These are chat archives for symengine/symengine

16th
Aug 2015
Sumith Kulal
@Sumith1896
Aug 16 2015 10:09
Hi,
Given const Polynomial U
What is the easiest way to get RCP<const Polynomial> W of it?
Isuru Fernando
@isuruf
Aug 16 2015 10:10
rcp_from_this
Sumith Kulal
@Sumith1896
Aug 16 2015 10:12
Cool, thanks
Ondřej Čertík
@certik
Aug 16 2015 15:37
@Sumith1896 sure, what is your question
@isuruf if you need to modify the headers in teuchos in order to make it work, then yes.
Otherwise if we can make it work without modifications, that would be much better.
Isuru Fernando
@isuruf
Aug 16 2015 15:38
Nope. I got it working
To make #589 work, I had to install teuchos headers in the same PR as well
Ondřej Čertík
@certik
Aug 16 2015 15:40
@isuruf which of your PRs are ready to be merged?
Isuru Fernando
@isuruf
Aug 16 2015 15:40
I have only #589 at the moment
which is ready to be merged
Ondřej Čertík
@certik
Aug 16 2015 15:40
Ok, let me test it by hand.
Isuru Fernando
@isuruf
Aug 16 2015 15:44
cmake --find-package works only for CMake 2.8.6 or above
Ondřej Čertík
@certik
Aug 16 2015 15:45
@isuruf btw, we should stop using include_directories, but rather assign both flags and link libraries directly with targets. I started doing that in another project: https://github.com/certik/hfsolver/pull/65/files, see the target_link_libraries(hfsolver PUBLIC hfsolver::ffte) line and also the two lines starting here: https://github.com/certik/hfsolver/pull/65/files#diff-2fa8650c9becc7c6822b268fba95f078R10
It's much cleaner, since cmake then keeps track of the proper flags and link libraries for each target, and when you link the target, you do not need to worry about that.
The only problem is that it requires a more recent cmake than the one on Travis, so I install a new cmake binary there: https://github.com/certik/hfsolver/pull/65/files#diff-500f6a507006ba7116b7698c4029dc19R13, it's not a big deal.
(See also using the namespace in cmake.)
Isuru Fernando
@isuruf
Aug 16 2015 15:50
what's the downside of include_directories?
Ondřej Čertík
@certik
Aug 16 2015 15:52
That it gets included for all targets
the same with compile flags.
Sumith Kulal
@Sumith1896
Aug 16 2015 16:09
@certik That issue got resolved, I'll let you know if anything else pops up.
Ondřej Čertík
@certik
Aug 16 2015 16:32
@isuruf that's awesome, thanks for documenting it!
Sumith Kulal
@Sumith1896
Aug 16 2015 16:44
Something I came across, is this intentional?
Use of "symengine/monomials.h"
Ondřej Čertík
@certik
Aug 16 2015 16:50
@Sumith1896 what's your exact question about it?
Sumith Kulal
@Sumith1896
Aug 16 2015 16:51
Why are we using "symengine/monomials.h" instead of <symengine/monomials.h> ?
Ondřej Čertík
@certik
Aug 16 2015 16:57
Ah. That's probably a mistake.
Sumith Kulal
@Sumith1896
Aug 16 2015 16:58
No problem. I'll fix it. I am working on the benchmarks anyways
Isuru Fernando
@isuruf
Aug 16 2015 17:21
@certik, for #550, what do you think is a good service for travis-ci?
http://docs.travis-ci.com/user/deployment/
Ondřej Čertík
@certik
Aug 16 2015 22:12
@isuruf I have no idea. I think we should have a development channel somewhere. The Python wrappers should be possible to install using pip. So we can have a dev channel for the symengine.
As to the C++ library, we can use something like https://binstar.org/