Updates for Julia 0.7, drop 0.5 and 0.6

[ararslan]

No description provided.

[ararslan]

Looks like the remaining 0.7 failure is the fault of Tk:

ERROR: LoadError: LoadError: LoadError: LoadError: LoadError: UndefVarError: JULIA_HOME not defined
Stacktrace:
 [1] init() at /home/travis/.julia/v0.7/Tk/src/tkwidget.jl:35

[tknopp]

@ararslan this is extremely interesting. If 0.7 is running: what do you get timings for the first plot (see JuliaLang/julia#27083 for the issue I am talking about)

Regarding the remaining 0.7 fault I would simply disable Tk.

[tknopp]

After #288 is merged it would be interesting if tests are passing then @ararslan. Lets get this package all green in CI testing!

[tknopp]

@ararslan: I have some further fixes for your branch aa/updates. Can you give me commit access so that I can push them?

[tknopp]

@ararslan: thanks!!!

Are you fine with dropping 0.6 support in Winston?

[ararslan]

Yes, and I think it is likely necessary in order to support 1.0 fully.

[tknopp]

Yep, but I will need help for non-trivial issues in Gtk:
JuliaGraphics/Gtk.jl#358

The other really urgent thing is to fix the Homebrew issue that makes Cairo.jl failing. This propagates through all these packages.

[tknopp]

@ararslan: I still do not have commit access here in Winston, although I am team member of JuliaGraphics. I think you need to "manage my access" if you click on me in the group menu.

[ararslan]

You should have write access now

[tknopp]

great works. I have create release-julia-0.6 as our 0.6 branch. Will push my fixes to aa/updates and push that branch to master then, since the tests are now passing.

[ararslan]

Sure, just want to make sure it passes CI before merging to master

[tknopp]

Of course. Although I would only look at 0.7 linux, since 1.0 and mac will not pass

[tknopp]

(funnily enough I am doing this on a mac, but fixed the Homebrew thingy manually)

[tknopp]

very interesting. Tests pass locally on Mac, but they fail on Travis/Linux. Seems an issue with file loading.

[tknopp]

there is a timing issue going on, where a global variable is overwritten. That at least my guess. I cannot reproduce this on Mac but on linux.

[tknopp]

Tracking down this bug has been extremely annoying, but now I have it.