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

    mammoth-autumn-2551

    08/04/2015, 2:10 AM
    we have run into the python thrift issues I think you're alluding to in the past
  • e

    enough-analyst-54434

    08/04/2015, 2:10 AM
    OK - so you'r 0.9.2 python (default), but 0.5.0-finagle jvm - right?
  • m

    mammoth-autumn-2551

    08/04/2015, 2:10 AM
    but our code layout is somehow incompatible with the switch to simple code gen
  • e

    enough-analyst-54434

    08/04/2015, 2:10 AM
    aha
  • m

    mammoth-autumn-2551

    08/04/2015, 2:10 AM
    nah we're using 0.5.0-finagle for both
  • e

    enough-analyst-54434

    08/04/2015, 2:10 AM
    right
  • e

    enough-analyst-54434

    08/04/2015, 2:11 AM
    So what's the layout oddity?
  • m

    mammoth-autumn-2551

    08/04/2015, 2:11 AM
    I'm not sure exactly, but I think it has to do with using thrift 'includes'
  • m

    mammoth-autumn-2551

    08/04/2015, 2:11 AM
    hence why I'm trying to build a simple example that works in 0.0.39 but fails in 0.0.40
  • e

    enough-analyst-54434

    08/04/2015, 2:11 AM
    includes _msut_have corresponding dependencies iff the includes come from other targets
  • e

    enough-analyst-54434

    08/04/2015, 2:11 AM
    the old method of compile would mask missing deps
  • e

    enough-analyst-54434

    08/04/2015, 2:12 AM
    hit or miss?
  • m

    mammoth-autumn-2551

    08/04/2015, 2:13 AM
    well I think miss
  • e

    enough-analyst-54434

    08/04/2015, 2:13 AM
    k
  • m

    mammoth-autumn-2551

    08/04/2015, 2:14 AM
    because in the last 6 months we got fed up with keeping thrift dependencies up to date and turned them into a single target with `sources=rglobs('**/*.thrift') or something like that
  • e

    enough-analyst-54434

    08/04/2015, 2:14 AM
    k
  • e

    enough-analyst-54434

    08/04/2015, 2:14 AM
    If you have just one ... that's helpful actually
  • e

    enough-analyst-54434

    08/04/2015, 2:15 AM
    let me look at the code with that in mind
  • e

    enough-analyst-54434

    08/04/2015, 2:15 AM
    so - what's the namespace situation like across all those rglobs?
  • e

    enough-analyst-54434

    08/04/2015, 2:15 AM
    flat, varied?
  • e

    enough-analyst-54434

    08/04/2015, 2:16 AM
    are you basically 
    namespace java ...
    parallel with a treed dir structure?
  • e

    enough-analyst-54434

    08/04/2015, 2:17 AM
    pre-emptive info:
  • e

    enough-analyst-54434

    08/04/2015, 2:17 AM
    In simplecodegen - for a single target - say a single rglobbed target, we:
  • e

    enough-analyst-54434

    08/04/2015, 2:17 AM
    for each source - in no particular order, call thrift with 1 source at a time and proper -I (if you have 1 thrift source root, its just that)
  • e

    enough-analyst-54434

    08/04/2015, 2:18 AM
    these gens all go to a single chroot
  • e

    enough-analyst-54434

    08/04/2015, 2:18 AM
    then when we're done, the baseclass walks the chroot, gathers all java sources, and forms 1 synthetic target to inject in the graph for compile
  • m

    mammoth-autumn-2551

    08/04/2015, 2:19 AM
    "are you basically 
    namespace java ...
    parallel with a treed dir structure? " -- yep pretty much
  • e

    enough-analyst-54434

    08/04/2015, 2:19 AM
    if somehow one thrift source overwrte another in the linear gen phase, that would obviously cause issues
  • e

    enough-analyst-54434

    08/04/2015, 2:19 AM
    so that would be 2 different thrift source files with the same namespace
  • e

    enough-analyst-54434

    08/04/2015, 2:19 AM
    ... I think - mod handwave'
1...132133134...2420Latest