-
Notifications
You must be signed in to change notification settings - Fork 17
Unify options for compilation of Matlab-dependent packages #65
Comments
I rather add (instead of substituting) the |
I like a lot the idea of automatic installation. I agree with @francesco-romano that leaving the users finer options may be useful (e.g. controllers may downloaded in other places). |
I would prefer to reduce as much as possible the complexity of both the cmake options available to the users and the I agree that finer control of which projects are enabled on the superbuild must be useful, but this applies to all the projects, not only the matlab ones. cc @DanielePucci I intentionally omitted the |
If we are really interesting in giving the user fine control of which package compile or not, I would work on this upstream on YCM, instead of manually hard coding a lot of options: robotology/ycm-cmake-modules#8 . |
Matlab does not find QPOasis. It seems to me that the option
does not enable the MATLAB binding
of the project external/qpOasis. Aren't I right? @traversaro @francesco-romano |
You are right. We should add in https://github.com/robotology/codyco-superbuild/blob/master/cmake/BuildqpOASES.cmake an option to enable the |
@nunoguedelha will take care of this modification :) |
After a short conversation with @francesco-romano on this topic, we agreed on clarifying the requirements before any further implementation. So, enabling
or
(These use cases do not address the possibility for users to fine control the individual superbuild components options). I would go for the first option. Was this your choice? |
I would go for 1) ! |
I would go for 1) also. I think that if @francesco-romano agrees, the poll is over and we can proceed with the modification :) |
ok great. For validation I need to build with this option ON, but I'm getting a compile error when linking |
Yes it is using an experimental version of Swig, for this reason I commit the generated files (and I don't ask the user to generate the bindings, but just to compile them). Probably I forgot to regenerate the bindings on my local machine. (FYI: to generate the iDynTree Matlab bindings, you should compile from source a swig experimental version (as in https://github.com/casadi/casadi/wiki/matlab) and then generate the bindings to commit with the Let me reboot on Linux and I will check. |
No, I actually don't have any problem on my computer with the current iDynTree master branch... |
build command : error output:
I get the same issue when building target ALL_BUILD. |
in iDynTree, I have the variables : and get the same issue... |
Mhh I should probably generate the bindings on an old version of Matlab (such as 2012b) to be sure of being forward compatible (or perhaps is an OS X-specific issue)... I guess for the time being we can remove this line: https://github.com/robotology/codyco-superbuild/blob/master/cmake/BuildiDynTree.cmake#L24 so we can compile the matlab support in the superbuild without problems. |
Any news on this? @nunoguedelha |
I have a branch ready but I'm not able to push. Could someone grant me the permissions? |
@drdanz can you add @nunoguedelha to the group https://github.com/orgs/robotology/teams/codyco-developers ? Thanks! |
while waiting for the permission, I forked the repo and pushed the branch here: |
Well done 👍 |
@nunoguedelha You are listed as pending, you should be able to join the group either by accepting the invitation on the mail you should have received, or by going to the group page and accepting (I think), otherwise I can try cancelling the invitation and inviting you again. |
Sure. I didn't receive the mail, I'll try the group page... |
What about the status of this issue? |
We disable Matlab compilation for iDynTree because we had some issue, but now we can enable it and close the issue. |
Based on the experience in the past months, with @francesco-romano we propose to modify the option in this way: |
I have yet another proposal based on how users use this options (to be implemented in #113 ): Just keep As far as I could see, the only problem difference is among users is if they want to download the controllers as part of the superbuild or not, so let's have an option only for that, and cut the unnecessary complexity. |
As for suggestion of @francesco-romano , I left the See ae316f6 |
We are starting to have a lot of Matlab-dependent packages (WBI-Toolbox, mex-wholebodymodel, now qpOASES bindings). Given that the main two possible user configuration are "build everything except what requires Matlab" or "build everything, including what requires Matlab", we should simplify things. To streamline installation, I guess it would make sense to drop the project specific superbuild options (
CODYCO_USES_WBI_TOOLBOX
,CODYCO_USES_MEX_WHOLEBODYMODEL
, etc.. ) and just have a singleCODYCO_USES_MATLAB
(suggestions for a better name are welcome) that enables download/compilation of everything that requires Matlab.Eventually we can automatically check if Matlab is available and enable this option if Matlab is found.
cc @francesco-romano @naveenoid @jeljaik
The text was updated successfully, but these errors were encountered: