Обсуждение: BUG #14984: function "heap_tuple_from_minimal_tuple" lostHeapTupleHeader->t_len value

The following bug has been logged on the website:

Bug reference:      14984
Logged by:          bucoo hsiao
Email address:      bucoo@sohu.com
PostgreSQL version: 10.1
Operating system:   all
Description:

HeapTuple
heap_tuple_from_minimal_tuple(MinimalTuple mtup)
{
    HeapTuple    result;
    uint32        len = mtup->t_len + MINIMAL_TUPLE_OFFSET;

    result = (HeapTuple) palloc(HEAPTUPLESIZE + len);
    result->t_len = len;
    ItemPointerSetInvalid(&(result->t_self));
    result->t_tableOid = InvalidOid;
    result->t_data = (HeapTupleHeader) ((char *) result + HEAPTUPLESIZE);
    memcpy((char *) result->t_data + MINIMAL_TUPLE_OFFSET, mtup,
mtup->t_len);
    memset(result->t_data, 0, offsetof(HeapTupleHeaderData, t_infomask2));
/* when memset return, result->t_data->t_len value lost
   should append code:
   HeapTupleHeaderSetDatumLength(result->t_data, len);
*/
    return result;
}

Hi,

On 2017-12-20 09:49:58 +0000, PG Bug reporting form wrote:
> The following bug has been logged on the website:
> 
> Bug reference:      14984
> Logged by:          bucoo hsiao
> Email address:      bucoo@sohu.com
> PostgreSQL version: 10.1
> Operating system:   all
> Description:        

This isn't really a complete bug report. How did you arrive at this
being a problem? What triggers it?


> HeapTuple
> heap_tuple_from_minimal_tuple(MinimalTuple mtup)
> {
>     HeapTuple    result;
>     uint32        len = mtup->t_len + MINIMAL_TUPLE_OFFSET;
> 
>     result = (HeapTuple) palloc(HEAPTUPLESIZE + len);
>     result->t_len = len;
>     ItemPointerSetInvalid(&(result->t_self));
>     result->t_tableOid = InvalidOid;
>     result->t_data = (HeapTupleHeader) ((char *) result + HEAPTUPLESIZE);
>     memcpy((char *) result->t_data + MINIMAL_TUPLE_OFFSET, mtup,
> mtup->t_len);
>     memset(result->t_data, 0, offsetof(HeapTupleHeaderData, t_infomask2));
> /* when memset return, result->t_data->t_len value lost
>    should append code:
>    HeapTupleHeaderSetDatumLength(result->t_data, len);
> */

There is no result->t_data->t_len. It's result->t_len. And calling
HeapTupleHeaderSetDatumLength() wouldn't be right, because the returned
value is an actual tuple, not a datum. Use heap_copy_tuple_as_datum() if
you want that.

Greetings,

Andres Freund

=?utf-8?q?PG_Bug_reporting_form?= <noreply@postgresql.org> writes:
>     memset(result->t_data, 0, offsetof(HeapTupleHeaderData, t_infomask2));
> /* when memset return, result->t_data->t_len value lost
>    should append code:
>    HeapTupleHeaderSetDatumLength(result->t_data, len);
> */

As with #14985, you haven't shown any reason why this is a good change
to make.

            regards, tom lane