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

    witty-crayon-22786

    01/23/2017, 8:28 PM
    so you'd fetch a cache artifact indicating you're on branch X rather than running git to determine that it's actually Y
  • e

    enough-analyst-54434

    01/23/2017, 8:28 PM
    although the git info then becomes somewhat irrelevant, but true
  • w

    witty-crayon-22786

    01/23/2017, 8:28 PM
    i feel bad about this, because we have an implementation of approximately this task that we ought to upstrem
  • e

    enough-analyst-54434

    01/23/2017, 8:29 PM
    So does everyone
  • w

    witty-crayon-22786

    01/23/2017, 8:29 PM
    will create a ticket about that
  • e

    enough-analyst-54434

    01/23/2017, 8:29 PM
    apparently
  • h

    high-advantage-47550

    01/23/2017, 9:03 PM
    is there a way to have the 
    binary
    task generate a jar with the version appended to the filename?
  • w

    witty-crayon-22786

    01/23/2017, 9:04 PM
    versions don't really exist by default
  • w

    witty-crayon-22786

    01/23/2017, 9:04 PM
    but if you were to change the binary 
    basename
    to something containing a manually specified version, then that would do it
  • h

    high-advantage-47550

    01/23/2017, 9:05 PM
    ah, right, thanks!
  • h

    high-advantage-47550

    01/24/2017, 1:19 AM
    is anyone successfully using pants with nodejs/webpack projects? i’m looking at the plugin
  • r

    rough-minister-58256

    01/24/2017, 1:24 AM
    we use it fairly heavily at Twitter
  • r

    rough-minister-58256

    01/24/2017, 1:25 AM
    @brash-cricket-7489 is OOO for a few days, but he’s the lead for our Web Tools team who owns the pants+nodejs work
  • h

    high-advantage-47550

    01/24/2017, 1:29 AM
    if you don’t mind my asking, are you using https://github.com/pantsbuild/pants/tree/master/contrib/node or a fork, or something internal?
  • h

    high-advantage-47550

    01/24/2017, 1:30 AM
    (if you know, otherwise I can wait until he’s back 🙂 )
  • w

    wide-energy-11069

    01/24/2017, 1:31 AM
    i dont see any special settings on node, so should just that vanilla
  • h

    high-advantage-47550

    01/24/2017, 1:31 AM
    thanks!
  • a

    acceptable-scooter-37412

    01/24/2017, 1:35 PM
    @witty-crayon-22786 and @enough-analyst-54434 thanks for taking that time to explain the underlying mechanism and for pointing out the issue
  • a

    acceptable-scooter-37412

    01/24/2017, 1:37 PM
    I tried to find somewhere to set 
    invalidate_dependents
    , but the only place I see it being set is hardcoded as False in ResourcesTask here: https://github.com/pantsbuild/pants/blob/master/src/python/pants/backend/jvm/tasks/resources_task.py#L54
  • a

    acceptable-scooter-37412

    01/24/2017, 1:39 PM
    to get around that, I followed the call chain down to where that 
    invalidate_dependents
    is actually used and I see that is used in 
    create_cache_manager
    in 
    TaskBase
    here: https://github.com/pantsbuild/pants/blob/master/src/python/pants/task/task.py#L264
  • a

    acceptable-scooter-37412

    01/24/2017, 1:40 PM
    so in my BuildPropertiesTask I decided to try to override that method and force 
    invalidate_dependents
    to be 
    True
    . So something like:
    Copy code
    def create_cache_manager(self, invalidate_dependents, fingerprint_strategy=None):
        return super(BuildPropertiesTask, self).create_cache_manager(
            invalidate_dependents=True,
            fingerprint_strategy=fingerprint_strategy)
    p
    • 2
    • 2
  • a

    acceptable-scooter-37412

    01/24/2017, 1:40 PM
    and presto that seems to work
  • a

    acceptable-scooter-37412

    01/24/2017, 1:41 PM
    so that is a little cheeky workaround for now, but is there a better way to do that?
  • a

    acceptable-scooter-37412

    01/24/2017, 1:41 PM
    and as for the fingerprintstrategy stuff -- mindblown. I might have to spend a bit of time investigating and understanding that
  • e

    enough-analyst-54434

    01/24/2017, 3:21 PM
    I'd classify that as cheeky - yes. 2 reasons: - the method is not marked 
    :API: public
    so its not guaranteed to be stable when you upgrade to pants next - more generally, ResourcesTask is a high-level base, yet this is reaching into a low level to tweak
  • e

    enough-analyst-54434

    01/24/2017, 3:21 PM
    I think the best 2 paths are just return the transitive closure of the binary targets as the immediate workaround
  • e

    enough-analyst-54434

    01/24/2017, 3:22 PM
    Then, optionally, expose an 
    :API: public
    knob in ResourcesTask for invalidate_dependants and eventually switch to consuming that new API when it rolls out to a pants version you're comfortable upgrading to.
  • e

    enough-analyst-54434

    01/24/2017, 3:23 PM
    All this said, if you implement a custom fingerprint that uses the git sha, you really need nothing more - targets are irrelevant at that point if you actually care about getting "this branch's" sha in the build propreties - the detail Stu and I discussed with each other yesterday.
  • e

    enough-analyst-54434

    01/24/2017, 3:58 PM
    I was bit by 
    TestProjectsIntegrationTest
    pretty hard last night as was contributor James Bench. I filed https://github.com/pantsbuild/pants/issues/4204 but wanted to check in on status / knowledge of this problem. I'm not sure if I hit a size edge by bad luck or if this has been a known problem.
  • e

    enough-analyst-54434

    01/24/2017, 3:59 PM
    The problem being no test output for >10 minutes under travis which times out the build.
1...159815991600...2420Latest