Re: [HACKERS] an older problem? hash table out of memory

Bruce Momjian <maillist(at)candle(dot)pha(dot)pa(dot)us> writes:
> Here they are. Can you properly fix them? Looks like good news that I
> found one of the ones you found. The others may be OK:

Is that all? Great, I was afraid we had some major problems lurking.

> ./backend/commands/view.c: snprintf(buf, strlen(viewName) + 5, "_RET%s", viewName);

This one is OK since viewName is passed in (and is valid, we hope).

> ./backend/executor/nodeHash.c: snprintf(tempname, strlen(tempname), "HJ%d.%d", (int) MyProcPid, hjtmpcnt);

This is the one I found. I'm still working on nodeHash but hope to
commit fixes in the next day or so.

> ./backend/libpq/pqcomm.c: snprintf(PQerrormsg + strlen(PQerrormsg), ERROR_MSG_LENGTH,

This is a bit bogus --- ERROR_MSG_LENGTH is the overall size of
PQerrormsg, but we're concatenating to what's already in the buffer, so
snprintf's limit should really be ERROR_MSG_LENGTH - strlen(PQerrormsg).
I applied a patch for consistency's sake, although I doubt this
statement could ever overrun the buffer in practice.

regards, tom lane