Tom Lane wrote:
> typedef void (*PGeventProc) (PGeventId eventId, const void *eventInfo,
> void *passthrough);
> int PQregisterEventProc(PGconn *conn, PGeventProc proc, void *passthrough);
The above prototypes will work and we will add our 'event instance
pointer' to the event info structures. Should have a patch shortly.
libpqtypes doesn't need a passthrough/user-pointer. The object
events/hooks allocate memory when the object is created "part of a
conn/result object instance", it is not supplied by the API user
registering the event/hook callback.
I think this is where some confusion has been occurring, there are two
different pointers: user pointer and event instance pointer.
BTW, PQeventData and PQresultEventData return the event instance
pointer, not the passthrough. At least that is how we were using these
functions, being how our previous patches do not include a
passthrough/user-pointer feature because libpqtypes didn't need it.
every bit counts
In response to
pgsql-hackers by date
|Next:||From: Euler Taveira de Oliveira||Date: 2008-05-16 18:54:30|
|Subject: Re: ecpg localization|
|Previous:||From: Heikki Linnakangas||Date: 2008-05-16 15:56:41|
|Subject: Re: Arbitary file size limit in twophase.c|
pgsql-patches by date
|Next:||From: Oleg Bartunov||Date: 2008-05-16 18:37:27|
|Subject: Re: Partial match in GIN (next vesrion) |
|Previous:||From: Magnus Hagander||Date: 2008-05-16 18:31:03|
|Subject: Re: libpq thread-locking|