Andreas Seltenreich writes:
> Thomas Munro writes:
>
>> It is safe, as long as the segment remains mapped. Each backend that
>> attaches calls LWLockRegisterTranche giving it the address of the name
>> in its virtual address space.
>
> Hmok, I was under the impression only backends participating in the IPC
> call the attach function, not necessarily the ones that could possible
> want to resolve the wait_event_info they found in the procArray via
> pgstat_get_wait_event().
Erm, ignore that question: They'll find a NULL in their
LWLockTrancheArray and run into the "extension" case you mentioned.
> But I really feel like I need to study the code a bit more before
> commenting further…
Following this advise now :-)
regards,
Andreas