https://pantsbuild.org/ logo
Powered by
#development
  • b

    billowy-beach-68498

    07/07/2016, 9:34 PM
    do you know where does this conversion happen? (from target adaptor to target)
  • b

    billowy-beach-68498

    07/07/2016, 9:34 PM
    we don’t store Target in graph? do we store target adaptor in the graph instead?
    👍 1
  • r

    rough-minister-58256

    07/07/2016, 9:34 PM
    most likely in 
    LegacyBuildGraph
  • w

    witty-crayon-22786

    07/07/2016, 9:35 PM
  • w

    witty-crayon-22786

    07/07/2016, 9:35 PM
    yes
  • b

    billowy-beach-68498

    07/07/2016, 9:35 PM
    ah ok! thank you!
  • r

    rough-minister-58256

    07/07/2016, 10:03 PM
    is there anything else that sources-as-a-product nets us beyond unblocking codegen in the new engine?
  • r

    rough-minister-58256

    07/07/2016, 10:04 PM
    (iiuc, the idea behind it is to unblock codegen.next that wouldn’t require graph mutation)
  • w

    witty-crayon-22786

    07/07/2016, 10:04 PM
    @rough-minister-58256: dependency inference
  • r

    rough-minister-58256

    07/07/2016, 10:04 PM
    oic
  • w

    witty-crayon-22786

    07/07/2016, 10:05 PM
    the real important part of "sources as a product" is typesafety
  • w

    witty-crayon-22786

    07/07/2016, 10:05 PM
    namely, there are ScalaSources, PythonSources, etc
  • w

    witty-crayon-22786

    07/07/2016, 10:05 PM
    so if i have ThriftSources, and i want PythonSources, voila... a particular task can do that
  • r

    rough-minister-58256

    07/07/2016, 10:05 PM
    makes sense
  • r

    rough-minister-58256

    07/07/2016, 10:06 PM
    and fingerprinting of products vs targets is the way we effectively cache that?
  • w

    witty-crayon-22786

    07/07/2016, 10:07 PM
    well, fingerprinting of "targets" as a whole has always been weird
  • w

    witty-crayon-22786

    07/07/2016, 10:07 PM
    the reason we have different FingerprintStrategy classes is because no consumer wants "the whole target"
  • w

    witty-crayon-22786

    07/07/2016, 10:08 PM
    they just want bits of it: the Sources (product 1), the PythonRequirements (product 2), etc
  • r

    rough-minister-58256

    07/07/2016, 10:09 PM
    ah
  • w

    witty-crayon-22786

    07/07/2016, 10:09 PM
    @rough-minister-58256: this is demonstrated in the engine experiments
  • w

    witty-crayon-22786

    07/07/2016, 10:09 PM
    Target is just a collection of products
  • w

    witty-crayon-22786

    07/07/2016, 10:09 PM
    you don't have Target types anymore... just products
  • r

    rough-minister-58256

    07/07/2016, 10:11 PM
    trying to understand why we’d do this in the context of tracking towards the OKR
  • r

    rough-minister-58256

    07/07/2016, 10:11 PM
    like, what do we effectively gain from focusing on sources as a product?
  • w

    witty-crayon-22786

    07/07/2016, 10:12 PM
    it unblocks codegen
  • w

    witty-crayon-22786

    07/07/2016, 10:12 PM
    and codegen is the first thing that we could actually usefully remote
  • r

    rough-minister-58256

    07/07/2016, 10:13 PM
    unblocks it how, specifically?
  • r

    rough-minister-58256

    07/07/2016, 10:13 PM
    the general sense I’m getting is that it’d provide superior “noop" builds because it’s a more effective way to reference the granular product work done - yeah?
  • w

    witty-crayon-22786

    07/07/2016, 10:13 PM
    @rough-minister-58256: codegen is currently accomplished via synthetic targets
  • r

    rough-minister-58256

    07/07/2016, 10:13 PM
    right
1...112113114...906Latest