Re: BUG #15804: Assertion failure when using logging_collector with EXEC_BACKEND

Michael Paquier <michael@paquier.xyz> writes:
> On Tue, Sep 10, 2019 at 10:58:52AM -0400, Tom Lane wrote:
>> Here's a version that does it like that.  After studying things a bit,
>> it seemed like a good idea to also move RemovePgTempFiles() down,
>> so that that possibly-slow step still occurs after making the external
>> PID file.  Everything else that we've moved to before that step
>> should be reasonably quick.

> Good point.  Nothing more to add with this version.  Thanks for
> caring.

I did one final round of testing and noted that in a Linux EXEC_BACKEND
build, this last change causes the race condition that was noted upthread:
RemovePgTempFiles could remove the parameter file that the syslogger
process is supposed to consume.  So I split off the clearing of that
subdirectory as we'd discussed, and pushed it.  Thanks for reviewing!

            regards, tom lane