GTEST_ROOT is not present if not building GTest

[paskino]

https://github.com/CCPPETMR/SIRF-SuperBuild/blob/master/SuperBuild/External_GTest.cmake#L75 will not set GTEST_ROOT rather GTEST_LIBRARIES.

[paskino]

reading better this link it seems that FindGTest will (or better may???) set the GTEST_ROOT cached variable.

[KrisThielemans]

I think GTEST_ROOT "may be set" by the user, not by the FindGTest.cmake. Could check on the source.

Using GTEST_LIBRARIES is fraught with difficulties as you don't know what's in there, an din any case can't be passed on.

This is a common problem in the SuperBuild: we need to ensure that "ours" is found if we build it, but otherwise we have to rely on usual CMake stuff.

We've tried mark_as_superbuild to work-around this, but it wouldn't cope with the USE_SYSTEM_*=OFF case where the user sets GTEST_ROOT.

I suppose the thing to do is to only pass GTEST_ROOT if it is actually set (either by us, or by the user). That'd need a change here for Gadgetron.

I think the best strategy is probably to create a variable GTEST_CMAKE_FLAGS in External_GTest.cmake which would be setting GTEST_ROOT if it needs to, and otherwise be empty, and use that in dependent packages. That should work for all packages I believe.

[paskino]

I'm struggling to understand the problem...

[KrisThielemans]

maybe you're right. cases:

• USE_SYSTEM_GTEST=ON: we set GTEST_ROOT
• USE_SYSTEM_GTEST=OFF:
  • the user sets GTEST_ROOT, so we'll pass it on
  • the user doesn't set GTEST_ROOT, so we set it to empty when calling CMake for Gadgetron.

Actually seems fine. If you agree, please close this issue.

I still think that a strategy if introducing GTEST_CMAKE_FLAGS, STIR_CMAKE_FLAGS etc would be nice and clean, but that's for another issue then (if you agree with it, please create a new one)