You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lucy.apache.org by "David E. Wheeler" <da...@justatheory.com> on 2013/07/23 19:01:44 UTC

[lucy-dev] Travis CI?

Lucy Hackers,

I�ve been fiddling with Travis CI, mostly because it�s dead simple to get it to build and run tests on many different versions of Perl for every push to GitHub. I forked Lucy and tried it by adding the attached file to the root of the Git repository. There were some failures. :-(

  https://travis-ci.org/theory/lucy/builds/9400267

Anyway, if you want to run it on Travis, I don�t know if they have some way to hook into Apache stuff, but if someone wants has write access to Apache repos on GitHub, you can give Travis access then add and push the attached file to get it going. Details here:

  http://about.travis-ci.org/docs/user/getting-started/

Best,

David

Re: [lucy-dev] Travis CI?

Posted by Marvin Humphrey <ma...@rectangular.com>.
On Tue, Aug 13, 2013 at 5:34 AM, Marvin Humphrey <ma...@rectangular.com> wrote:
> On Tue, Aug 13, 2013 at 2:34 AM, Nick Wellnhofer <we...@aevum.de> wrote:
>> On 05/08/2013 04:38, Marvin Humphrey wrote:
>>> Anyway, the RAT buildbot is producing readable output once again.  It's
>>> not passing -- we've got a few files to take care of:
>>>
>>>      http://ci.apache.org/projects/lucy/rat-output.html
>>>
>>> But that's straightforward enough.
>>
>>
>> I added license headers to a couple of files but it seems that the RAT
>> report hasn't been updated for some reason.
>
> Thanks for working on those.
>
> I think the buildslave that runs our RAT report is offline, due to some
> problem with the upgrades that were done as a result of our issue.

The buildslave which runs our RAT buildbot task is back.  I've committed a few
more license headers and master is now passing.

I've also added a new sh script which runs all tests; I hope to switch all of
our CI over to it, including Travis.  Windows testing is still an open
question, though.

    https://git-wip-us.apache.org/repos/asf?p=lucy.git;a=blob;f=devel/bin/test_all.sh;h=60f311b5f96374063606fa4cf4c7e1d514040129;hb=omni-test

Marvin Humphrey

Re: [lucy-dev] Travis CI?

Posted by Marvin Humphrey <ma...@rectangular.com>.
On Tue, Aug 13, 2013 at 2:34 AM, Nick Wellnhofer <we...@aevum.de> wrote:
> On 05/08/2013 04:38, Marvin Humphrey wrote:
>>
>> Anyway, the RAT buildbot is producing readable output once again.  It's
>> not
>> passing -- we've got a few files to take care of:
>>
>>      http://ci.apache.org/projects/lucy/rat-output.html
>>
>> But that's straightforward enough.
>
>
> I added license headers to a couple of files but it seems that the RAT
> report hasn't been updated for some reason.

Thanks for working on those.

I think the buildslave that runs our RAT report is offline, due to some
problem with the upgrades that were done as a result of our issue.  Gav wrote
in our INFRA Jira:

     http://s.apache.org/M2o

    Both slaves it looks like were using default 1.7.1 ant install rather than
    other available Ant 1.8.[2|3]

    I've since upgraded both default Ant installs to the latest 1.9.x

    Other work needed included a reboot, and so far Portunus (slave2) hasn't
    come back so waiting for that one to return to do a new compare.

    Note also that one slave used rat 0.8-snapshot whilst the other used
    0.9-snapshot. Both will use 0.9

>From IRC:

    creamygoodness: lucy-bot: force build lucy-trunk-rat
    lucy-bot: The build has been queued, I'll give a shout when it starts

The next step would be to thank Gav for the efforts so far and ask him to
either address the downed machine or reassign our builds to another box.  We
might be able to do something ourselves -- both Logan and I seem to have
commit privs on the buildslave area of svn, but I'm not sure how far we're
supposed to go in using it.

I'll work on this task later today.

Marvin Humphrey

Re: [lucy-dev] Travis CI?

Posted by Nick Wellnhofer <we...@aevum.de>.
On 05/08/2013 04:38, Marvin Humphrey wrote:
> Anyway, the RAT buildbot is producing readable output once again.  It's not
> passing -- we've got a few files to take care of:
>
>      http://ci.apache.org/projects/lucy/rat-output.html
>
> But that's straightforward enough.

I added license headers to a couple of files but it seems that the RAT 
report hasn't been updated for some reason.

Nick

Re: [lucy-dev] Travis CI?

Posted by Marvin Humphrey <ma...@rectangular.com>.
On Wed, Jul 24, 2013 at 6:15 PM, Marvin Humphrey <ma...@rectangular.com> wrote:
> On Wed, Jul 24, 2013 at 2:15 PM, David E. Wheeler <da...@justatheory.com> wrote:
>> Yeah, the more testing the better!
>
> +1
>
> I've started by trying to repair our RAT buildbot, which is currently
> malfunctioning.

Sheesh, that was harder than it should have been.  Turns out the weird
behavior was due to differing versions of Ant, and my workaround didn't work
because I hadn't grokked that our rat-excludes file uses Ant fileset syntax.

    https://issues.apache.org/jira/browse/INFRA-6625
    http://mail-archives.apache.org/mod_mbox/creadur-dev/201308.mbox/%3CCAAS6%3D7jqPDG8v7%3Dqn0Qc0AkjZ-F4bhyFLn9LgUtDv2JcHsNMSg%40mail.gmail.com%3E

Anyway, the RAT buildbot is producing readable output once again.  It's not
passing -- we've got a few files to take care of:

    http://ci.apache.org/projects/lucy/rat-output.html

But that's straightforward enough.

