On 16.11.2018 11:23, Michael Paquier wrote:
> On Thu, Nov 15, 2018 at 06:12:38PM +0900, Kyotaro HORIGUCHI wrote:
>> This patch eliminates unnecessary copying that was done for
>> non-continued records. Now the return value of XLogReadRecord
>> directly points into page buffer holded in XLogReaderStats. It is
>> safe because no caller site uses the returned pointer beyond the
>> replacement of buffer content at the next call to the same
>> function.
>
> I was looking at this patch, and shouldn't we worry about compatibility
> with plugins or utilities which look directly at what's in readRecordBuf
> for the record contents? Let's not forget that the contents of
> XLogReaderState are public.
According to my experience, I clarify some comments to avoid this
mistakes in the future (see attachment).
--
Andrey Lepikhov
Postgres Professional
https://postgrespro.com
The Russian Postgres Company