fixes for Julia v0.7

[fredrikekre]

No description provided.

[coveralls]

Coverage remained the same at 100.0% when pulling 163461a on fredrikekre:fe/0.7 into df96c3c on JuliaDiff:master.

[fredrikekre]

Is this OK or is it worth the trouble to keep this compatible with 0.6?

[jrevels]

Let's leave this uncommented.

[fredrikekre]

Why? nightly is closer to v0.7.0 than v0.7-beta (which the 0.7 travis target points to).

[jrevels]

Wait, -nightly and -0.7 are the same right now for Travis?

I feel that's it bad form to predicate CI pass/fail on unfixed upstream dependencies, since you can't really expect contributors to update the package for every new upstream commit. In contrast, fixing the upstream dependency to a specific tag (v0.7-beta, v0.7-rc, etc.) is a clear specification of what contributors intend to support, such that upgrading requires explicit intervention/testing.

But if -0.7 is actually targeting -nightly then that whole philosophy falls through I guess 😛

[jrevels]

Well, regardless, if indeed -0.7 targets the same commits as -nightly, it should still be fine to uncomment nightly as an allowed failure in the build matrix, since the same failures would still be captured for the 0.7 runs (assuming I understand Travis' build system correctly).

[fredrikekre]

- 0.7 points to the latest release of the 0.7 series, which currently is 0.7-beta, but soon will be 0.7-rc1 or w/e so it is not the same as nightly which is the front of julia master.
Reason for commenting out - 0.7 here is that StaticArrays is broken since it includes the Statistics stdlib as a dependency (my fault 😛) but that stdlib is not in the beta release, it was merged some commits after. That was the ForwardDiff PR.

[fredrikekre]

Now passes on nightly and 0.7-beta2.

[fredrikekre]

bump

[jrevels]

Thanks again!