[DONT MERGE] Travis experimentation

[tshort]

I'm trying to get travis building to work. There's other junk in here that should be pulled elsewhere.

[tshort]

This is irritating. I can't get gfortran to work right to compile dassl. I'm tempted to depreciate dassl. It's also a pain on windows. I'm not happy about that because I find dassl to be simpler and easier to fiddle with than sundials. Comments @iraikov?

[iraikov]

Actually, I also like dassl for its simplicity, and I think a bit better performance on problems with structural events. But why not just default to sundials if dassl cannot be compiled?

[tshort]

Sounds like a good approach...

On Fri, Jan 30, 2015 at 10:08 PM, Ivan Raikov notifications@github.com
wrote:

Actually, I also like dassl for its simplicity, and I think a bit better
performance on problems with structural events. But why not just default to
sundials if dassl cannot be compiled?

—
Reply to this email directly or view it on GitHub
#40 (comment).

[tshort]

The same gfortran errors noted are showing up in Ipopt, too:

jump-dev/Ipopt.jl#23

[tkelman]

You can fix the gfortran problem while keeping language: julia if you like by adding a line rm -f ~/julia/lib/julia/libgfortran.so.3. We should have this fixed for our next point release, it required building our own gcc for use on the buildbot that compiles the generic linux binaries.

[tkelman]

No guarantees it'll work, but you might also want to try @nolta's Fortran-to-Julia converter on dassl. See pass0.sh and pass1.jl here: https://github.com/nolta/SpecialFunctions.jl/tree/master/src/amos

[tshort]

Cool @tkelman! I'll have to try @nolta's converter. Thanks also for your input on language: julia. It looked like a fix was in the works.

[tkelman]

Fixing JuliaPackaging/BinDeps.jl#123 should hopefully help make this kind of thing easier to debug in the future.