Обсуждение: Possible mistake in Section 63.6 - 9.6devel Documentation

On Tue, Jul 21, 2015 at 6:54 AM, Vignesh Raghunathan
<vignesh.pgsql@gmail.com> wrote:
> Hello,
>
> It has been mentioned in Section 63.6 that the first two fields in
> PageHeaderData track the most recent WAL entry related to the page. However,
> I am not sure how pd_checksum is related to WAL. Could it be possible that
> the sentence has been carried over from previous versions of the
> documentations without considering the change to the second field in
> PageHeaderData?

Yes, the documentation is mistaken. The two bytes of pd_tli have been
switched to pd_checksum in 9.3, hence only the first field is relevant
for WAL, aka pd_lsn. Looking at this portion of the docs I think that
it should be updated as attached, mentioning pd_checksum as well.
--
Michael

On Tue, Jul 21, 2015 at 9:59 AM, Michael Paquier
<michael.paquier@gmail.com> wrote:
> On Tue, Jul 21, 2015 at 6:54 AM, Vignesh Raghunathan
> <vignesh.pgsql@gmail.com> wrote:
>> Hello,
>>
>> It has been mentioned in Section 63.6 that the first two fields in
>> PageHeaderData track the most recent WAL entry related to the page. However,
>> I am not sure how pd_checksum is related to WAL. Could it be possible that
>> the sentence has been carried over from previous versions of the
>> documentations without considering the change to the second field in
>> PageHeaderData?
>
> Yes, the documentation is mistaken. The two bytes of pd_tli have been
> switched to pd_checksum in 9.3, hence only the first field is relevant
> for WAL, aka pd_lsn. Looking at this portion of the docs I think that
> it should be updated as attached, mentioning pd_checksum as well.

Also the type of pd_lsn in the Table 63-3 should be PageXLogRecPtr.

Regards,

--
Fujii Masao

On Thu, Sep 10, 2015 at 12:42 AM, Fujii Masao <masao.fujii@gmail.com> wrote:
> On Tue, Jul 21, 2015 at 9:59 AM, Michael Paquier
> <michael.paquier@gmail.com> wrote:
>> On Tue, Jul 21, 2015 at 6:54 AM, Vignesh Raghunathan
>> <vignesh.pgsql@gmail.com> wrote:
>>> Hello,
>>>
>>> It has been mentioned in Section 63.6 that the first two fields in
>>> PageHeaderData track the most recent WAL entry related to the page. However,
>>> I am not sure how pd_checksum is related to WAL. Could it be possible that
>>> the sentence has been carried over from previous versions of the
>>> documentations without considering the change to the second field in
>>> PageHeaderData?
>>
>> Yes, the documentation is mistaken. The two bytes of pd_tli have been
>> switched to pd_checksum in 9.3, hence only the first field is relevant
>> for WAL, aka pd_lsn. Looking at this portion of the docs I think that
>> it should be updated as attached, mentioning pd_checksum as well.
>
> Also the type of pd_lsn in the Table 63-3 should be PageXLogRecPtr.

Yep. See attached FWIW.
--
Michael

On Thu, Sep 10, 2015 at 3:05 PM, Michael Paquier
<michael.paquier@gmail.com> wrote:
> On Thu, Sep 10, 2015 at 12:42 AM, Fujii Masao <masao.fujii@gmail.com> wrote:
>> On Tue, Jul 21, 2015 at 9:59 AM, Michael Paquier
>> <michael.paquier@gmail.com> wrote:
>>> On Tue, Jul 21, 2015 at 6:54 AM, Vignesh Raghunathan
>>> <vignesh.pgsql@gmail.com> wrote:
>>>> Hello,
>>>>
>>>> It has been mentioned in Section 63.6 that the first two fields in
>>>> PageHeaderData track the most recent WAL entry related to the page. However,
>>>> I am not sure how pd_checksum is related to WAL. Could it be possible that
>>>> the sentence has been carried over from previous versions of the
>>>> documentations without considering the change to the second field in
>>>> PageHeaderData?
>>>
>>> Yes, the documentation is mistaken. The two bytes of pd_tli have been
>>> switched to pd_checksum in 9.3, hence only the first field is relevant
>>> for WAL, aka pd_lsn. Looking at this portion of the docs I think that
>>> it should be updated as attached, mentioning pd_checksum as well.
>>
>> Also the type of pd_lsn in the Table 63-3 should be PageXLogRecPtr.
>
> Yep. See attached FWIW.

Thanks! Applied.

Regards,

--
Fujii Masao