On Wed, Aug 10, 2011 at 3:08 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Peter Geoghegan <peter@2ndquadrant.com> writes:
>> On 10 August 2011 01:35, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>> Actually, I'm nearly done with it already. Perhaps you could start
>>> thinking about the other polling loops.
>
>> Fair enough. I'm slightly surprised that there doesn't need to be some
>> bikeshedding about my idea to treat the PGPROC latch as the generic,
>> per-process latch.
>
> No, I don't find that unreasonable, especially not since Simon had made
> that the de facto situation anyhow by having it be initialized for all
> backends in proc.c and set unconditionally by some of the standard
> signal handlers. I am working on renaming it to procLatch (I find
> "waitLatch" a bit too generic) and
That was the direction I wanted to go in anyway, as you guessed.
> fixing a bunch of pre-existing bugs
> that I now see in that code, like failure to save/restore errno in
> signal handlers that used to only set a flag but now also call SetLatch.
Thanks for looking at the code; sounds like we nipped a few
would-have-been-bugs there.
--
Simon Riggs http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services