What I'd like to do next is get continuous integration set up for multiple
hosts, and also for running the Clownfish and CFC test suites.  Instead of
managing multiple CI setups, how about one mondo script which runs all
possible congigs in our tree (probing for system capabilities at the start and
cleaning up in between as necessary)?

Marvin Humphrey

Re: [lucy-dev] Travis CI?

Posted by Marvin Humphrey <ma...@rectangular.com>.
On Wed, Jul 24, 2013 at 2:15 PM, David E. Wheeler <da...@justatheory.com> wrote:
> Yeah, the more testing the better!

+1

I've started by trying to repair our RAT buildbot, which is currently
malfunctioning.

http://mail-archives.apache.org/mod_mbox/www-builds/201307.mbox/%3CCAAS6%3D7iwJJvphkNz1sZu_iiZfuDS9Nyg63A9kT7VxRQdxBhq3Q%40mail.gmail.com%3E

Marvin Humphrey

Re: [lucy-dev] Travis CI?

Posted by "David E. Wheeler" <da...@justatheory.com>.
On Jul 24, 2013, at 3:56 AM, Marvin Humphrey <ma...@rectangular.com> wrote:

> They don't support Perl 5.8 FWIW:
> 
>    http://about.travis-ci.org/docs/user/languages/perl/#Versions-earlier-than-5.10-Are-Not-Provided

That's clearly out-dated documentation.

  http://about.travis-ci.org/docs/user/ci-environment/#Perl-VM-images

As you can see, the tests passed on 5.8. :-)

> We don't have administrative access to the Apache github mirrors, because
> Github doesn't support fine-grained permissions and Infra is understandably
> reluctant to give everybody access to everything.  However, it looks like a
> small number of projects have Travis set up.
> 
>    https://issues.apache.org/jira/browse/INFRA-6161
>    https://issues.apache.org/jira/browse/INFRA-4698
>    https://issues.apache.org/jira/browse/OAK-145
>    https://issues.apache.org/jira/browse/INFRA-6124
> 
> I'd say go ahead and commit .travis.yml to master.  After that, I'll get
> Travis set up on my github fork just for practice, then take up the matter
> with Infra and see what they say.

Okay.

> We may want to be more aggressive about taking advantage of the capabilities
> ci.apache.org already offers before pushing too hard, though.

Yeah, the more testing the better!

Best,

David

Re: [lucy-dev] Travis CI?

Posted by Marvin Humphrey <ma...@rectangular.com>.
On Tue, Jul 23, 2013 at 10:01 AM, David E. Wheeler
<da...@justatheory.com> wrote:
> I’ve been fiddling with Travis CI, mostly because it’s dead simple to get it
> to build and run tests on many different versions of Perl for every push to
> GitHub.

I'm cool with this.

Note that the ASF already provides continuous integration services:

    http://ci.apache.org/buildbot.html

We're already using buildbot.  Logan set up our config:

    https://svn.apache.org/repos/infra/infrastructure/buildbot/aegis/buildmaster/master1/projects/lucy.conf

Every push to master triggers a build/test run with stock Perl on one of the
FreeBSD machines.

I'm not sure how many different versions of Perl are available to us, but
there's certainly some operating system diversity we aren't taking advantage
of yet.

> I forked Lucy and tried it by adding the attached file to the root
> of the Git repository.

The dev@lucy mailing list strips attachments, but here's the content of that
file:

    https://github.com/theory/lucy/blob/1e49a7ccab4bfb62dc045f4a82d02e0f4f0b11e0/.travis.yml

    language: perl
    perl:
      - "5.19"
      - "5.18"
      - "5.16"
      - "5.14"
      - "5.12"
      - "5.10"
      - "5.8"
    before_install:
      - cd perl

They don't support Perl 5.8 FWIW:

    http://about.travis-ci.org/docs/user/languages/perl/#Versions-earlier-than-5.10-Are-Not-Provided

> There were some failures. :-(

We're on it. :)

> Anyway, if you want to run it on Travis, I don’t know if they have some way
> to hook into Apache stuff, but if someone wants has write access to Apache
> repos on GitHub, you can give Travis access then add and push the attached
> file to get it going. Details here:
>
>   http://about.travis-ci.org/docs/user/getting-started/

We don't have administrative access to the Apache github mirrors, because
Github doesn't support fine-grained permissions and Infra is understandably
reluctant to give everybody access to everything.  However, it looks like a
small number of projects have Travis set up.

    https://issues.apache.org/jira/browse/INFRA-6161
    https://issues.apache.org/jira/browse/INFRA-4698
    https://issues.apache.org/jira/browse/OAK-145
    https://issues.apache.org/jira/browse/INFRA-6124

I'd say go ahead and commit .travis.yml to master.  After that, I'll get
Travis set up on my github fork just for practice, then take up the matter
with Infra and see what they say.

We may want to be more aggressive about taking advantage of the capabilities
ci.apache.org already offers before pushing too hard, though.

Marvin Humphrey