Re: [HACKERS] possible self-deadlock window after bad ProcessStartupPacket

On Fri, Jul 20, 2018 at 4:53 AM, Heikki Linnakangas <hlinnaka@iki.fi> wrote:
> ISTM that no-one has any great ideas on what to do about the ereport() in
> quickdie(). But I think we have consensus on replacing the exit(2) calls
> with _exit(2). If we do just that, it would be better than the status quo,
> even if it doesn't completely fix the problem. This would prevent the case
> that Asim reported, for starters.

+1 for trying to improve this by using _exit rather than exit, and for
not letting the perfect be the enemy of the good.

But -1 for copying the language "if some idiot DBA sends a manual
SIGQUIT to a random backend".  I think that phrase could be deleted
from this comment -- and all of the other places where this comment
appears already today -- without losing any useful informational
content.  Or it could be rephrased to "if this process receives a
SIGQUIT".  It's just not necessary to call somebody an idiot to
communicate the point.

-- 
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company