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

    witty-crayon-22786

    12/11/2016, 2:31 AM
    which is easier, and not what buildgen is about, iirc
  • p

    purple-leather-29803

    12/11/2016, 2:31 AM
    hm ok i thought it was powering the “keep” shading rule that supposedly keeps in the JAR all classes that are actually reachable from things you’ve “kept"
  • w

    witty-crayon-22786

    12/11/2016, 2:31 AM
    post compile, there are at least 3 ways to get the dependency info... pants has one built in based on using the zinc analysis
  • w

    witty-crayon-22786

    12/11/2016, 2:32 AM
  • w

    witty-crayon-22786

    12/11/2016, 2:33 AM
    @purple-leather-29803 : oh, i see. yea, good point.
  • w

    witty-crayon-22786

    12/11/2016, 2:33 AM
    that's how the zinc analysis works for java as well (EDIT: inspecting the classfiles)
  • w

    witty-crayon-22786

    12/11/2016, 2:33 AM
    for scala, zinc uses a compiler plugin
  • p

    purple-leather-29803

    12/11/2016, 2:33 AM
    ah crazy, hadn’t realized that zinc is solving this problem too
  • w

    witty-crayon-22786

    12/11/2016, 2:33 AM
    but again: post compile
  • p

    purple-leather-29803

    12/11/2016, 2:34 AM
    because that is the level at which it is determining that a change to class X has or has not resulted in a need to recompile Y?
  • w

    witty-crayon-22786

    12/11/2016, 2:34 AM
    (for java...)
  • w

    witty-crayon-22786

    12/11/2016, 2:34 AM
    yea
  • w

    witty-crayon-22786

    12/11/2016, 2:34 AM
    it's building the sub-target graph
  • p

    purple-leather-29803

    12/11/2016, 2:34 AM
    you mean sub-pants-targets? as in it knows the graph at the class/method to class/method level?
  • w

    witty-crayon-22786

    12/11/2016, 2:34 AM
    yep, zinc does.
  • p

    purple-leather-29803

    12/11/2016, 2:35 AM
    that is pretty neat
  • w

    witty-crayon-22786

    12/11/2016, 2:35 AM
    and buggy, unfortunately =/
  • p

    purple-leather-29803

    12/11/2016, 2:35 AM
    lol, still?
  • w

    witty-crayon-22786

    12/11/2016, 2:35 AM
    the compiler keeps changing, as does zinc
  • w

    witty-crayon-22786

    12/11/2016, 2:36 AM
    our (twitter's) long term goal is to get the graph to the point where sub-target (edit: sub) incremental isn't necessary
  • w

    witty-crayon-22786

    12/11/2016, 2:36 AM
    and then drop zinc
  • p

    purple-leather-29803

    12/11/2016, 2:37 AM
    hm seems like it would always be better to know that info though, no?
  • w

    witty-crayon-22786

    12/11/2016, 2:37 AM
    if it were really well integrated with the compiler, maybe
  • p

    purple-leather-29803

    12/11/2016, 2:37 AM
    gotcha
  • p

    purple-leather-29803

    12/11/2016, 2:38 AM
    seems like in a perfect world you’d just declare your 3rd-party libs in one dir like today, and then the tooling would just figure out all the other targets+deps at the most-granular-possible level
  • w

    witty-crayon-22786

    12/11/2016, 2:41 AM
    yes, absolutely.
  • w

    witty-crayon-22786

    12/11/2016, 2:42 AM
    headed out for dinner... let's chat more. will take a look at your project
  • p

    purple-leather-29803

    12/11/2016, 2:42 AM
    whatever union of dep-resolution and compiler-integrations are in intellij these days feels somewhat close to this
  • p

    purple-leather-29803

    12/11/2016, 2:43 AM
    cool, thanks for all the info! the repo i posted earlier is not that interesting, quickest possible hack to get an inter-pkg dep graph for one of my projects before i attempted to factor out a package to a new repo.
  • w

    witty-crayon-22786

    12/11/2016, 2:43 AM
    yea... smaller scope (at least in our repo) though
1...157215731574...2420Latest