> My present theory about the cause is that the backend lost its CPU
> quantum immediately after doing the send() that responded to the last
> INSERT, and was interrupted by SIGQUIT before it could continue.
> If you look at pq_flush() you'll see that it does not reset
> PqSendPointer until it exits. This means that when quickdie()
> pushes the WARNING into the send buffer, it will be added onto
> what is there --- and then the pq_flush call at the end of ereport
> will transmit that bufferload a second time.
I have applied a patch to prevent this scenario in 7.4 and 8.0 branches.
Next question is whether your original problem was due to this, or
something else. As far as I can see, this problem occurs *only* when
a SIGQUIT arrives at just the right time, so it seems unlikely to
explain any problems you'd have seen in the field.
regards, tom lane
In response to
pgsql-hackers by date
|Next:||From: Bruce Momjian||Date: 2004-09-26 00:58:37|
|Subject: Re: Use of zlib|
|Previous:||From: Bruce Momjian||Date: 2004-09-25 22:58:01|
|Subject: Re: 8.0.0beta1: Ownership of implicit sequences after dump/restore|