Split fmt and spdlog dependencies #842
Conversation
In order for nvcc (CUDA compiler) to not fail the DAPHNE build after switching to C++20, we introduce fmt as a separate dependency and upgrade the spdlog to the tip of the spdlog default branch as of 2024-10-03 for the time being (we'll switch back to a release once that compiles without troubles) because the released spdlog 1.14.1 did not compile with the newer fmt 11.0.2. Closes #842
corepointer
force-pushed
the
842-split-fmt-spdlog
branch
from
3800ebe to
1f4bf92
Compare
In order for nvcc (CUDA compiler) to not fail the DAPHNE build after switching to C++20, we introduce fmt as a separate dependency and upgrade the spdlog to the tip of the spdlog default branch as of 2024-10-03 for the time being (we'll switch back to a release once that compiles without troubles) because the released spdlog 1.14.1 did not compile with the newer fmt 11.0.2. Closes #842
corepointer
force-pushed
the
842-split-fmt-spdlog
branch
from
1f4bf92 to
97300d8
Compare
In order for nvcc (CUDA compiler) to not fail the DAPHNE build after switching to C++20, we introduce fmt as a separate dependency and upgrade the spdlog to the tip of the spdlog default branch as of 2024-10-03 for the time being (we'll switch back to a release once that compiles without troubles) because the released spdlog 1.14.1 did not compile with the newer fmt 11.0.2. Closes #842
corepointer
force-pushed
the
842-split-fmt-spdlog
branch
from
97300d8 to
40a6e2b
Compare
|
The CI build failure seems to be unrelated and the tests run fine on my laptop and the scale-up box. I'll proceed with the merge no not cause any further troubles (see #843). |
In order for nvcc (CUDA compiler) to not fail the DAPHNE build after switching to C++20, we introduce fmt as a separate dependency and upgrade the spdlog to the tip of the spdlog default branch as of 2024-10-03 for the time being (we'll switch back to a release once that compiles without troubles) because the released spdlog 1.14.1 did not compile with the newer fmt 11.0.2. Closes #842
corepointer
force-pushed
the
842-split-fmt-spdlog
branch
from
40a6e2b to
596b9e1
Compare
|
Updating the docker images right away to accommodate this change. At least for x86-64. ARM folks please get in touch if you run into troubles that you can not resolve on your own 🙈 |
Hi @corepointer , we have a user that runs into this problem with |
|
I don't have a working ARM development environment at hand to do this quickly (my ARM VM from the OCI always-free tier that I used for this task is occupied with something else atm). Maybe that user of yours can try running the container build scripts themselves (put an "exit" into the container build script after the BASE dev container and before the CUDA stuff). |
In order for nvcc (CUDA compiler) to not fail the DAPHNE build after switching to C++20, we introduce fmt as a separate dependency and upgrade the spdlog to the tip of the spdlog default branch as of 2024-10-03 for the time being (we'll switch back to a release once that compiles without troubles) because the released spdlog 1.14.1 did not compile with the newer fmt 11.0.2.