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

    rough-minister-58256

    06/06/2018, 5:41 PM
    i.e. “read this file without caching it and give me its content”
  • r

    rough-minister-58256

    06/06/2018, 5:42 PM
    then maybe those are read from disk always and anything relative to the buildroot would be cached?
  • w

    witty-crayon-22786

    06/06/2018, 5:42 PM
    transient requires a feature we don't have yet i think: https://github.com/pantsbuild/pants/issues/5788
  • r

    rough-minister-58256

    06/06/2018, 5:42 PM
    yeah, this would all be nascent
  • w

    witty-crayon-22786

    06/06/2018, 5:43 PM
    it's pretty likely that we'll end up adding an intrinsic for this, yea. but it would be a big chunk of feature work
  • r

    rough-minister-58256

    06/06/2018, 5:43 PM
    k
  • w

    witty-crayon-22786

    06/06/2018, 5:43 PM
    my understanding about this ticket though was that bootstrap was the only thing that needed to go outside of the buildroot
  • w

    witty-crayon-22786

    06/06/2018, 5:43 PM
    and that after that, it was reasonable to assume that everything specified in options would be located inside
  • w

    witty-crayon-22786

    06/06/2018, 5:44 PM
    if that's not the case, we might want to break out a ticket about actually adding an intrinsic for a "polled absolute path" or something
  • r

    rough-minister-58256

    06/06/2018, 5:44 PM
    yeah, I was trying to move bootstrap into a rule
  • w

    witty-crayon-22786

    06/06/2018, 5:44 PM
    yea, i wouldn't do that yet.
  • r

    rough-minister-58256

    06/06/2018, 5:44 PM
    k
  • w

    witty-crayon-22786

    06/06/2018, 5:45 PM
    for one thing, it needs to happen in the pantsd client anyway
  • w

    witty-crayon-22786

    06/06/2018, 5:45 PM
    where we don't have a warm daemon
  • w

    witty-crayon-22786

    06/06/2018, 5:45 PM
    so there is less advantage
  • r

    rough-minister-58256

    06/06/2018, 5:45 PM
    well, the core advantage I see is just that “rules don’t do I/O unless that’s cooperatively done with the engine”
  • w

    witty-crayon-22786

    06/06/2018, 5:47 PM
    @rough-minister-58256: i agree. but i think that applies to options parsing in general, but not to bootstrap (yet, because we don't have features in place needed to support it)
  • w

    witty-crayon-22786

    06/06/2018, 5:47 PM
    sorry, i thought i had put that in the ticket, but apparently i didn't =(
  • r

    rough-minister-58256

    06/06/2018, 5:47 PM
    In order to safely memoize this @rule (ie, use it in pantsd), we’d need to move that filesystem access to usage of PathGlob -> FileContents.
  • r

    rough-minister-58256

    06/06/2018, 5:47 PM
    was the tack I was on
  • w

    witty-crayon-22786

    06/06/2018, 5:48 PM
    sure, but for things like 
    file_option
    and 
    dir_option
  • r

    rough-minister-58256

    06/06/2018, 5:48 PM
    ah. heh
  • w

    witty-crayon-22786

    06/06/2018, 5:48 PM
    (...was where i was going with that. not enough explanation)
  • r

    rough-minister-58256

    06/06/2018, 5:48 PM
    k
  • r

    rough-minister-58256

    06/06/2018, 5:49 PM
    I’ll wrap this bootstrap retrofit up as-is because I think we can use it later then
  • w

    witty-crayon-22786

    06/06/2018, 5:49 PM
    if you think it would be worthwhile to dive into the design of the intrinsic for absolute file access, that's potentially a thing too
  • w

    witty-crayon-22786

    06/06/2018, 5:49 PM
    it just feels like a diversion right now
  • r

    rough-minister-58256

    06/06/2018, 5:49 PM
    yeah, I think we can punt for now and wire that up later
  • w

    witty-crayon-22786

    06/06/2018, 5:49 PM
    sorry for the miscommunication
  • r

    rough-minister-58256

    06/06/2018, 5:49 PM
    nbd
1...510511512...906Latest