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

    fancy-queen-20734

    10/26/2017, 4:52 PM
    well, if a non-codegen target says it exports a codegen dependency, should it also export the synthetic target generated from the codegen dep, or not?
  • w

    witty-crayon-22786

    10/26/2017, 4:52 PM
    ...i think i see now.
  • w

    witty-crayon-22786

    10/26/2017, 4:52 PM
    yes, it should.
  • w

    witty-crayon-22786

    10/26/2017, 4:53 PM
    @fancy-queen-20734: but i think you can rely on export transitivity to solve/work-around that... if you inject the synthetic target as an export of the original target
  • w

    witty-crayon-22786

    10/26/2017, 4:54 PM
    ...doh. that's not a thing either.
  • f

    fancy-queen-20734

    10/26/2017, 4:54 PM
    yeah, but then I need …
  • f

    fancy-queen-20734

    10/26/2017, 4:54 PM
    yeah, you see it now
  • w

    witty-crayon-22786

    10/26/2017, 4:54 PM
    blargh.
  • f

    fancy-queen-20734

    10/26/2017, 4:54 PM
    I find myself wishing the buildgraph had edge properties
  • w

    witty-crayon-22786

    10/26/2017, 4:54 PM
    mutability is a such a pain.
  • w

    witty-crayon-22786

    10/26/2017, 4:55 PM
    @fancy-queen-20734: Variants.
  • f

    fancy-queen-20734

    10/26/2017, 4:55 PM
    then scopes, exports, and a few other things could be more cleanly represented, at the cost of making other things way more complex
  • w

    witty-crayon-22786

    10/26/2017, 4:55 PM
    but remember that these are mutable build graph problems.
  • f

    fancy-queen-20734

    10/26/2017, 4:55 PM
    yeah
  • w

    witty-crayon-22786

    10/26/2017, 4:56 PM
    the graph as it exists is problematic.
  • f

    fancy-queen-20734

    10/26/2017, 4:56 PM
    that’s part of the “what if we could split creating targets from the code gen” bit. With v2 it’s less complex because you just delay resolving the compile deps until all of the code gen is done
  • f

    fancy-queen-20734

    10/26/2017, 4:57 PM
    s/complex/mutation dependent
  • w

    witty-crayon-22786

    10/26/2017, 4:57 PM
    yea.
  • w

    witty-crayon-22786

    10/26/2017, 4:57 PM
    well... no
  • w

    witty-crayon-22786

    10/26/2017, 4:57 PM
    @fancy-queen-20734: no delay... it will all happen concurrently =P
  • f

    fancy-queen-20734

    10/26/2017, 4:57 PM
    to be more pedantic
  • w

    witty-crayon-22786

    10/26/2017, 4:57 PM
    but i get what you're saying
  • f

    fancy-queen-20734

    10/26/2017, 4:58 PM
    resolving deps becomes lazy
  • f

    fancy-queen-20734

    10/26/2017, 4:58 PM
    or at least lazier
  • w

    witty-crayon-22786

    10/26/2017, 4:58 PM
    once we don't build the BuildGraph, it should be fully lazy, yea (edit: BuildGraph)
  • w

    witty-crayon-22786

    10/26/2017, 4:59 PM
    @fancy-queen-20734: anywho. now that i understand: are there any more changes you're planning to make, or should i re-review
  • f

    fancy-queen-20734

    10/26/2017, 5:00 PM
    I’m going to undo the change to target.py, because we need it for the non-codegen -> codegen case
  • f

    fancy-queen-20734

    10/26/2017, 5:00 PM
    and I could remove the bits from codegen that tried to make it unnecessary
  • f

    fancy-queen-20734

    10/26/2017, 5:01 PM
    I think the extraction of calculating extra deps / exports into a separate step is still reasonable though
  • f

    fancy-queen-20734

    10/26/2017, 5:01 PM
    because it moves towards a separation of concerns
1...184618471848...2420Latest