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

    witty-crayon-22786

    07/10/2018, 9:20 PM
    @rough-minister-58256: https://github.com/pantsbuild/pants/pull/6088 looks great.
  • w

    witty-crayon-22786

    07/10/2018, 9:21 PM
    had a few questions about the goalrunner/context/pantsrunner split
  • r

    rough-minister-58256

    07/10/2018, 9:21 PM
    rad, have a handful more test failures to fixup in CI - got distracted fire fighting - but will wrap it up shortly
  • w

    witty-crayon-22786

    07/10/2018, 9:22 PM
    happy to talk about how to resolve the caching issue... oddly enough, we already have a 
    cacheable
    flag on the rust task struct, that i think is just completely unused right now: https://github.com/pantsbuild/pants/blob/master/src/rust/engine/src/tasks.rs#L17
  • r

    rough-minister-58256

    07/10/2018, 9:22 PM
    and yeah. the GoalRunner bit was largely to satisfy non-daemon execution - bit weird, I agree. might be able to improve that - will take another look after lunch.
  • r

    rough-minister-58256

    07/10/2018, 9:24 PM
    but ultimately without the daemon there’s no pre-GR graph helper instance
  • r

    rough-minister-58256

    07/10/2018, 9:25 PM
    ..yet 😉
  • w

    witty-crayon-22786

    07/10/2018, 9:32 PM
    hm. that is odd to me i think, because we compute targetroots before creating the GR
  • w

    witty-crayon-22786

    07/10/2018, 9:33 PM
    which means we have a scheduler at least.
  • r

    rough-minister-58256

    07/10/2018, 9:34 PM
    the only case we have targetroots pre-GR(F) is with the daemon: https://github.com/pantsbuild/pants/blob/master/src/python/pants/bin/goal_runner.py#L88
  • r

    rough-minister-58256

    07/10/2018, 9:35 PM
  • w

    witty-crayon-22786

    07/10/2018, 9:36 PM
    ah, ok.
  • w

    witty-crayon-22786

    07/10/2018, 9:36 PM
    yea, more decomposition. but happy to land and then iterate.
  • r

    rough-minister-58256

    07/10/2018, 9:37 PM
    sg
  • a

    aloof-angle-91616

    07/11/2018, 6:15 AM
    i'm seeing an error on importing the native engine module when invoking the scheduler in tests, this is happening on all branches and i've removed 
    ~/.cache/pants
    as well as 
    .pants.d
    , and run a 
    git clean -xfd
    at the repo root. i'm seeing the beginning of the log when i switch branches or do something else to make rust recompile (this is not the error, but i haven't seen this before). the error is after the first ellipsis, where there appears to be an error importing the native engine (I think?) for some reason. where could i start trying to figure out where this came from?
    -.txt
  • a

    aloof-angle-91616

    07/11/2018, 6:16 AM
    i thought it might have been an issue with my system or something but updating and upgrading all my packages did not change anything
  • a

    aloof-angle-91616

    07/11/2018, 6:17 AM
    i feel like this is obvious i just don't understand what cache to blow away or what config i could have screwed with
  • a

    aloof-angle-91616

    07/11/2018, 6:17 AM
    this happens on master, to clarify, and i don't think master is broken
  • a

    aloof-angle-91616

    07/11/2018, 6:18 AM
    i guess i could start by trying to do that import of that shared object
  • a

    aloof-angle-91616

    07/11/2018, 6:19 AM
    didn't even google that symbol yet i am off my game today
  • a

    aloof-angle-91616

    07/11/2018, 6:21 AM
    almost definitely a python version issue, sorry
  • a

    aloof-angle-91616

    07/11/2018, 6:53 AM
    still not quite able to figure out the issue, currently running again with 
    PATH=/bin:/usr/bin'
    so if anyone has thoughts on the above they would still be applicable
  • r

    rough-minister-58256

    07/11/2018, 7:14 AM
    the problem there is a clear ABI type mismatch for 
    cffi-1.11.1-cp27-cp27m-manylinux1_x86_64.whl
    vs whatever interpreter you’re executing with (which if on linux, is usually 
    cp27mu
    where the 
    u
    is “wide unicode” aka UCS4 vs UCS2)
    a
    • 2
    • 2
  • r

    rough-minister-58256

    07/11/2018, 7:16 AM
    whatever resolved that 
    cffi
    wheel was lax about the ABI type matching - which is pex’ default mode for a short form platform, but can be specified more concretely
  • r

    rough-minister-58256

    07/11/2018, 7:17 AM
    platforms=['linux-x86_64-cp-27-mu']
    vs 
    platforms=['linux-x86_64']
  • a

    average-vr-56795

    07/11/2018, 2:36 PM
    @rough-minister-58256 Took a look at the moon landing patch, looks awesome and super exciting 😄 “TODO: Running this command a second time with the daemon will result in no output because of product caching.” - which piece needs fixing there?
  • r

    rough-minister-58256

    07/11/2018, 4:34 PM
    that’s the second half of the change that I mentioned in standup
  • w

    witty-crayon-22786

    07/11/2018, 5:17 PM
    i think the cleanest way would be to make it a native concept in the graph, with an assertion/panic (for now!) if an uncacheable node is anything other than a root
  • w

    witty-crayon-22786

    07/11/2018, 5:17 PM
    that might be cross-pollinationable though
  • w

    witty-crayon-22786

    07/11/2018, 5:18 PM
    (ie, not huge)
1...553554555...906Latest