CMake update - Githubissues

Tudat / tudatBundle

NOTE: This Tudat version is no longer supported. See https://docs.tudat.space/en/stable/ and https://github.com/tudat-team/tudat-bundle for the new version

http://tudat.tudelft.nl

BSD 3-Clause "New" or "Revised" License

11 stars 113 forks source link

CMake update #9

Closed magnific0 closed 7 years ago

magnific0 commented 7 years ago

Part of larger CMake overhaul. New features include

The CMake modules path (CMAKE_MODULES_PATH) is now appended everywhere instead of overwritten.
Multiple paths can be contained, for instance when a project includes another project and both add an entry and all will be referenced if a CMake module is required (include or find_package).
If the same module exists in both locations, the parent module has priority (earlier addition to the variable).
A README file was added to the directories containing CMake modules to notify users of the latter behaviour and general purpose of the folder.
CMake modules that were not used by project were removed.
A collection of unique FindXXX.cmake modules is now located in tudatBundle/external/CMake
Copyright comment headers were fixed.
The CMake notice messages of the project name, release type and whether it detected being included from another project are not combined in a single message.
A flag is introduced to force Apple machines to use the include_directory command (just like Windows and Linux).
This flag is put in the Tudat bundle CMakeLists.txt and turned on by default.
add_boost is rewritten to use a separate cache dir (for downloading, unzipping and building).

magnific0 commented 7 years ago

Just noticed that I managed to pull in the new eigen sources (again). @DominicDirkx can you reset them to the version as you did before?

DominicDirkx commented 7 years ago

Hi Jacco,

Thanks for this update. The code compiles/runs fine on Ubuntu 14.04 gcc/clang.

Under Windows I got an error from cmake on this line:

file(RENAME ${ExtractedDir} ${BoostSourceDir})

Complaining that the folder didn't exist. However, I can't seem to reproduce the problem, it runs fine now. Perhaps it's an issue of it caching some variable after cmake terminates,, which is then used on the following iteration?

I'll try to get it to reproduce the error. Do you have a Windows machine where you could test this?

DominicDirkx commented 7 years ago

I'm really confused.... I did a clean install of Tudat, and everything worked like a charm. Any thoughts?

magnific0 commented 7 years ago

@DominicDirkx I assume you mean clean install on Windows. Could it be that some variable names were cached? I did change a number of variables inside add_boost.cmake.

DominicDirkx commented 7 years ago

Yes, a clean install on Windows of the tudat bundle. Did not reinstall Qt, though.

I'm running CMake from a shell now, with a different build folder and MSVC as compiler. Hopefully that should be completely independent from any cached variables from previous installs.

DominicDirkx commented 7 years ago

Well, with MSVC boost won't build, but it does move over the build step and doesn't crash until way later...

magnific0 commented 7 years ago

Do we support MSVC? I know we have some sections in add_boost.cmake and compiler.cmake that deal with this, but my reasoning for not removing those is that it would hamper future / user adaptation for MSVC.

The boost cache dir is now outside of the regular build folder. So to really start fresh, you should also remove tudatBundle/boost/build.

magnific0 commented 7 years ago

What is left to be tested? I tried clang on OS X and gcc on Linux. Is there a specific Windows test you want me to run?

DominicDirkx commented 7 years ago

Concerning MSVC - as far as I know we never intended to support it, but I figured it was a test that's as independent as possible from previous compilations.

If you could run the CMakeLists test from Qt on Windows/MinGW to see if you can reproduce my strange error, that'd be great. If it runs properly for you, then I think we're all set.

magnific0 commented 7 years ago

Great, I'll do that right away and report back.

Regarding MSVC, it's indeed been something I wanted to test for a while and see how difficult it would be to get it working for say the latest version. The MSVC IDE is pretty neat actually, so it would be a bonus to Windows users if this worked.

magnific0 commented 7 years ago

I was able to successfully build the bundle under Windows MinGW 4.9.2. I have attached my cmake log, I don't get the error/warning you mentioned.

magnific0 commented 7 years ago

I also reverted the Eigen version, so it should be ready to commit now.