-
-
Notifications
You must be signed in to change notification settings - Fork 23
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Avoid setting CFLAGS, CPPFLAGS, and CXXFLAGS #59
Conversation
This is addressed in conda-forge/docker-images#134
Hi! This is the friendly automated conda-forge-linting service. I just wanted to let you know that I linted all conda-recipes in your PR ( |
@conda-forge-admin, please rerender |
…da-forge-pinning 2020.05.14.17.14.20
@isuruf @h-vetinari @jakirkham Either conda-forge/docker-images#135 or conda-forge/docker-images#136 caused all CUDA builds to fail here, see https://dev.azure.com/conda-forge/feedstock-builds/_build/results?buildId=163080&view=results. In this feedstock, we rely on the fact that |
I found there're a bunch of recent efforts on CF that attempt to add |
OK, tests on CUDA 9.2 passed, because there is no |
@jakirkham as a workaround I kinda undo #53 in 0ad5edb. |
It wouldn't overwrite it. Do you have a log where it would overwrite? |
@leofang
It seems like relying on a failed import does not yield much information regarding the soundness of the package - is there no other way to do this? In fact, I have the same problem (of a GPU device being necessary) in conda-forge/faiss-split-feedstock#1 to actually test the code, but that can only be solved with CI-instance that have GPUs.
I agree with @isuruf that this wouldn't happen. The default installation for cuda drivers is in |
@isuruf @h-vetinari Thanks for replying!
Right, but as I said, this effort at most makes something simple like
I second your request, although I have mixed views. I already left a comment in the issue conda-forge/conda-forge.github.io#1062 you opened.
On the machines I have access to, they are always in |
@leofang, in the docker image, where are the |
@isuruf Not sure what's this for, but here you go.
one of our local machines:
|
(By the way, how to get root access in the docker images? |
@h-vetinari, how about we add the existing paths in front of the cuda compat directory.
? |
Sure. I don't know the precedence of
I don't think you can. The entrypoint script switches to a user |
Entries in |
Soooooo, one more PR for the dockerfiles, I'm guessing. |
Noticed something weird with our CI, let me check here... |
NVM. I made a mistake in my local tests. This is good to go. |
1. libcudart is loaded at import 2. import is tested
Thanks, @isuruf, for suggestions! The PR is ready to go. Let's wait a little to see if @jakirkham has any feedback. |
Sorry was out last week. So not caught up on all of the changes that went in. Could someone please summarize? 🙂 |
Hi @jakirkham, in conda-forge/docker-images#134 the CUDA headers living outside of the CUDA root were symlinked back, so your Moreover, Another change to the docker images was to deal with the linker discovery order, but I don't think it is critical here. |
Close #9, close #58.
Checklist
0
(if the version changed)conda-smithy
(Use the phrase@conda-forge-admin, please rerender
in a comment in this PR for automated rerendering)