Re: warning in code while building on windows

On Mon, Aug 19, 2013 at 9:33 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> Andrew Dunstan <andrew(at)dunslane(dot)net> writes:
>>> Amit Kapila escribió:
>>>> 1>.\src\backend\utils\cache\relfilenodemap.c(213) : warning C4101:
>>>> 'isnull' : unreferenced local variable
>
>> The macro is pretty gcc-specific, isn't it?
>
> If that's the problem, why isn't Amit seeing a boatload of similar
> warnings elsewhere?

If I try to change other place similar to relfilenodemap.c, I am
getting similar warning. For example:

Original Code
--------------------
ExecMaterial()
{
..
/*
* Allocate a second read pointer to serve as the mark. We know it
* must have index 1, so needn't store that.
*/
int ptrno PG_USED_FOR_ASSERTS_ONLY;

ptrno = tuplestore_alloc_read_pointer(tuplestorestate,
node->eflags);
Assert(ptrno == 1);
..
}

Modified Code
---------------------
ExecMaterial()
{
..
/*
* Allocate a second read pointer to serve as the mark. We know it
* must have index 1, so needn't store that.
*/
int ptrno PG_USED_FOR_ASSERTS_ONLY;

#ifdef USE_ASSERT_CHECKING

ptrno = tuplestore_alloc_read_pointer(tuplestorestate,
node->eflags);
Assert(ptrno == 1);
#endif
..
}

After above modification it gives below compilation error:
1>.\src\backend\executor\nodeMaterial.c(69) : warning C4101: 'ptrno' :
unreferenced local variable

Coming to original warning, changing the code as below removes warning:

RelidByRelfilenode()
{
..
#ifdef USE_ASSERT_CHECKING
if (assert_enabled)
{
Oid check;
bool isnull PG_USED_FOR_ASSERTS_ONLY;
check = fastgetattr(ntp, Anum_pg_class_reltablespace,
RelationGetDescr(relation),
&isnull);
Assert(!isnull && check == reltablespace);

check = fastgetattr(ntp, Anum_pg_class_relfilenode,
RelationGetDescr(relation),
&isnull);
Assert(!isnull && check == relfilenode);
}
#endif

Moving isnull declaration inside if block resolves current compilation
warning, I think in any case it is better to declare inside if block
as it is used in that block only.

I think resolving the bigger problem such that it should not give
warning for such usage in MSVC is important, but can be dealt as a
separate thread/patch.

With Regards,
Amit Kapila.
EnterpriseDB: http://www.enterprisedb.com