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

    witty-crayon-22786

    10/03/2019, 9:43 PM
    hermeticity
    when referring to an environment in general means "clean" or "isolated"
  • w

    witty-crayon-22786

    10/03/2019, 9:43 PM
    "bit for bit reproducible" is just 
    reproducible
    , in general
  • w

    witty-crayon-22786

    10/03/2019, 9:44 PM
    hermeticity can contribute to things being reproducible, but putting a non-deterministic thing inside of a hermetic box will still cause it to not be reproducible
  • a

    aloof-angle-91616

    10/03/2019, 9:44 PM
    i have changed my message above to refer to reproducibility instead
  • h

    hundreds-breakfast-49010

    10/03/2019, 9:45 PM
    ok so we distinguish hermeticity and reproducibility
  • w

    witty-crayon-22786

    10/03/2019, 9:45 PM
    pex accesses the network, and thats a violation for most definitions of "hermeticity"... in practice though, network access is necessary in a bunch of cases.
  • w

    witty-crayon-22786

    10/03/2019, 9:46 PM
    when you think of hermeticity in terms of isolation, there are sortof endless degrees of isolation... no network, no filesystem, etc
  • w

    witty-crayon-22786

    10/03/2019, 9:47 PM
    docker gets you pretty close if you don't open up network access
  • w

    witty-crayon-22786

    10/03/2019, 9:47 PM
    (...but we do in RBE)
  • a

    aloof-angle-91616

    10/03/2019, 9:47 PM
    if the requirements resolution was separated from the rest of pex building, for example, then at least the rest of the pex build process can be hermetic
  • w

    witty-crayon-22786

    10/03/2019, 9:50 PM
    yea
  • w

    witty-crayon-22786

    10/03/2019, 9:50 PM
    it soooortof is? in that we build the requirements pex separately
  • w

    witty-crayon-22786

    10/03/2019, 9:50 PM
    so you could turn off network access for the second half, i think.
  • r

    red-balloon-89377

    10/04/2019, 1:41 PM
    I’m starting to doubt the value of making spawing nailguns asynchronous in the first PR for nailgunnable EPRs…
  • r

    red-balloon-89377

    10/04/2019, 1:41 PM
    It’s becoming a bit of a massive pain, given the APIs
  • a

    average-vr-56795

    10/04/2019, 1:58 PM
    Feels pretty happily deferrable 🙂
  • a

    average-vr-56795

    10/04/2019, 1:58 PM
    You can spawn functions on the `Executor`’s 
    io_pool
    which will avoid blocking the general future executor
  • r

    red-balloon-89377

    10/04/2019, 2:06 PM
    I think the main issue here is that expressing “I want to block until the process has had one line of output, then run some stuff while the process is still alive” is very complicated, there’s not a clear way to sequence “I want to run something when the process has started, but before it finishes” with “I want to run something when exactly one line of output is available”.
  • r

    red-balloon-89377

    10/04/2019, 2:06 PM
    So yeah, I’m now trying to integrate it with blocking spawning of nailguns
  • r

    red-balloon-89377

    10/04/2019, 2:06 PM
    and then running client requests async
  • a

    average-vr-56795

    10/04/2019, 2:11 PM
    Sounds good 🙂
  • a

    average-vr-56795

    10/04/2019, 2:32 PM
    @fancy-motherboard-24956 Have fun with those failing unit tests which may be formatting-sensitive…
    😄 2
    f
    h
    • 3
    • 4
  • a

    aloof-angle-91616

    10/04/2019, 4:14 PM
    https://twitter.com/mountain_ghosts/status/1180133138200178688?s=21 interesting description of why i like rust that i was paraphrasing to @hundreds-father-404 last night
    👍 1
    • 1
    • 1
  • a

    aloof-angle-91616

    10/04/2019, 4:14 PM
    (this is from a jealous lover of js)
  • a

    aloof-angle-91616

    10/04/2019, 4:16 PM
    and of course, the lazy semantics could absolutely be implemented in js if you just allowed a 
    .collect()
    method instead of automatically collecting, it’s just a statement about the stdlib & culture of the languages in general
  • h

    hundreds-breakfast-49010

    10/04/2019, 6:15 PM
    @aloof-angle-91616 I'm not 100% sure what this is getting at
  • a

    aloof-angle-91616

    10/04/2019, 6:16 PM
    just that it's cool that rust has really ergonomic iterator methods that expose the same functional interface as in more dynamic languages, but doesn't require you to make any performance tradeoffs for those ergonomics
  • h

    hundreds-breakfast-49010

    10/04/2019, 6:16 PM
    in javascript the first 
    filter
    will return a (potentially-smaller) array, and then the second 
    filter
    will iterate over that smaller array
  • h

    hundreds-breakfast-49010

    10/04/2019, 6:17 PM
    and in rust the first 
    filter
    would create an iterator object that the second 
    filter
    would also iterate over, but I don't think it reduces the amount of total iteration
  • a

    aloof-angle-91616

    10/04/2019, 6:17 PM
    in rust, the semantics are more similar to e.g. 
    a.filter(x => f(x) && g(x))
    , where the iteration only occurs once
1...658659660...906Latest