cgo hanging failure with go1.5 on travis

[Dan Kortschak]

We are seeing a consistent hanging failure with our travis-ci build of
the gonum cgo-based LAPACK implementation, but only on go1.5rc1 and only
on the travis builder (I have repeated the failure consistently on
travis, but cannot get a hang on my local machine with the same source
and deps - github.com/xianyi/OpenBLAS in this case).

An example of this is here: https://travis-ci.org/gonum/lapack/builds/75132168

Basically, it seems that the cgo call never returns.

Does anyone have any suggestion about why this might be? I have some
suspicions that the hardward that is being used on the travis builders
differs from our local hardware since we see different pure C called
OpenBLAS LAPACK behaviour on that infrastructure than locally:

https://github.com/xianyi/OpenBLAS/issues/610

A hacked together view of /proc/cpuinfo is visible here:
https://travis-ci.org/gonum/lapack/jobs/73869174

[Sebastien Binet]

Dan,

gonum/lapack is running tests on the "legacy infrastructure" of
travis-ci, ie: on VMs
(FYI, the new infra is container-based).
so it might be a syscall thing?

(not sure if I am helping out or have just muddied the waters some more...)

-s

[Dan Kortschak]

That wouldn't surprise me. You can see in the failure that the cgo call never returns, sitting there waiting for 9 minutes.