[aerogear-dev] Travis-ci and CloudBees problems on Android land

[Karel Piwko]

On Thu, 2014-12-11 at 10:04 -0500, Summers Pittman wrote:
> On 12/11/2014 09:46 AM, Karel Piwko wrote:
> > Hi Passos,
> >
> > have you enabled XVnc / Framebuffer on Jenkins? It looks that it might
> > be related to the log output.
> >
> > If you have issues with Android SDK version and/or applying any
> > configuration to existing installation, I'd suggest to use Arquillian
> > Gradle Spacelift, which is able to bring Android SDK runtime there for
> > you very cheaply. Stefan has automated testsuite for our Android
> > ftesting tool that way, so if you have a Linux machine (adding support
> > for Mac OS X should be pretty easy), you just need to clone and run
> > Gradle.
> And port the project to Gradle from Maven.

Well, not really, Gradle actually calls Maven, so the project itself
stays unchanged.

> >
> > https://github.com/arquillian/arquillian-droidium/tree/master/tests
> >
> > Maybe it could be reused for Android integration tests?
> >
> > Also I'd suggest not to use ARMv7 ABI but x86 based one. For most CI
> > setups ARMv7 means a VM in VM and that is terribly slow unless you have
> > nested VM support in both HW and SW - which you most likely don't
> > have ;-)
> x86 has the same problem unless the environment has KVM installed. I 
> know Travis CI does not and I know that in all of CloudBees' docs it 
> refers to the arm emulator instead of using x86.

Thanks, that's a very valuable piece of information.

> >
> > Karel
> We've also got the build running in Travis again by being much smarter 
> about when we start and stop the emulator and project compiling.

Sweet!

> >   
> > On Mon, 2014-12-08 at 13:59 -0200, Daniel Passos wrote:
> >> Hi All,
> >>
> >>
> >> The Android team has been working on resolving a travis build issue
> >> last week (and part of weekend) with travis-ci[1][2]. I has been
> >> testing with proting the build to cloudbees[3][4].
> >>
> >>
> >> travis-ci:
> >>
> >>
> >> In Travis the VM is running out of memory and getting killed. We fixed
> >> this by starting the emulator after compiling and I was able to make
> >> it pass 2 times[5][6], but when a PR[7] for AG was sent, we were back
> >> to the old problem[8] (travis-ci killed our build, because it's using
> >> a lot of memory)
> >>
> >>
> >> CloudBees:
> >>
> >>
> >> We are having trouble running the emulator[9] on CloudBees ([android]
> >> Emulator did not appear to start; giving up). We found a fix[10] but
> >> are not sure if we can apply it on CloudBees
> >>
> >>
> >> [1]
> >> https://travis-ci.org/danielpassos/aerogear-android-security/builds
> >> [2] https://travis-ci.org/secondsun/aerogear-android-security/builds
> >> [3] https://aerogear.ci.cloudbees.com/job/aerogear-android-security
> >> [4] https://aerogear.ci.cloudbees.com/job/aerogear-android-pipe
> >> [5]
> >> https://travis-ci.org/danielpassos/aerogear-android-security/builds/43219446
> >> [6]
> >> https://travis-ci.org/danielpassos/aerogear-android-security/builds/43256877
> >> [7] https://github.com/aerogear/aerogear-android-security/pull/17
> >> [8]
> >> https://travis-ci.org/aerogear/aerogear-android-security/builds/43295194
> >> [9]
> >> https://aerogear.ci.cloudbees.com/job/aerogear-android-pipe/6/jdk=openjdk8/console
> >> [10]:
> >> http://stackoverflow.com/questions/19349222/jenkins-android-emulator-emulator-did-not-appear-to-start-giving-up
> >>
> >>
> >> -- Passos
> >> _______________________________________________
> >> aerogear-dev mailing list
> >> aerogear-dev at lists.jboss.org
> >> https://lists.jboss.org/mailman/listinfo/aerogear-dev
> >
> > _______________________________________________
> > aerogear-dev mailing list
> > aerogear-dev at lists.jboss.org
> > https://lists.jboss.org/mailman/listinfo/aerogear-dev
> 
>