Fix build from source with Clang v4

[jgoldfar]

(makes ordered comparison between pointer and zero an error)

[jgoldfar]

Not sure how this would be tested through CI; would perhaps be worthwhile to be able to select a BinDeps provider manually and test more combinations? Homebrew install was not working locally, hence this fix

[tkelman]

This isn't the right fix, the comparison shouldn't be on the pointer but on the dereferenced value, as was done upstream in coin-or/SYMPHONY@907c0be

[jgoldfar]

Fair enough (and evident in retrospect.) Worth updating this PR or close?

[tkelman]

Happy to merge if you want to update, whether it's worth it is up to you. See discussion in #44

[jgoldfar]

PR is updated (but this code path remains untested in any automated way...)

[tkelman]

Suppose the homebrew provider could be put behind an env var flag so it would be possible to disable without needing code modifications, if you wanted to get fancy

[jgoldfar]

My comment above is mostly a dupe of JuliaPackaging/BinDeps.jl#334, but I wonder what the point of such an option would be if not also tested. Not that every project even could build on e.g. Travis before timing out; there's definitely merit to having one provider that works for 99% of cases. Looking forward to BinaryBuilder making our lives easier!
PS many thanks for your contributions to Julia and the surrounding ecosystem

[tkelman]

You would want to test things naturally, but Travis can't really do some of the more obscure cases like freebsd, and non-ubuntu distros or non-default configurations of julia would require more docker work than most people probably want to undertake.

Thanks for the kind words. I'm hardly involved much any more, hence the few month gap where I wasn't looking at things like this.