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

    proud-alligator-68508

    06/08/2015, 8:19 PM
    that includes downloads, i assume?
  • r

    rough-minister-58256

    06/08/2015, 8:19 PM
    yup
  • p

    proud-alligator-68508

    06/08/2015, 8:19 PM
    with a full local 
    ~/.ivy2
    cache, our worst ivy resolve times were just a few minutes, when we were using ivy. full download.. well, depended on network.
  • b

    bored-art-40741

    06/08/2015, 8:19 PM
    Ours takes about 1-2 minutes of CPU burn time doing real resolving, and then however long it takes to do the downloads in parallel
  • r

    rough-minister-58256

    06/08/2015, 8:20 PM
    yeah, once cached it’s not terrible but still seems like huge room for improvement there from a perf standpoint
  • b

    bored-art-40741

    06/08/2015, 8:20 PM
    It doesn't download anything except poms (which it maintains a small local cache for) until it's done resolving
  • b

    bored-art-40741

    06/08/2015, 8:20 PM
    Then it does a batch fetch of all of the artifacts it determined it needs
  • r

    rough-minister-58256

    06/08/2015, 8:20 PM
    yup
  • b

    bored-art-40741

    06/08/2015, 8:20 PM
    (By "ours" I mean "our resolver", not our ivy setup)
  • r

    rough-minister-58256

    06/08/2015, 8:20 PM
    ah
  • r

    rough-minister-58256

    06/08/2015, 8:21 PM
    yeah, ivy works the same way
  • r

    rough-minister-58256

    06/08/2015, 8:21 PM
    two sequential phases w/ sequential work in each
  • a

    ancient-stone-50795

    06/08/2015, 8:21 PM
    In our ivy setup, the "worst case" resolve is on the order of 10 minutes. Our .ivy2 has about 1GB in it.
  • p

    proud-alligator-68508

    06/08/2015, 8:21 PM
    the python resolver i think is on par with ivy for pure resolution time, but having the globally consistent resolution means “resolving” a subset of targets is almost free if you’ve previously done a global resolve
  • a

    ancient-stone-50795

    06/08/2015, 8:22 PM
    Of course, we have a wickedfast connection to the internet. Maven is definitely faster when you have a cold cache.
  • p

    proud-alligator-68508

    06/08/2015, 8:23 PM
    so that 1-2 minute cpu burn is basically one-time, and then when you throw ad-hoc sets of targets, it is almost free, whereas when we had ivy doing resolve on the fly, it might pick different things on each subset, so it was doing a relatively expensive resolve for each new combination of targets it saw.
  • p

    proud-alligator-68508

    06/08/2015, 8:24 PM
    and since most of our engineers use a “find the set of targets that is transitively invalidated by what my commit changed” workflow, we were basically always handing ivy a set of targets it had never seen before.
  • m

    mammoth-autumn-2551

    06/09/2015, 7:08 PM
    Another random question for the java + thrift crowd: after upgrading to pants 0.0.32, we now observe compilation warnings from the thrift-generated source files
  • m

    mammoth-autumn-2551

    06/09/2015, 7:08 PM
    I know we have global java compile warnings specified in pants.ini, but can we disable them just for the thrift generated files?
  • m

    mammoth-autumn-2551

    06/09/2015, 7:09 PM
    also we don’t use scrooge, currently
  • a

    ancient-stone-50795

    06/09/2015, 7:34 PM
    If you are using jmake, you can pass an argument to jmake 
    -Tnowarnregex
    that will keep warnings that match the expression from showing up.
  • m

    mammoth-autumn-2551

    06/09/2015, 7:36 PM
    that sounds perfect.
  • a

    ancient-stone-50795

    06/09/2015, 7:37 PM
    @mammoth-autumn-2551: See the --compile-java-args option: 
    ./pants compile.java --help
    for the current defaults. You can just copy those and tack a new -Tnowarnregex onto the end
  • a

    ancient-stone-50795

    06/09/2015, 7:37 PM
    zinc will soon support a similar 
    -log-filter
    option
  • m

    mammoth-autumn-2551

    06/09/2015, 7:39 PM
    perfect, we already seem to have this in our pants.ini: # Suppress the warning for annotations with no processor - we know there are many of these! '-C-Tnowarnregex', '-C^(warning: )?No processor claimed any of these annotations: .*'
  • m

    mammoth-autumn-2551

    06/09/2015, 7:39 PM
    so I’ll just extend that
  • m

    mammoth-autumn-2551

    06/09/2015, 7:52 PM
    Thanks @ancient-stone-50795 ! For reference this is what I needed to add my pants.ini:
  • m

    mammoth-autumn-2551

    06/09/2015, 7:52 PM
    -.txt
    -.txt
  • a

    ancient-stone-50795

    06/10/2015, 9:10 PM
    I have been working on an issue for about a week dealing with setting 
    Class-Path
    in MANIFEST.MF for a jar file. We use a fat (shaded) jar for our deploys but sometimes need to include separate jars that have been signed. When I did this in the pants generated .jar file it could never find the extra jars I added:
    Copy code
    Manifest-Version: 1.0
Class-Path: gns-server-signed-jars/bcprov-jdk15on-1.50.jar
Created-By: org.pantsbuild.tools.jar.JarBuilder
Main-Class: com.squareup.gns.server.GnsServerApp
  • a

    ancient-stone-50795

    06/10/2015, 9:11 PM
    When I ran the code with 
    java -jar gns-server.jar
    and 
    gns-server-signed-jars/bcprov-jdk15on-1.50.jar
    in a directory locally, it couldn't resolve classes in the jar, but running:`
    Copy code
    java -cp gns-server.jar: gns-server-signed-jars/bcprov-jdk15on-1.50.jar com.squareup.gns.server.GnsServerApp
1...678...2420Latest