[erlang-questions] travis-ci erlang support broken?

Tuncer Ayaz tuncer.ayaz@REDACTED
Wed Jan 14 21:34:07 CET 2015


On Sun, Jan 4, 2015 at 12:38 PM, Tuncer Ayaz wrote:
> On Fri, Jan 2, 2015 at 10:20 PM, Tuncer Ayaz wrote:
> > On Fri, Jan 2, 2015 at 9:49 PM, Tuncer Ayaz wrote:
> > > Does anybody know if the following travis-ci Erlang distro
> > > breakage is due to some kind of recent compartmentalization
> > > changes, and if so, what changes are required to continue using
> > > travis-ci for Erlang?
> > >
> > > Can we use only short names for nodes now?
> > >
> > > curl
> > > https://s3.amazonaws.com/archive.travis-ci.org/jobs/45606731/log.txt
> > > | tail -n 45
> >
> > The relevant bit:
> > {error_logger,{{2015,1,1},{13,48,39}},
> >     "Can't set long node name!\nPlease check your configuration\n",[]}
> >
> > Also, I should have tried -sname before asking the list:
> > https://github.com/rebar/rebar/pull/421
>
> After switching -name to -sname, the test runs into:
> {error_logger,{{2015,1,2},{21,22,11}},
>   \"Protocol: ~tp: the name ct_rt3@REDACTED seems to be in use by
> another Erlang node\", [\"inet_tcp\"]}
>
> Clearly, it didn't fix the problem, and since the tests ran fine
> previously, I'll have to assume something's changed in travis-ci's
> Erlang support. They're out-of-office for the moment, so we'll
> hopefully find out what changed once they're back.

This has since been fixed by switching to Travis-CI's new container-based
workers by adding

sudo: false

to .travis.yml.

During the tests, it came to light that the host config wasn't correct
enough for `erl -name ct_rt3` to work automatically. We could have
resorted to using ct_rt@REDACTED, but that would be IPv4-specific and
require us to remember it for any and all distributed node spun up in
tests.

Thanks to Hiro Asari from Travis-CI for suggesting the new workers, as
that env works out of the box by returning a proper FQDN.



More information about the erlang-questions mailing list