Thread: A configure.in patch check (fwd)
Helps if I attach the patch... ---------- Forwarded message ---------- Date: Sun, 25 Aug 2002 14:36:19 +0100 (BST) From: Nigel J. Andrews <nandrews@investsystems.co.uk> To: pgsql-hackers@postgresql.org Subject: A configure.in patch check Would someone apply the attached patch to the development source and let me know if the autoconf step fails or works. I've only got autoconf 2.13 available and the file needs 2.53 apparently. If it works could I also have a copy of the resulting configure script, or patch, please. For the record, this is related to reserving the last few backend slots for the superuser and I just need to test what I've done. TIA -- Nigel J. Andrews Director --- Logictree Systems Limited Computer Consultants
"Nigel J. Andrews" <nandrews@investsystems.co.uk> writes:
> + AC_MSG_CHECKING([for default superuser reserved number of connections])
> + PGAC_ARG_REQ(with, reservedbackends, [ --with-reservedbackends=N set default superuser reserved number of
connections[2]],
> + [],
> + [with_reservedbackends=2])
This will be rejected anyway; what you want is to set up
reserved_backends as a GUC parameter, not as something that has to be
hard-wired at configure time. I can't see any reason to make it
hard-wired...
regards, tom lane
On Sun, 25 Aug 2002, Tom Lane wrote: > "Nigel J. Andrews" <nandrews@investsystems.co.uk> writes: > > + AC_MSG_CHECKING([for default superuser reserved number of connections]) > > + PGAC_ARG_REQ(with, reservedbackends, [ --with-reservedbackends=N set default superuser reserved number of connections[2]], > > + [], > > + [with_reservedbackends=2]) > > This will be rejected anyway; what you want is to set up > reserved_backends as a GUC parameter, not as something that has to be > hard-wired at configure time. I can't see any reason to make it > hard-wired... It is a GUC. It's exactly like max_backends. I took the easy route out and just followed where DEF_MAXBACKENDS was being set rather than hard wiring the value any where. Rather distressingly in order to get this new value into where it's needed I had to hit quite a few files, more than I would have expected. Again I just followed how MaxBackends was being sent to where it was needed but is there any particular reason why storage/ipc/sinvaladt.c:SIBackendInit() can't access MaxBackends and my new ReservedBackends directly? The are global variables afterall, I think #include "miscadmin.h" would need to be added but is that bad? -- Nigel J. Andrews Director --- Logictree Systems Limited Computer Consultants
"Nigel J. Andrews" <nandrews@investsystems.co.uk> writes:
> It is a GUC. It's exactly like max_backends. I took the easy route out and
> just followed where DEF_MAXBACKENDS was being set rather than hard wiring
> the value any where.
Oh. Well, skip the configure part: the only reason there's still a
configure parameter for maxbackends is backwards compatibility with
ancient configure scripts (from days when it was in fact frozen at
configure time). I don't see a need to provide one for reserved_slots.
> Rather distressingly in order to get this new value into where it's needed
> I had to hit quite a few files, more than I would have expected. Again I
> just followed how MaxBackends was being sent to where it was needed but is
> there any particular reason why storage/ipc/sinvaladt.c:SIBackendInit()
> can't access MaxBackends and my new ReservedBackends directly?
Probably not. Again, the way that MaxBackends is handled is largely
legacy code. You'd have been better off looking at almost any other
GUC parameter as a template ;-)
regards, tom lane