CentOS 5 release machine changes

[rvagg]

I'm noting this here in an issue because I don't have time to mess with the Ansible configs while I push forward on v0.12 and v0.10 builds.

/etc/sysconfig/jenkins now has this JENKINS_ENV="PRE_1_PATH=/home/iojs/bin:$PATH ARCH=...., and the Jenkins config for the release jobs overwrite PATH if PRE_1_PATH exists and it's a <= v0.12 build.

[rvagg]

Also, xz is needed on these machines as we discovered today #258

[jbergstroem]

Is this how we juggle gcc?

[rvagg]

if [[ "X$PRE_1_PATH" != "X" && $NODE_VERSION =~ ^[0] ]]; then
  export PATH=$PRE_1_PATH
fi

yep

[jbergstroem]

its almost like we should invert this? i mean, we do the exact same thing to "enable" gcc48.

[rvagg]

maybe but I've taken the approach that >=v4.x releases are going to be more frequent and will soon be the only releases we're pushing and that <=v0.12 releases are the exception

[jbergstroem]

Yeah, its just that double negation is kind of awkward. Perhaps create functions available through env that's called gcc48 and gcc41 we call regardless of default?

[rvagg]

we can easily add POST_1_PATH and do the same there, I'd be happy to do that. Not so sure about using gcc version numbers because they are not consistent across the *nix environments we have.

[jbergstroem]

@rvagg i didn't see this being used everywhere, just the exceptions which are centos5 and 6(gcc44, right?).

[rvagg]

mm, in practice it's turned out to be the only thing needed for <=0.12

[jbergstroem]

How about we do this instead:

if [[ $NODE_VERSION =~ ^[0] ]]; then
  export PATH=${PATH//\/opt\/rh\/devtoolset-2\/root\/usr\/bin:/}
fi

..then remove all the PRE_ variable stuff? I don't see above path changing anytime soon, and if it doesn't match we should be fine anyway.

[rvagg]

fine, I guess. would be nice to be testing 0.10 and 0.12 with vanilla compilers on wheezy but since I don't think we're doing that then whatever.

[jbergstroem]

Just implemented this. If things blow up for 0.x you know who to blame.