Punting on this for now.... going with major version bump instead.

Doing this a slightly different way, abandoning this rev.

Just revert, kill the other changes.
Commit message locally updated to remove bit about whining, that will be a separate,
future commit.

Just revert, kill the other changes.
Commit message locally updated to remove bit about whining, that will be a separate,
future commit.

Just revert, kill the other changes.
Commit message locally updated to remove bit about whining, that will be a separate,
future commit.

In D45875#1045923, @kib wrote:

Then what about sx_try_xlock()?

Why didn't sx_lock() complain?

I support moving forward with this plan. The risk seems to be in the new jail stuff, which should make some flakey tests today more reliable, so what's not to love about that?

And something like https://reviews.freebsd.org/D45868 for the compat stuff... It's a little different than I originally thought, so there will need to be a new ioctl to cover this case since the sizes stay the same, we have to change the number so we don't overflow the receiver's conf buffer if it's run against a new kernel. I suspect the new ioctl# and the freebsd14 ioctl parts from that differential belong here, and then you can take mine for compat, make sure it works and then commit it maybe (or if you want me to, let me know)

So something like https://reviews.freebsd.org/D45867 for the rename (I can commit this if you'd like)

Thanks for removing this from all the old versions...

Updated the commit message locally...

gcc 11 and newer does the right thing, so just test for that. gcc 10 no longer
is supported as a 'freebsd' build tool, so we don't have to check for the
internal version number.

In D45836#1045523, @jhb wrote:

Compiler explorer does not have pcc, but tcc with -Wall doesn't raise a warning either, so instead I think what you want is something like this:

/*
  * Like __printflike, but allows fmtarg to be NULL.
  * GCC versions prior to 11 explicitly warned for NULL format strings, but other
  * compilers do not.
  */
#if defined(__clang__) || __GNUC_PREREQ(11, 0)
#define  __printf0like __printflike
#else
#define __printf0like(fmtarg, firstvararg)
#endif

And I will just drop my patches to GCC 12+ to add printf0. This could really benefit from a "is actually GCC for real and not some other compiler pretending to be GCC" helper so that we could truly stub out only for real GCC < 11.

In D45434#1045067, @jhb wrote:

@imp did you get a chance to test this?

Oh, didn't hit submit before, though this doesn't affect the commit

update with jhb's suggestions. Hopefully I got them right

simplify, and move some info into style(9)

Update per jrtc27

In D45825#1044897, @rlibby wrote:

In D45825#1044894, @imp wrote:

Seems fine... what compilers / build env hits this?

CROSS_TOOLCHAIN=amd64-gcc13 make buildworld produces about 120 warnings like this one:

gcc13/usr/src/freebsd/amd64.amd64/bin/ln/ln.o.meta:/usr/src/freebsd/bin/ln/ln.c:248:26: warning: left-hand operand of comma expression has no effect [-Wunused-value]