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

    aloof-angle-91616

    05/20/2018, 3:23 AM
    ^also it totally makes sense to me now why that (mild) slowdown occurs, we're storing `PathStat`s and `PathGlob`s in a map instead of just using a set, i'm not super bothered by that slowdown anymore (although i haven't gotten to the point of inspecting a flamegraph of the native calls to check)
  • a

    aloof-angle-91616

    05/20/2018, 3:26 AM
    unclear if this is something people have already seen but this paper from a now-defunct MSR lab in 2012 describes monadic incremental computation with execution traces and has a high-quality implementation in rust (the repo containing this pdf) and shouts out twitter specifically many times as an example application https://github.com/frankmcsherry/differential-dataflow/blob/master/differentialdataflow.pdf
  • a

    aloof-angle-91616

    05/20/2018, 3:27 AM
    transitively linked from https://github.com/frankmcsherry/blog/blob/master/posts/2018-05-19.md which was on hn today
  • e

    enough-analyst-54434

    05/20/2018, 10:51 PM
    A heads-up about changes to the rust build that just landed in https://github.com/pantsbuild/pants/commit/195f4d5312680b5a570e3d9906d05891aff0bdb9 Fairly large change, but I expect a smooth transition including branch switches back and forth from old to new. Best to read https://github.com/pantsbuild/pants/blob/master/build-support/bin/native/README.md and re-read https://github.com/pantsbuild/pants/blob/master/src/rust/engine/README.md if there is any confusion about command line builds / hacking or IDEA setup. Both should now be simpler / more uniform.
    ✨ 1
    a
    • 2
    • 1
  • a

    aloof-angle-91616

    05/21/2018, 6:09 AM
    omg i just realized i lost all the integration tests for the source glob matching pr and i was about to scream but the sha was on the pants clone in my osx laptop thank god
  • a

    aloof-angle-91616

    05/21/2018, 6:10 AM
    now i get to see if the code passes the testing at all 🤣
  • a

    aloof-angle-91616

    05/21/2018, 6:15 AM
    lol i had added the target in the build file but forgot the test source file, this is something that would have been caught by this fix 🤣 🤣 🤣 🤣 🤣 🤣 🤣
  • a

    aloof-angle-91616

    05/21/2018, 6:19 AM
    WARN] In target tests/python/pants_test/engine/legacy:graph_integration with sources=['test_graph_integration.py']: glob pattern 'test_graph_integration.py' did not match any files.
  • a

    aloof-angle-91616

    05/21/2018, 6:26 AM
    the tests pass except for the one small todo i hadn't gotten to yet 🤣 🤣 🤣 🤣 🤣 🤣 🤣 🤣 🤣 🤣 🤣 🤣
  • a

    average-vr-56795

    05/21/2018, 4:07 PM
    Idle question… How do folks feel about the following rust struct initialisers?
    Copy code
    Foo {
  bar,
}

Foo {
  bar: bar,
}

Foo {
  bar,
  baz: bash,
}

Foo {
  bar: bar,
  baz: bash,
}
    ?
  • r

    rough-minister-58256

    05/21/2018, 5:28 PM
    don’t have a ton of context, but generally speaking: explicit form is generally better for readability
  • w

    witty-crayon-22786

    05/21/2018, 6:58 PM
    @average-vr-56795: i like them, because i think it encourages meaningful, non-colliding variable names.
    ☝️ 1
    a
    a
    • 3
    • 27
  • w

    witty-crayon-22786

    05/21/2018, 6:58 PM
    (...in a way that spreads out of the struct and into nearby code)
  • w

    witty-crayon-22786

    05/21/2018, 6:59 PM
    @aloof-angle-91616: i'm guessing i'm missing something big, but: with regard to https://github.com/pantsbuild/pants/pull/5769/files#r189682029 ... can we just eagerly fail in 
    VFS::expand
    if we've been configured to?
    a
    • 2
    • 6
  • a

    aloof-angle-91616

    05/21/2018, 9:17 PM
    @witty-crayon-22786 @average-vr-56795 i'm thinking of adding a field to 
    PathGlobs
    (in rust and python) to denote "fail if one of the globs expands to nothing" -- this (along with failing in 
    VFS#expand()
    ) would remove the extra intrinsic and all changes to 
    native.py
    and 
    <http://externs.rs|externs.rs>
    , i think. does that sound reasonable?
  • w

    witty-crayon-22786

    05/21/2018, 9:18 PM
    @aloof-angle-91616: i think that since it is a global setting, you should be able to just stash the setting on 
    context.core
    and then pass it explicitly to VFS::expand
  • w

    witty-crayon-22786

    05/21/2018, 9:18 PM
    ...hm. actually, perhaps that would fail in some cases we don't want to fail for.
  • a

    aloof-angle-91616

    05/21/2018, 9:18 PM
    i hadn't thought of it as a global setting, but i guess where else are we expanding pathglobs into snapshots except in sources args?
  • w

    witty-crayon-22786

    05/21/2018, 9:19 PM
    (...
    BUILD
    file globbing in particular)
  • a

    aloof-angle-91616

    05/21/2018, 9:19 PM
    yeah
  • w

    witty-crayon-22786

    05/21/2018, 9:19 PM
    @aloof-angle-91616: yea, ok: i think adding it as a field to PathGlobs would be reasonable.
  • a

    aloof-angle-91616

    05/21/2018, 9:20 PM
    ok, neat. it probably wouldn't be too hard to move it elsewhere if we want to for some reason
  • w

    witty-crayon-22786

    05/21/2018, 9:20 PM
    yea.
  • w

    witty-crayon-22786

    05/21/2018, 9:20 PM
    (lemme look at a few codepaths that create PathGlobs to confirm)
  • a

    aloof-angle-91616

    05/21/2018, 9:21 PM
    sick
  • w

    witty-crayon-22786

    05/21/2018, 9:21 PM
    yea, i think it needs to be a per-match setting. so yea, on PathGlobs is good.
    👌 1
  • a

    aloof-angle-91616

    05/21/2018, 9:21 PM
    still moving slowly due to back issues so you don't need to hurry too much in confirming
  • a

    aloof-angle-91616

    05/21/2018, 9:22 PM
    sick
  • w

    witty-crayon-22786

    05/21/2018, 9:22 PM
    yep, no problem.
  • r

    rough-minister-58256

    05/21/2018, 10:15 PM
    pub fn get_stdout_raw(&self) -> &[u8]
1...491492493...906Latest