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

    rough-minister-58256

    06/01/2018, 10:12 PM
    ah right
  • r

    rough-minister-58256

    06/01/2018, 10:14 PM
  • w

    witty-crayon-22786

    06/01/2018, 10:29 PM
    thanks!
  • r

    rough-minister-58256

    06/01/2018, 10:29 PM
    👍
  • r

    rough-minister-58256

    06/01/2018, 11:15 PM
    btw @enough-analyst-54434 how are you doing? back home yet?
  • e

    enough-analyst-54434

    06/01/2018, 11:18 PM
    Making progress. Got back home Monday night. I've had a few local Bozeman Dr visits to setup care / recovery here.
  • r

    rough-minister-58256

    06/01/2018, 11:19 PM
    nice, glad to hear that
  • a

    average-vr-56795

    06/04/2018, 1:08 PM
    @rough-minister-58256 Thanks for the investigation and fix!
  • r

    rough-minister-58256

    06/04/2018, 4:24 PM
    👍
  • a

    average-vr-56795

    06/04/2018, 4:28 PM
    @witty-crayon-22786 Currently the 
    work_dir
    in the engine is unused. Is there something we should be using it for?
  • a

    average-vr-56795

    06/04/2018, 4:29 PM
    It used to be used for snapshots, but those are now elsewhere
  • a

    average-vr-56795

    06/04/2018, 4:29 PM
    We make tempdirs for process execution, but those are just in $TMP
  • r

    rough-minister-58256

    06/04/2018, 5:33 PM
    my $0.02: we should probably avoid direct use of $TMP in favor of keying somewhere in the work_dir, so that e.g. 
    clean-all
    is more effective if we’re utilizing a ton of disk space
  • w

    witty-crayon-22786

    06/04/2018, 5:56 PM
    they should be eagerly cleaned up, so the long term disk usage won't be a thing... but yea, in general, moving toward a world where you can flip one switch to move the whole 
    --workdir
    to a ramdisk is good... avoiding a proliferation of flags would be ideal
  • w

    witty-crayon-22786

    06/04/2018, 5:57 PM
    @average-vr-56795: tempdirs somewhere under the workdir would be reasonable. and based on the above, i think leaving the option in place and encouraging folks to use it is reasonable
  • a

    average-vr-56795

    06/04/2018, 6:03 PM
    Sounds good 🙂 Thanks!
  • w

    witty-crayon-22786

    06/04/2018, 8:14 PM
    alright, the merge is going to be a pain in the butt, but i think i have to dive into some pre-cleanup in rule_graph before https://github.com/pantsbuild/pants/issues/5788
  • a

    average-vr-56795

    06/05/2018, 1:27 PM
    @witty-crayon-22786 I imagine we may need to talk about https://github.com/pantsbuild/pants/pull/5908 after you’ve had a read of the description 🙂
  • w

    witty-crayon-22786

    06/05/2018, 4:15 PM
    iiinteresting.
  • w

    witty-crayon-22786

    06/05/2018, 4:17 PM
    from the description it's not completely clear what the motivation was
  • a

    average-vr-56795

    06/05/2018, 4:17 PM
    Yeah… Maybe controversial… But if we’re going to shove source discovery through the engine, it feels like we should do so in a general way. The implicit promotion of the 
    kwarg
    to the 
    Target
    is a little sketchy… But feels like it’s something we’re going to need to do at some point
  • a

    average-vr-56795

    06/05/2018, 4:18 PM
    2nd paragraph is the motivation. Unifying all source discovery to be populating `EagerFilesetWithSpec`s in the engine.
  • a

    average-vr-56795

    06/05/2018, 4:18 PM
    (Which is a step towards being able to use `Snapshot`s of 
    sources
    from any particular 
    Target
    )
  • w

    witty-crayon-22786

    06/05/2018, 4:19 PM
    by having fewer 
    Adaptor
    types i guess? Adaptors are/were supposed to be a stepping stone toward some more solid target API
  • w

    witty-crayon-22786

    06/05/2018, 4:19 PM
    unfortunately there hasn't been much design there yet: https://github.com/pantsbuild/pants/issues/4535
  • w

    witty-crayon-22786

    06/05/2018, 4:20 PM
    but... hm
  • a

    average-vr-56795

    06/05/2018, 4:20 PM
    In my ideal world, we’d be describing the arguments to a 
    Target
    with types, rather than just as names in kwargs, and the engine would know how to instantiate types
  • w

    witty-crayon-22786

    06/05/2018, 4:20 PM
    yes, absolutely.
  • a

    average-vr-56795

    06/05/2018, 4:20 PM
    In the mean time, we already do the magic for `sources`; applying it to 
    source
    too doesn’t feel crazy…
  • w

    witty-crayon-22786

    06/05/2018, 4:21 PM
    certainly a typed argument / typed field world would support plural and singular sources args... but let me read through it. will comment today
1...508509510...906Latest