This doesn't seem to be used anywhere? Other QMD_*_CHECK macros are internal macros used in queue.h directly. Is this intended to be used in consumers, or is the patch incomplete?

Sorry, I should have provided more context, and probably not have submitted this as is.

Initially, I intended QMD_STAILQ_CHECK_EMPTY() to be called by STAILQ_EMPTY() (only in the case the STAILQ is indeed determined empty), but (too) quickly inferred that this didn't seem possible without using statement expressions, which I suspected were not allowed in sys/queue.h.

Then, I began using it directly in some consumer (which is not in this series) but didn't change the name.

Ideally, I would like this check to be performed by STAILQ_EMPTY(). So, is it possible to use statement expressions in sys/queue.h? If not, would you see an alternative to make it possible? One way may be to express the check in terms of the ternary operator by wrapping panic() into an inline function returning a boolean, which should work but is not particularly elegant...

QMD_STAILQ_CHECK_EMPTY() should also be called from STAILQ_INIT(), which is easy to do as the latter already expands to a statement, and I intend to add that after your feedback to this message.

Regardless, I've experienced it is also useful to call QMD_STAILQ_CHECK_EMPTY() directly at points where the callers know that the list must be empty, which are not necessarily initializations or explicit empty tests. Re-thinking about that, I'm now finding that a better solution would be to create something like a STAILQ_ASSERT_EMPTY() macro and have consumers call it instead. This macro would call STAILQ_EMPTY() and panic on it returning false. As already evoked above, STAILQ_EMPTY() would call QMD_STAILQ_CHECK_EMPTY() to perform the additional integrity check when the list is empty. This would also have the benefit that QMD_*_CHECK() macros would continue to be called by sys/queue.h exclusively.

Does that sound like an acceptable plan?

The new version illustrates what I'd like to do. In the meantime, reading sys/cdefs.h, I saw some statement expression, so I was probably too conservative in assuming it can't be used in sys/queue.h.

As QMD_STAILQ_CHECK_TAIL() is internal, I've removed its unused field parameter, making it diverge from its QMD_TAILQ_CHECK_TAIL() model.

If by statement expressions you mean the '({ ..., a; })' GNU C extension, yes we use that a lot. It's true that <sys/queue.h> is widely used outside of the base system, but so is <sys/cdefs.h>.

If by statement expressions you mean the '({ ..., a; })' GNU C extension (...)

Exactly. With respect to the ISO standard, that's arguably the most logical name and the one GNU uses in its GCC manual.