https://pantsbuild.org/ logo
Powered by
#general
  • e

    enough-analyst-54434

    07/21/2015, 7:37 PM
    But again, not a nice piece of advice for many users... as evidenced by the 1751 issue thread, he had no clue how to prebuild
  • w

    witty-crayon-22786

    07/21/2015, 7:38 PM
    Copy code
    $ build-support/bin/release-changelog-helper.sh | vim -
What sha was the last release made from?: Vim: Reading from stdin...
2a0316a583882edbaadf4c95c9423548ae21931f
    ignored 3488a6aec78363a70e5ea4ea6d7e92a28ca57f86 for some reason
  • e

    enough-analyst-54434

    07/21/2015, 7:38 PM
    I use the tags
  • e

    enough-analyst-54434

    07/21/2015, 7:38 PM
    release_0.0.37
  • w

    witty-crayon-22786

    07/21/2015, 7:38 PM
    @adventurous-twilight-44110: the release is out, but we'll try to get you in the next changelog for that one. sorry!
  • w

    witty-crayon-22786

    07/21/2015, 7:39 PM
    @adventurous-twilight-44110: would you mind following up to add docs about how/when to use 
    register(.., fingerprint=True)
    ?
  • r

    rough-minister-58256

    07/21/2015, 7:39 PM
    hmm, but that seems to be an error on the packagers part - one that could be fixed by the dev explicitly defining what should otherwise be a transitive dep; perhaps a more strategic approach would be to submit upstream fixes to matplotlib/scipy et al to correct their use of setup.py
  • a

    adventurous-twilight-44110

    07/21/2015, 7:39 PM
    sure, like on the website?
  • r

    rough-minister-58256

    07/21/2015, 7:40 PM
    maybe a doc on ‘how to pre-build’ too
  • w

    witty-crayon-22786

    07/21/2015, 7:40 PM
    @adventurous-twilight-44110: http://pantsbuild.github.io/docs.html
  • r

    rough-minister-58256

    07/21/2015, 7:40 PM
    or a brute-builder
  • e

    enough-analyst-54434

    07/21/2015, 7:40 PM
    @rough-minister-58256: take a look at my pip example - the issue is controlling order. For the pip repro I had no control over this and pip tried scipy 1st
  • r

    rough-minister-58256

    07/21/2015, 7:41 PM
    k, will digest this a bit
  • e

    enough-analyst-54434

    07/21/2015, 7:41 PM
    yeah - I think docs may be the only practical hope
  • w

    witty-crayon-22786

    07/21/2015, 8:16 PM
    @happy-kitchen-89482: we have @mammoth-pharmacist-54215 making a change to allow cleaning up artifacts in the local artifact cache. it adds a new option to the CacheSetup subsystem, and so he essentially needs that option set in all unit tests that use the cache. should he just set it in TaskTestBase somewhere?
  • w

    witty-crayon-22786

    07/21/2015, 8:16 PM
    i'm guessing we set 
    cache.read=False
    and 
    cache.write=False
    somewhere globally in order to not get an attribute error from TestOptiuonsValues
  • m

    mammoth-pharmacist-54215

    07/21/2015, 8:31 PM
    I suspect this may have actually been an issue with accessing as a global, give me a little bit to re-run ci and confirm.
  • h

    happy-kitchen-89482

    07/21/2015, 9:05 PM
    No unit tests should use the cache
  • h

    happy-kitchen-89482

    07/21/2015, 9:05 PM
    yes, those are both set to False
  • h

    happy-kitchen-89482

    07/21/2015, 9:05 PM
    in tests
  • h

    happy-kitchen-89482

    07/21/2015, 9:05 PM
    and then re-enabled selectively, e.g., so tests don't bootstrap jvm tools every time
  • h

    happy-kitchen-89482

    07/21/2015, 9:06 PM
    Can we get more details on that change though?
  • m

    mammoth-pharmacist-54215

    07/21/2015, 9:07 PM
  • m

    mammoth-pharmacist-54215

    07/21/2015, 9:09 PM
    I've changed the code to use _options.target_entry_max instead of the global private version and that seems to resolve the issue, I'll update the RB shortly
  • h

    happy-kitchen-89482

    07/21/2015, 9:19 PM
    Cool
  • h

    happy-kitchen-89482

    07/21/2015, 9:19 PM
    added some comments, since I was in there
  • m

    mammoth-pharmacist-54215

    07/21/2015, 9:30 PM
    thanks for the feedback, perhaps I'm doing something wrong but when I use the option and none was specified I don't get an int type I get a RankedValue, this is why I was doing the type checking. Is there a cleaner way to do this?
  • w

    witty-crayon-22786

    07/21/2015, 9:51 PM
    @mammoth-pharmacist-54215: reference the field directly with 
    get_options().option_name
    and you should not see a RankedValue.
  • w

    witty-crayon-22786

    07/21/2015, 9:51 PM
    lots of examples around in Tasks
  • w

    witty-crayon-22786

    07/21/2015, 9:52 PM
    you're in a Subsystem, but it should still be the same: 
    _options.option_name
1...676869...2420Latest