Remove the friend declaration with an attribute

[kingcrimsontianyu]

Description

The logging_resource_adaptor has a friend declaration (which is not a definition) preceded by an attribute, whereas C++ standard requires that such declaration must be a definition. gcc 11.4.0 used in the dev container does not correctly identify this, but gcc 12.1 and newer are able to, and will report a compile error (converted from a warning) when building RMM:

an attribute that appertains to a friend declaration that is not a definition is ignored

This simple PR removes the friend declaration from the logging_resource_adaptor class. This is valid since in recent releases of RMM the friended function no longer accesses any private (or protected) data or methods of logging_resource_adaptor.

closes #1668

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[wence-]

/ok to test

[harrism]

Thanks for the fix and for noticing that we no longer need this friend declaration!

Just wondering about the cmake test GPU configuration changes.

[harrism]

/ok to test

[harrism]

I don't think adding a single signed commit is enough. There have to be zero unsigned commits in the history of the PR.

[harrism]

/merge

[harrism]

Blocking merge while we discuss the GPUS 1 PERCENT 100 further.

[bdice]

Thanks for reverting the CMake changes for now — we can explore improving the tests in a follow-up.

[harrism]

Thanks!