Travis test failure on XCODE5 about grid

[paboyle]

annoys the hell out of me too. We are doing several builds in one, and splitting clearly right
thing. Haven't seen easily how to though, but didn't really try yet to be honest.

[aportelli]

It's seems that it is going through now. Maybe they upgraded their VM to something faster. I'll close that, we can reopen the issue if the problem comes back.

[coppolachan]

Reopen this, 'cause it looks that the problem is not gone, see the failure of my last commit
https://travis-ci.org/paboyle/Grid/builds/164965163

[aportelli]

Aaaaaah. Ok but it looks different, the job didn't run out of time but was killed because it stalled for more than 10 min. I restarted it, maybe we were just unlucky.

[coppolachan]

it is not the first time I see the Xcode5 gcc version stalling in this way. Not sure why it is happening only for the gcc compilation, the VM should be the same as for the clang compilation.

[aportelli]

It passed, it looks like it was just an accident. Let's see how it goes from here.

[coppolachan]

good,
let's monitor for few more commits.

[paboyle]

Following discussion with Guido today I dropped the MacOS GCC build.

We test:
MacOS and Ubuntu

GCC (2 versions) and Clang.

The OS varies the sys headers, but these need not be varied against both compilers.

Of course, if the MacOS / gcc were reliable I would include it, but think on balance that
doing both OS's and all three compilers on Linux is sufficient.