The mechanism that occurs to me (and I'm not wedded to
this idea) is:
typedef uint8 T_HOFF_TYPE;
typedef struct xl_heap_header
{
uint16 t_infomask2;
uint16 t_infomask;
T_HOFF_TYPE t_hoff;
} xl_heap_header;
#define SizeOfHeapHeader (offsetof(xl_heap_header, t_hoff) +
sizeof(T_HOFF_TYPE))
On Thursday, January 2, 2014 3:39 PM, Tom Lane <t...@sss.pgh.pa.us> wrote:
Mark Dilger <markdil...@yahoo.com> writes:
> I still don't understand why this case in src/include/pgstat.h
> is different from cases elsewhere in the code.
The reason why I'm exercised about it is that (a) somebody
actually made a
mistake of this type, and (b) it wasn't caught by any automated testing.
The catalog and WAL-related examples you cite would probably crash
and burn in fairly obvious ways if somebody broke them --- for instance,
the most likely way to break SizeOfHeapHeader would be by adding another
field after t_hoff, but we'd notice that before long because said field
would be corrupted on arrival at a replication slave.
In contrast, messing up the pgstats message sizes would have no
consequences worse than a hard-to-detect, and probably platform-specific,
performance penalty for stats transmission. So unless we think that's
of absolutely zero concern, adding a mechanism to make such bugs more
apparent seems useful.
I'm not against adding more assertions elsewhere, but it's a bit hard to
see what those asserts should test. I don't see any practical way to
assert that field X is the last one in its struct, for instance.
regards, tom lane
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
