar’
  • j

    jolly-chef-92794

    01/28/2016, 7:44 PM
    It seems like ivy is fetching arbitrarily otherwise, or based on some subtle transitive dependency ordering shenanigans
  • j

    jolly-chef-92794

    01/28/2016, 7:44 PM
    hadoop.png
  • j

    jolly-chef-92794

    01/28/2016, 7:45 PM
    Here we want the 
    .jar
    , but we’re getting the 
    -tests.jar
    in some instances
  • b

    bored-art-40741

    01/28/2016, 7:45 PM
    @witty-crayon-22786: If the injection is disabled then we have no handle on how to look up the resolution reports
1...934935936...2420Latest