Date: Mon, 30 Nov 2020 10:43:28 -0700 From: Ian Lepore <ian@freebsd.org> To: mmel@freebsd.org, src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: Re: svn commit: r368187 - head/sys/dev/nvme Message-ID: <53cedfd91e07e2cfc4b2f77ff0795c1d5cf68707.camel@freebsd.org> In-Reply-To: <b32dcdee-52ea-8eed-4dde-1f9f7990a54f@freebsd.org> References: <202011301451.0AUEpn9w002536@repo.freebsd.org> <5bd6eb969c3a198246ab915257375b02c7b14e0c.camel@freebsd.org> <b32dcdee-52ea-8eed-4dde-1f9f7990a54f@freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 2020-11-30 at 17:56 +0100, Michal Meloun wrote: > > On 30.11.2020 17:02, Ian Lepore wrote: > > On Mon, 2020-11-30 at 14:51 +0000, Michal Meloun wrote: > > > Author: mmel > > > Date: Mon Nov 30 14:51:48 2020 > > > New Revision: 368187 > > > URL: https://svnweb.freebsd.org/changeset/base/368187 > > > > > > Log: > > > Unbreak r368167 in userland. Decorate unused arguments. > > > > > > Reported by: kp, tuexen, jenkins, and many others > > > MFC with: r368167 > > > > > > Modified: > > > head/sys/dev/nvme/nvme.h > > > > > > Modified: head/sys/dev/nvme/nvme.h > > > ================================================================= > > > ==== > > > ========= > > > --- head/sys/dev/nvme/nvme.h Mon Nov 30 14:49:13 2020 ( > > > r368186) > > > +++ head/sys/dev/nvme/nvme.h Mon Nov 30 14:51:48 2020 ( > > > r368187) > > > @@ -1728,9 +1728,15 @@ extern int nvme_use_nvd; > > > > > > #endif /* _KERNEL */ > > > > > > +#if _BYTE_ORDER != _LITTLE_ENDIAN > > > +#define MODIF > > > +#else > > > +#define MODIF __unused > > > +#endif > > > + > > > /* Endianess conversion functions for NVMe structs */ > > > static inline > > > -void nvme_completion_swapbytes(struct nvme_completion *s) > > > +void nvme_completion_swapbytes(struct nvme_completion *s > > > MODIF) > > > > IMO, this is pretty ugly, it causes the brain to screech to a halt > > when > > you see it. Why not just add an unconditional __unused to the > > functions? The unused attribute is defined as marking the variable > > as > > "potentially unused" -- there is no penalty for having it there and > > then actually using the variable. > > > > I understand, (and I have significant tendency to agree) but I did > not > find more correct way how to do it. > Are you sure that __unused is defined as *potentially* unused? I > cannot > find nothing about this and you known how are compiler guys creative > with generating of new warnings... > I known that C++17 have 'maybe_unused' attribute, but relationship > to > standard '__unused' looks unclear. > > In any case, I have not single problem to change this to the > proposed > style if we found that this is the optimal way. > > Michal The __unused macro is defined in cdefs.h under #if __GNUC_PREREQ__(2, 7) as attribute((unused)) and the gcc docs for that attribute say "This attribute, attached to a [variable|function], means that the variable is meant to be possibly unused. GCC will not produce a warning for this variable." -- Ian
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?53cedfd91e07e2cfc4b2f77ff0795c1d5cf68707.camel>