Compile issue on Ubuntu 20.4 and Windows 10 with RTX3090, cuda 11.1 and Optix 7.1 about pbrt-v4 HOT 16 CLOSED

I see the ubuntu one as well (pretty much the same set up here locally). It eventually build if you run make a few times in a row, the issue seems to be one of a missing dependency.

This seems to have been caused by ec742b2, so I've just reverted that for now--I haven't been able to see what the issue is, given mediocre cmake skills over her. (cc @pierremoreau, since it'd be nice to have that in there otherwise.) Please let me know if it builds for you with top of tree now.

from pbrt-v4.

For the windows part, are you using MSVC2017? If so, I think this is the same as the (yet unresolved) issue #44.

from pbrt-v4.

from pbrt-v4.

I saw nothing in your output that indicated which version you are using. I asked about 2017 because the symptoms are the same as an issue that had seemed to be VS2017 only.

from pbrt-v4.

Under ubuntu I can now build top of tree. Killeroos-gold 8.9s.

from pbrt-v4.

As to the Windows build I saw issue #44 but it appeared to be resolved by going to VS2019. I'm definitely running VS2019.

from pbrt-v4.

Mmh, I wonder if instead of depending on the file itself, it would work better if the dependency was on the library, i.e. changing this line from DEPENDS $<TARGET_OBJECTS:${lib_name}> to DEPENDS "${lib_name}".

from pbrt-v4.

I'm still working on it, but #57 seems to be building fine with Visual Studio 2019 and the built-in CMake support; the only issue I'm having are those "definition of dllimport function not allowed" messages from the dependencies though they go away if you re-run CMake and build again, and seem to happen on the current master.

I'll try to test it on Linux later today.

from pbrt-v4.

#57 is building just fine with both Ninja and Make on Linux for me, so hopefully it does for you as well.

from pbrt-v4.

from pbrt-v4.

Could you please try with the branch in #57 (since I'm assuming that by top of tree, you mean master)? It won't fix the Windows build, but I would like some additional feedback that it does not break Linux builds.

from pbrt-v4.

from pbrt-v4.

With #57, I am referring to the changes suggested in that MR which is some version of master + an additional commit. You can check it out by doing (assuming origin is the name of the remote for this repository)

git fetch origin pull/57/head:BRANCHNAME && git switch BRANCHNAME

which will create a new branch named BRANCHNAME with the content of the MR.

from pbrt-v4.

from pbrt-v4.

Awesome, thank you for testing!

For Windows, the workaround I found was to force-save one of the CMakeLists.txt files to force CMake to re-run, and then it would build just fine. But obviously that is not a great solution.

from pbrt-v4.

...and now that MR is merged, so I'll go ahead and close this. Thanks, @pierremoreau!

from pbrt-v4.