Re: PostmasterIsAlive() in recovery (non-USE_POST_MASTER_DEATH_SIGNAL builds)

On Fri, Jun 11, 2021 at 1:18 PM Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> Heikki Linnakangas <hlinnaka(at)iki(dot)fi> writes:
> > On 09/04/2021 07:01, Thomas Munro wrote:
> >> This seems to work on Linux, macOS, FreeBSD and OpenBSD (and I assume
> >> any other BSD). Can anyone tell me if it works on illumos, AIX or
> >> HPUX, and if not, how to fix it or disable the feature gracefully?
> >> For now the patch assumes that if you have SIGIO then you can do this;
> >> perhaps it should also test for O_ASYNC? Perhaps HPUX has the signal
> >> but requires a different incantation with I_SETSIG?
>
> I took a look on HPUX 10.20 (gaur's host):

Thanks both for looking!

Unfortunately I'll have to withdraw this patch in its current form.
On closer inspection, only the last backend to start up and do
F_SETOWN on the pipe receives the signal. We'd probably have to
create a separate pipe for each backend, or something like that, which
seems unwarranted so far.

> * I found I_SETSIG, but ...
>
> $ grep -r SETSIG /usr/include
> /usr/include/sys/stropts.h:#define I_SETSIG _IO('S', 9) /* request SIGPOLL signal on events */
>
> stropts.h seems to be for a feature called "streams", which is
> probably nonstandard enough that we don't want to deal with it.

Agreed. It is technically POSIX, but optional and marked obsolescent.
It's annoying to see that I_SETSIG did allow multiple processes to
register to receive signals for the same event on the same underlying
stream, unlike F_SETOWN. Oh well.