Can one of the admins verify this patch?

ok to test

Add --disable-tests to allow the user to disable tests. As a side-effect, this will avoid the following build failure when check is found:

I have no objections to add --disable-tests, tho:

libstat_wrapper.c:11:10: fatal error: gnu/lib-names.h: No such file or directory
   11 | #include <gnu/lib-names.h>
      |          ^~~~~~~~~~~~~~~~~

This build failure is raised since version 2.0.5 and 78df90b

this part I don´t understand. We run CI on each build and we have never seen this problem before.

What distribution / container / platform are you building on? Is the version of check too old?

The build failure is raised by buildroot autobuilders.
check is in version 0.15.2 but check doesn't provide gnu/lib-names.h.
This file is glibc-specific, it will raise the above build failure with uclibc(-ng) and musl.

I think it would be better just to disable that test. It's a good thing to test libqb against other libcs where possible.

Indeed, you could disable or remove this test however it still makes sense to disable all tests to save build time and disk space.

The build failure is raised by buildroot autobuilders. check is in version 0.15.2 but check doesn't provide gnu/lib-names.h. This file is glibc-specific, it will raise the above build failure with uclibc(-ng) and musl.

Apologize for my complete ignorance, but what is buildroot.org?

+1 on disabling the test when gnu/lib-names.h is not available.

Regardless of the libc is in use, test suite should be built and executed. Build only doesn´t help much for the final results. Shaving 3/4 seconds to build is meaningless if the result doesn´t work on the libc in use.

Why is gnu/lib-names.h even included? I don't see anything from this header file that gets used in libstat_wrapper.c. Removing the gnu/lib-names.h solves the build issue.

I'll remove it. I suspect it was needed for some earlier OS version, but without that it's passed all the CI tests, so it's obviously fine without!