https://pantsbuild.org/ logo
Powered by
#pex
  • r

    rough-minister-58256

    07/07/2016, 12:14 AM
  • r

    rough-minister-58256

    07/07/2016, 12:15 AM
    guessing these failures are happening on travis?
  • f

    faint-holiday-3863

    07/07/2016, 2:06 AM
    nope, on my laptop believe it or not
  • f

    faint-holiday-3863

    07/07/2016, 2:07 AM
    That’s exactly the bit of code I’m working with, too. Trying to actually get wheel working in there.
  • f

    faint-holiday-3863

    07/07/2016, 2:07 AM
  • r

    rough-minister-58256

    07/07/2016, 5:27 AM
    bonus points if you kill that entire manual interpreter setup bit.. pex should be able to handle it natively, no?
  • f

    faint-holiday-3863

    07/07/2016, 9:10 PM
    In theory
  • f

    faint-holiday-3863

    07/07/2016, 9:12 PM
    ideally bazel would be supplying the entire python distribution at runtime
  • f

    faint-holiday-3863

    07/07/2016, 9:13 PM
    I guess that doesn’t help with when you distribute and run the pex, unless you install the special python distribution with it
  • f

    faint-holiday-3863

    07/07/2016, 9:13 PM
    which is what google does
  • r

    rough-minister-58256

    07/07/2016, 9:15 PM
    yeah - their pex equivalent (par) embeds a python as well.
  • f

    faint-holiday-3863

    07/07/2016, 9:16 PM
    I’m pretty sure it does not. There’s a standard runtime python in /usr/grte/v<something>/stuff
  • f

    faint-holiday-3863

    07/07/2016, 9:16 PM
    Of course things could have changed since I left, it’s been a few years.
  • f

    faint-holiday-3863

    07/07/2016, 9:17 PM
    iirc the shebang lines end up being like 
    #!/usr/grte/v2/bin/python2.7
  • r

    rough-minister-58256

    07/07/2016, 9:17 PM
    perhaps just for the initial bootstrap? all the literature I’ve found on the topic and anecdotal stories seem to point to a fully embedded python distribution
  • f

    faint-holiday-3863

    07/07/2016, 9:18 PM
    It’s possible that my memory is just wrong.
  • r

    rough-minister-58256

    07/07/2016, 9:19 PM
    seems a shame they haven’t open sourced that
  • f

    faint-holiday-3863

    07/07/2016, 9:19 PM
    the actual implementation of par is a bit of a disaster internally 😛
  • r

    rough-minister-58256

    07/07/2016, 9:19 PM
    oh is it
  • f

    faint-holiday-3863

    07/07/2016, 9:19 PM
    it just happens to work correctly in their environment
  • r

    rough-minister-58256

    07/07/2016, 9:19 PM
    heh
  • f

    faint-holiday-3863

    07/07/2016, 9:20 PM
    main() -> start() -> actual_start() -> real_actual_start() -> really_real_actual_start_shim_wrapper() -> __main__() -> app.run()
  • f

    faint-holiday-3863

    07/07/2016, 9:21 PM
    that kind of thing.
  • r

    rough-minister-58256

    07/07/2016, 9:21 PM
    heh nice
  • f

    faint-holiday-3863

    07/07/2016, 9:22 PM
    fundamentally it’s pretty similar to pex, but there’s nothing that tries to resolve requirements or embed eggs/wheels; just the startup wrapper.
  • f

    faint-holiday-3863

    07/07/2016, 9:22 PM
    which does various gymnastics to make gflags do the right thing, iirc
  • f

    faint-holiday-3863

    07/07/2016, 9:27 PM
    invites rwilliams@lyft.com to this slack
  • d

    delightful-student-56861

    07/15/2016, 8:05 PM
    @faint-holiday-3863: haha, I remember 
    really_real_actual_start_shim_wrapper
    ! What a nightmare
  • f

    faint-holiday-3863

    07/15/2016, 8:05 PM
    It was some hilarity in there, yeah
  • f

    faint-holiday-3863

    07/15/2016, 8:07 PM
    I understand a bit more of what goes on inside pex now, and the bazel pex wrapper
1...131415...74Latest