use the latest tag instead of the master branch of papi in the ci#8239
use the latest tag instead of the master branch of papi in the ci#8239mherkazandjian wants to merge 3 commits intoapache:mainfrom
Conversation
|
Our current approach is to not fix versions of dependencies in the docker images. The problem is that we hit things like this. I'm not sure what we want to do as a solution. |
|
|
||
| cd / | ||
| git clone https://bitbucket.org/icl/papi.git | ||
| git clone -b stable-6.0 --single-branch https://bitbucket.org/icl/papi.git |
There was a problem hiding this comment.
This will not work as we need to include this commit (https://bitbucket.org/icl/papi/commits/a0d7cf0d5929f2115636e77317daa0f00d075e08) in order for papi to work with CUDA.
There was a problem hiding this comment.
i see. well, this is quite inconvenient at the moment. I think this error can be fixed by solving the papi compile issue
in the papi repo, that is another solution, but i am not sure how long that will take them to fix.
we could alternatively checkout the commit you refernced above instead of the stable tag (as a temporary fix)?
that would get the docker ci_cpu image build to succeed at least.
There was a problem hiding this comment.
i pointed the checkout to the latest commit in papi where the build succeeds
mherkazandjian@eb6e7d2
There was a problem hiding this comment.
papi just reverted the offending commit
https://bitbucket.org/icl/papi/issues/84/build-of-the-master-branch-is-failing-with
we can close close this PR and the issue #8173
2 participants
The CI docker build tvm.ci_cpu fails with the current master of papi
https://github.com/apache/tvm/blob/main/docker/install/ubuntu_install_papi.sh#L29
the offending papi commit is
https://bitbucket.org/icl/papi/commits/1f48bb789bb8a42e7f1e3faec4e49487af14bf46
apparently the papi CI is broken too
i tried to build the docker ci_cpu while going back in the commits on the master branch of papi
and i think the safest to use is the use the latest papi stable tag.
with this patch the tvm ci_cpu docker container builds successfully.