On 05/02/2020 18:17, Rich Felker wrote:
> On Wed, Feb 05, 2020 at 05:06:11PM -0600, A. Wilcox wrote:
> This is not a difference in the signature; the constness in the
> declaration is meaningless (explicitly ignored by the language). In
> the definition, it causes the local variable corresponding to the
> argument to be const, but that's not terribly useful.


Sorry, I was confused and thought it was a pointer type.  Disregard.


>> It is implemented not only in glibc, but additionally in Solaris 11:
>>
>> https://docs.oracle.com/cd/E88353_01/html/E37843/pthread-sigqueue-3c.html
> 
> Without _np -- interesting. I'm not opposed if this seems to be the
> consensus.


Notably, no one in the BSD family has implemented it.  I have not found
any discussion of this interface on the Austin Group defect tracker.  It
is a very recent addition to Solaris – appearing to have been added last
year.

WebKit's JSC mentions the desire to utilise this interface in a comment
but stops short of actual use due to portability concerns.  Firefox
appears to use this interface for a test case for its sandbox, on glibc
hosts only.  Intriguingly, Xenomai does define the interface, but as
pthread_sigqueue_np.

The only seemingly-legitimate use of this interface comes from a
perplexingly wrong usage in Intel's graphics driver toolset at:

https://cgit.freedesktop.org/xorg/app/intel-gpu-tools/tree/lib/igt_core.c?id=497e13d2b4#n2465

...where they call pthread_sigqueue only if the system is /not/ Linux.
I would assume this would break the build on FreeBSD, but perhaps they
don't support these tools on non-Linux platforms.  If that's the case,
then I'm not sure why this code block even exists.

Best,
--arw

-- 
A. Wilcox (awilfox)
Project Lead, Adélie Linux
https://www.adelielinux.org