Re:

"Robert A. Knop Jr." <rknop(at)lilys(dot)lbl(dot)gov> writes:
>>>> No. Unless the doc for the function explicitly says you should free
>>>> its result, it's a pointer into libpq-managed space.
>>
>> Experience contradicts this.
>> I rewrote the thing explicitly freeing everything I got back from
>> PQfname() and PQgetvalue(). The memory leak was gone, and I didn't get
>> any segfaults or crashes.

I don't understand how. In recent versions, not only is the pointer
you get back not a freshly-malloced chunk, it's not a malloc chunk at
all, but a pointer into the middle of a much larger malloc chunk.
You must have an extremely forgiving malloc library, one wherein free()
doesn't crash if handed a pointer that doesn't point at a malloc chunk.

>> Is there a Postgres developer who can confirm this one way or the other?

I am a Postgres developer, and if you don't want to take my word for it,
read the source code. src/interfaces/libpq/fe-exec.c sez:

char *
PQgetvalue(const PGresult *res, int tup_num, int field_num)
{
if (!check_tuple_field_number("PQgetvalue", res, tup_num, field_num))
return NULL;
return res->tuples[tup_num][field_num].value;
}

char *
PQfname(const PGresult *res, int field_num)
{
if (!check_field_number("PQfname", res, field_num))
return NULL;
if (res->attDescs)
return res->attDescs[field_num].name;
else
return NULL;
}

Actual loading of the PGresult structure is elsewhere in the same file,
but you can certainly see that you're not getting your very own copy
back here.

regards, tom lane