16.03.2023 00:13, Thomas Munro wrote:
> Thanks. We were wondering if the retry mechanism might somehow be
> hiding this in non-assert builds, but, looking more closely, that is
> tied specifically to the memory reservation operation.
As to hiding, when analyzing the Assert issue, I was wondered if
PMSignalShmemInit() can hide an error when calling ShmemInitStruct() for backends?
If I understand correctly, backends get PMSignalState through backend_variables,
so maybe ShmemInitStruct() should not be executed for backends at all?
Best regards,
Alexander