On 2 December 2013 19:37, Alvaro Herrera <alvherre@2ndquadrant.com> wrote:
> Dean Rasheed escribió:
>
>> +/*
>> + * If a schema was explicitly specified, test if it exists. If it does not,
>> + * report the schema as missing rather than the child object.
>> + */
>> +static bool
>> +schema_does_not_exist_skipping(List *objname,
>> + const char **msg,
>> + char **name)
>> +{
>> + RangeVar *rel;
>> +
>> + rel = makeRangeVarFromNameList(objname);
>> +
>> + if (rel->schemaname != NULL &&
>> + !OidIsValid(LookupNamespaceNoError(rel->schemaname)))
>> + {
>> + *msg = gettext_noop("schema \"%s\" does not exist, skipping");
>> + *name = rel->schemaname;
>> +
>> + return true;
>> + }
>> +
>> + return false;
>> +}
>
> In success cases, are we leaking a lot of memory? In the error case I
> guess it doesn't matter that the RangeVar is getting leaked (we're
> aborting anyway), but if we're called and everything turns out to work,
> are things cleaned up timely?
>
I think that memory gets freed at the end of the DROP command, so I
don't think this is a concern. In any case, that RangeVar is only of
order 50 bytes. If we were concerned about memory leakage here, a
bigger concern would be the calling code in does_not_exist_skipping(),
which is using NameListToString() which allocates at least 1024 bytes
for the name of the non-existent object without freeing it.
Regards,
Dean