Depends on symengine/symengine#782

The changes look good, +1 if tests pass. You should probably update the hash to the latest master.

All tests pass, except the Sage ones, that it can't find GMP.

I think we need to set -DCMAKE_PREFIX_PATH in setup.py to point to $SAGE_LOCAL, then it should find it. Do you know a best place to do this in the setup.py? We should just add it on a command line to setup.py only when Sage is used.

Any ideas how it was finding GMP before?

@certik, sage issue is because of the new sage archive's naming extracted folder name is different. It's SageMath now. I fixed this in another branch, but there seems to be a problem with sage's gcc. Let me get back to you after checking

I think they use gcc 4.9.2 or so. If there are any compilation issues, then we need to fix them.

But why cannot we use the old version of Sage for this PR that worked? We can port to new Sage in a new PR.

Old sage version are not available anymore in the link. Maybe they are archived somewhere

I can't find it either. That's terrible.

How is it that this line

We should probably call it without the source.

Because set -e isn't added to the top of install_travis.sh. source has to be used because the script is setting up the environment.

I tried to fix it in #41, but I think there is an issue with setting set -e and sourcing it, because then it propagates into the parent shell and actually makes Travis fail. So the right solution is to install things as a regular script with set -e. And then just have a tiny script that sets up environment (without set -e), and that one will be sourced.

Because when installation fails, we should simply fail immediately.

You can also do set +e at the end

Except when it fails, then set +e will never be executed. But then I guess it will fail anyway, so maybe that's ok.

Set LD_LIBRARY_PATH to $SAGE_LOCAL. That should fix it.

@isuruf let's get the Sage stuff out of this PR, and merge #42 which restores the old Sage version and all tests pass there. In this PR let's just update the SymEngine and make sure everything passes.

We can then update Sage in a new PR.

I like the changes up to 95d4aa0, did anything fail at this point (I am testing it in #43)? Let's fiddle with Travis in another PR, as we talked, we have to be a bit more careful here.

Closing in favor of #45.