Обсуждение: pgsql: Remove IS_AF_UNIX macro

Remove IS_AF_UNIX macro

The AF_UNIX macro was being used unprotected by HAVE_UNIX_SOCKETS,
apparently since 2008.  So the redirection through IS_AF_UNIX() is
apparently no longer necessary.  (More generally, all supported
platforms are now HAVE_UNIX_SOCKETS, but even if there were a new
platform in the future, it seems plausible that it would define the
AF_UNIX symbol even without kernel support.)  So remove the
IS_AF_UNIX() macro and make the code a bit more consistent.

Discussion: https://www.postgresql.org/message-id/flat/f2d26815-9832-e333-d52d-72fbc0ade896%40enterprisedb.com

Branch
------
master

Details
-------
https://git.postgresql.org/pg/commitdiff/797129e5910144a2a937b88e145874a15b83578a

Modified Files
--------------
src/backend/libpq/hba.c             |  4 ++--
src/backend/libpq/pqcomm.c          | 24 ++++++++++++------------
src/backend/postmaster/postmaster.c |  4 ++--
src/include/common/ip.h             |  6 ------
src/interfaces/libpq/fe-connect.c   | 14 +++++++-------
5 files changed, 23 insertions(+), 29 deletions(-)

Peter Eisentraut <peter@eisentraut.org> writes:
> Remove IS_AF_UNIX macro
> The AF_UNIX macro was being used unprotected by HAVE_UNIX_SOCKETS,
> apparently since 2008.

I hadn't looked closely at this patch, but are you referring to
this bit in ip.h?

#ifdef    HAVE_UNIX_SOCKETS
#define IS_AF_UNIX(fam) ((fam) == AF_UNIX)
#else
#define IS_AF_UNIX(fam) (0)
#endif

That's by no means "unprotected": we will not try to reference
AF_UNIX unless HAVE_UNIX_SOCKETS is set.  I think this change
will fail to break because we set HAVE_UNIX_SOCKETS everywhere,
but I believe it was a mistake.  We might as well just nuke
all the HAVE_UNIX_SOCKETS conditional compilation if we let
this stand.

(Now, maybe we should indeed do that.  I don't have much
interest in the possibility that we'll worry about such
platforms in future.)

            regards, tom lane

On 15.02.22 16:41, Tom Lane wrote:
> Peter Eisentraut <peter@eisentraut.org> writes:
>> Remove IS_AF_UNIX macro
>> The AF_UNIX macro was being used unprotected by HAVE_UNIX_SOCKETS,
>> apparently since 2008.
> 
> I hadn't looked closely at this patch, but are you referring to
> this bit in ip.h?
> 
> #ifdef    HAVE_UNIX_SOCKETS
> #define IS_AF_UNIX(fam) ((fam) == AF_UNIX)
> #else
> #define IS_AF_UNIX(fam) (0)
> #endif
> 
> That's by no means "unprotected": we will not try to reference
> AF_UNIX unless HAVE_UNIX_SOCKETS is set.

In src/backend/utils/adt/pgstatfuncs.c there is a use of AF_UNIX that 
has been there unprotected by any #ifdef since about 2008.

> We might as well just nuke
> all the HAVE_UNIX_SOCKETS conditional compilation if we let
> this stand.
> 
> (Now, maybe we should indeed do that.  I don't have much
> interest in the possibility that we'll worry about such
> platforms in future.)

Maybe/probably.  But there is a difference between platforms having the 
AF_UNIX symbol (which is required by POSIX unconditionally) and 
platforms actually having Unix sockets or not (which might require 
different default configurations or run-time behavior), so the two 
questions are not that closely connected.

Peter Eisentraut <peter.eisentraut@enterprisedb.com> writes:
> On 15.02.22 16:41, Tom Lane wrote:
>> I hadn't looked closely at this patch, but are you referring to
>> this bit in ip.h?

> In src/backend/utils/adt/pgstatfuncs.c there is a use of AF_UNIX that 
> has been there unprotected by any #ifdef since about 2008.

Oh, I see it :-(.  Poor coding, but it evidently no longer matters.

> Maybe/probably.  But there is a difference between platforms having the 
> AF_UNIX symbol (which is required by POSIX unconditionally) and 
> platforms actually having Unix sockets or not (which might require 
> different default configurations or run-time behavior), so the two 
> questions are not that closely connected.

Fair point.

            regards, tom lane