Re: [gpsd-dev] [PATCH] Use pselect() to reduce wakeups and avoid race conditions

[Greg Troxel]

Mike Frysinger <address@hidden> writes:

> On Saturday 14 January 2012 00:43:08 Håkan Johansson wrote:
>> Regarding (p)select and signal races, I did not see the 'self-pipe trick'
>> mentioned in the discussion.  Quoting
>> http://www.kernel.org/doc/man-pages/online/pages/man2/select.2.html:
>> 
>> "On systems that lack pselect(), reliable (and more portable) signal
>> trapping can be achieved using the self-pipe trick (where a signal handler
>> writes a byte to a pipe whose other end is monitored by select() in the
>> main program.)"
>
> neat trick, but i wonder if we even have to bother.  pselect() has
> been in the Linux kernel (and thus race free) for almost 6 years.  the
> select() based code, while admittedly racy, isn't that big of a deal
> -- at worse, the code takes 1 second to respond.  i think this is an
> acceptable trade off -- install a kernel from the last 6 years, or
> suck it up ;).

gpsd does not specify Linux as a requirement; it is in theory portable
to any reasonable posix-like system (which perhaps includes windows, and
perhaps doesn't).

But, pselect is specified by posix:

  http://pubs.opengroup.org/onlinepubs/009695399/functions/select.html

and is available in NetBSD 5 (perhaps earlier) and FreeBSD from 2002
(and also in OS X 10.7, with the same FreeBSD man page), so I don't see
any issue in relying on it.

pgpK8VxiYPCZP.pgp
Description: PGP signature