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

    witty-crayon-22786

    09/19/2019, 5:02 PM
    if we think that we're going to lean way in here (on spawning JVMs), that might be reasonable.
  • w

    witty-crayon-22786

    09/19/2019, 5:02 PM
    (as in, much further than we have currently planned, i think?)
  • a

    average-vr-56795

    09/19/2019, 5:03 PM
    The actual full request would look like:
    Copy code
    ExecuteProcessRequest(
  args = ["compile blah"],
  digest = blah,
  nailgun_execute_process_request = ExecuteProcessRequest(
    argv = ["-XmX10G", "org.pantsbuild.Zinc"],
    digest = blurgle,
  ),
)
    and the logic would only activate if 
    nailgun_execute_process_request
    is non-empty
  • w

    witty-crayon-22786

    09/19/2019, 5:03 PM
    mm, you lost me again. don't think you need the subrequest if you do https://pantsbuild.slack.com/archives/C0D7TNJHL/p1568912438098200?thread_ts=1568906384.076900&cid=C0D7TNJHL ? ... except maybe implicitly for a second while you lookup in your nailgun pool
  • a

    average-vr-56795

    09/19/2019, 5:03 PM
    You do - the digests are different
  • a

    average-vr-56795

    09/19/2019, 5:04 PM
    The outer EPR includes source files, the inner is just a classpath
  • a

    average-vr-56795

    09/19/2019, 5:04 PM
    (Which is important so we know when to restart the nailgun)
  • w

    witty-crayon-22786

    09/19/2019, 5:04 PM
    well, who is composing it? the user, or us?
  • w

    witty-crayon-22786

    09/19/2019, 5:04 PM
    (that's what this boils down to, i think)
  • a

    average-vr-56795

    09/19/2019, 5:05 PM
    Yeah, that’s fair. I don’t really mind whether we ask the user to do it with variables, or we do it for them. Having the user do it is a bit more in the spirit of “rules are just python” imo, and we can maybe offer a function in the standard library to nailgunify an EPR if we really want?
  • w

    witty-crayon-22786

    09/19/2019, 5:06 PM
    anyway, this doesn't need to be perfect in a first cut, my hope is just that we can lean more toward a Drip-py "awareness of the JVM args means much better UX" model
  • w

    witty-crayon-22786

    09/19/2019, 5:08 PM
    and if 90% of the way there (we only parse the 
    -cp
    arg to find your classpath, and strip anything before our "best guess at the main class"... in the first PR) is fairly easy to evolve into a 100% solution, then that would be lovely.
  • r

    red-balloon-89377

    09/19/2019, 5:08 PM
    Hmm In practice, Yi needed to pass the argfile to Drip via an env var, so I don’t think Drip got the magic working 100% either
  • w

    witty-crayon-22786

    09/19/2019, 5:09 PM
    OR, we lean the other way and say "we think that there are benefits from a UX perspective to putting the JVM front and center"
  • w

    witty-crayon-22786

    09/19/2019, 5:09 PM
    @red-balloon-89377: for which tool...?
  • r

    red-balloon-89377

    09/19/2019, 5:09 PM
    Drip
  • r

    red-balloon-89377

    09/19/2019, 5:09 PM
    Oh
  • w

    witty-crayon-22786

    09/19/2019, 5:09 PM
    which underlying tool?
  • r

    red-balloon-89377

    09/19/2019, 5:09 PM
    Zinc, sorry
  • w

    witty-crayon-22786

    09/19/2019, 5:10 PM
    thaaaat is very weird
  • w

    witty-crayon-22786

    09/19/2019, 5:10 PM
    and a good sign of potential perils. but i think that you are right about the main class
  • w

    witty-crayon-22786

    09/19/2019, 5:10 PM
    @red-balloon-89377: re the main class: from experimenting with:
    Copy code
    $ java -cp . Thing -- hello!
Hello world! 2
$ java -Dsomething=something -cp . Thing -- hello!
Hello world! 2
$ java -cp . Thing -- -Dsomething=something hello!
Hello world! 3
$ java -cp . -Dsomething=somethin Thing -- hello!
Hello world! 2
  • w

    witty-crayon-22786

    09/19/2019, 5:12 PM
    (ignoring the 
    --
    for a second)
  • r

    red-balloon-89377

    09/19/2019, 5:12 PM
    😧
  • w

    witty-crayon-22786

    09/19/2019, 5:12 PM
    it passes everything after 
    Thing
    (the main class name here) to the program
  • r

    red-balloon-89377

    09/19/2019, 5:12 PM
    I couldn’t be that easy
  • r

    red-balloon-89377

    09/19/2019, 5:12 PM
    Oh, wait, Thing reads the envvar
  • w

    witty-crayon-22786

    09/19/2019, 5:12 PM
    well, as previously mentioned: the nice thing about this is that if we get it wrong, either the JVM will (definitely) fail or the program will (probably/maybe) fail
    ☝️ 1
  • r

    red-balloon-89377

    09/19/2019, 5:12 PM
    cool 🙂
  • w

    witty-crayon-22786

    09/19/2019, 5:13 PM
    @red-balloon-89377: no, it doesn't: it's just a random arg i threw in there
1...648649650...906Latest