On Sat, Jul 28, 2018 at 2:00 AM, Robert Haas <robertmhaas@gmail.com> wrote:
> On Wed, Jul 11, 2018 at 8:58 PM, Masahiko Sawada <sawada.mshk@gmail.com> wrote:
>> While reading the replication slot codes, I found a wrong assignment
>> in pg_logical_slot_get_changes_guts() function as follows.
>>
>> if (PG_ARGISNULL(2))
>> upto_nchanges = InvalidXLogRecPtr;
>> else
>> upto_nchanges = PG_GETARG_INT32(2);
>>
>> Since the upto_nchanges is an integer value we should set 0 meaning
>> unlimited instead of InvalidXLogRecPtr. Since InvalidXLogRecPtr is
>> actually 0 this function works fine so far.
>
> If somebody changes InvalidXLogRecPtr to (uint64)-1, then it breaks as
> the code is written. On the other hand, if somebody reverted
> 0ab9d1c4b31622e9176472b4276f3e9831e3d6ba, it would keep working as
> written but break under your proposal.
I might be missing something but I think the setting either 0 or
negative values to it solves this problem. Since the upto_nchanges is
int32 we cannot build if somebody reverted
0ab9d1c4b31622e9176472b4276f3e9831e3d6ba.
Regards,
--
Masahiko Sawada
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center