Bruce Momjian <maillist@candle.pha.pa.us> writes:
> Applied.
>
>> Just in case you'd like to see what I was talking about, I am attaching
>> my patch to src/interfaces/libpq/fe-exec.c to prevent utility functions
>> called from SPI from locking up the client.
Uh, I didn't actually believe that that patch was a good idea. Hacking
libpq to survive a protocol violation committed by the backend is *not*
a solution; the correct answer is to fix the backend. Otherwise we will
have to discover similar workarounds for other clients that do not
use libpq (ODBC, for example).
Please reverse out that patch until someone can find some time to look
at the issue. (I will, if no one else does, but it would probably be
more efficient for someone who already knows something about SPI to
fix it...)
regards, tom lane