On Tue, Jan 7, 2014 at 12:46 AM, Stephen Woodbridge
<woodbri(at)swoodbridge(dot)com> wrote:
> On 1/6/2014 10:00 AM, Pfuntner, John wrote:
>>
>> If I've done a palloc() to get storage inside a user-defined function and
>> raise an error using ereport(), should I be using pfree() to release the
>> storage before the ereport()?
>>
>> Consider this example in C:
>>
>> PG_FUNCTION_INFO_V1(Example);
>> Datum
>> Example(PG_FUNCTION_ARGS) {
>> VarChar* pstring=PG_GETARG_VARCHAR_P(0);
>> VarChar* ret=NULL;
>>
>> int len = VARSIZE(pstring) - VARHDRSZ;
>> char *string=palloc(len+1);
>> memcpy(string, VARDATA(pstring), len);
>> string[len] = '\0';
>>
>> /* ... */
>>
>> if ( /* some bad condition */ ) {
>> ereport(ERROR, (errcode(ERRCODE_DATA_EXCEPTION), errmsg("some bad
>> condition occurred!")));
>> }
>>
>> /* ... */
>>
>> pfree(string);
>>
>> if (ret == NULL)
>> PG_RETURN_NULL();
>> else
>> PG_RETURN_VARCHAR_P(ret);
>> }
>>
>> I only have the pfree() at the end before the return if there is no error.
>> If I fail to call pfree() before ereport(), do I have a memory leak?
>
>
> No, this is why it is important to use palloc and malloc.
palloc allocates memory in the context of a transaction, meaning that
it will be automatically freed when transaction finishes by either a
commit or an abort. When calling ereport the transaction will be
automatically aborted, hence free'ing the memory. Have a look here
for more details:
http://www.postgresql.org/docs/devel/static/xfunc-c.html#AEN54284
Regards,
--
Michael