added missing typename. #208
Conversation
replaced not with ! because intel compiler do not want to compile this. replaced integrate(double dt) with integrate(S dt) to allow fcl compile with float.
include/fcl/CMakeLists.txt
Outdated
| @@ -1,3 +1,5 @@ | |||
| cmake_minimum_required(VERSION 3.6) | |||
There was a problem hiding this comment.
It would be good to remove this line unless there is a compelling reason we need this. FCL is supposed to support Ubuntu 14.04 LTS, and the CMake version on Ubuntu 14.04 LTS is 2.8.12.
Removed cmake_minimum_required(VERSION 3.6).
|
CMakeLists.txt was changed by accident. But I already pushed other changes before "Moving easy to inline code from *.cpp to header files." What do you think? |
|
Why do we need this explicit inlining? As I know, the inlining decision is implicitly made by compilers if it's needed for optimization even it's not in the same compilation unit. I believe having definition in the header file would only makes headers less readable. |
|
It is not about inlining it is because of One Definition Rule. It is not needed for methods declared inside of class/struct like BV_node_base.
Well this is only a couple of lines, not that much and without it FLC is for sure hot header only because it requires *.cpp files to compile and link. Of corse another way would be to put such code into plane-inl.h files. |
|
I think there is a miscommunication here. FCL is not a header-only project yet. Sorry if I brought this confusion by the previous comment. We've templatized FCL for the scalar type so that we can use AutoDiffScalar to get derivatives (see @sherm1's suggestion and related PRs: #154, #165). The templatizing was necessarily applied to most of the code, so it seems easy to make FCL an (optional) header-only project. However, we haven't made a decision on that yet. It would be good to open a new issue about this to discuss with other contributors and postpone the inlining of this PR until then. |
replaced not with ! because intel compiler do not want to compile this.
replaced integrate(double dt) with integrate(S dt) to allow fcl compile with float.
This change is