On Wed, May 8, 2024 at 6:47 AM Sean Christopherson [off-list ref] wrote:

On Tue, May 07, 2024, Edward Liaw wrote:

quoted

809216233555 ("selftests/harness: remove use of LINE_MAX") introduced
asprintf into kselftest_harness.h, which is a GNU extension and needs
_GNU_SOURCE to either be defined prior to including headers or with the
-D_GNU_SOURCE flag passed to the compiler.

v1: https://lore.kernel.org/linux-kselftest/20240430235057.1351993-1-edliaw@google.com/ (local)
v2: add -D_GNU_SOURCE to KHDR_INCLUDES so that it is in a single
location.  Remove #define _GNU_SOURCE from source code to resolve
redefinition warnings.

Edward Liaw (5):
  selftests: Compile kselftest headers with -D_GNU_SOURCE
  selftests/sgx: Include KHDR_INCLUDES in Makefile
  selftests: Include KHDR_INCLUDES in Makefile
  selftests: Drop define _GNU_SOURCE
  selftests: Drop duplicate -D_GNU_SOURCE

Can you rebase this on top of linux-next?  I have a conflicting fix[*] for the
KVM selftests queued for 6.10, and I would prefer not to drop that commit at
this stage as it would require a rebase of a pile of other commits.

Ok, I'll do that.

And I doubt KVM is the only subsystem that has a targeted fix for the _GNU_SOURCE
mess.

If we want/need to get a fix into 6.9, then IMO we should just revert 809216233555
("selftests/harness: remove use of LINE_MAX"), as that came in quite late in the
6.9 cycle, and I don't think it's feasible to be 100% confident that globally
defining _GNU_SOURCE works for all selftests, i.e. we really should have a full
cycle for folks to test.

That sounds reasonable to me.  In this thread Tao suggested reverting
back to 809216233555 and using a fixed value in place of LINE_MAX to
fix 38c957f07038
https://lore.kernel.org/linux-kselftest/20240508070003.2acdf9b4@kernel.org/ (local)

[*] https://github.com/kvm-x86/linux/commit/730cfa45b5f4