Обсуждение: Mark ItemPointer arguments as const thoughoutly

Hi,

On Wed, 10 Sept 2025 at 05:27, Chao Li <li.evan.chao@gmail.com> wrote:
>
> Hi Hacker,
>
> This is a follow up 991295f. I searched over the src/ and make all ItemPointer arguments as const as much as
possible.
>
> I made clean build and no waring found. And "make check" also passes. I will create a patch on CF to see if CI
passes.

If you would like to run CI yourself, you can do so by forking the
Postgres GitHub repository and enabling Cirrus CI on your fork. Once
that is done, each commit will trigger a CI run. More details are
available in the README [1].

[1] https://git.postgresql.org/cgit/postgresql.git/tree/src/tools/ci/README

-- 
Regards,
Nazir Bilal Yavuz
Microsoft

On 26.09.25 04:54, Chao Li wrote:
> On Wed, Sep 10, 2025 at 1:20 PM Chao Li <li.evan.chao@gmail.com 
> <mailto:li.evan.chao@gmail.com>> wrote:
> 
>>     On Sep 10, 2025, at 12:20, Chao Li <li.evan.chao@gmail.com
>>     <mailto:li.evan.chao@gmail.com>> wrote:
>>
>>     v2 tries to fix the CI failure.
>>
>>     Chao Li (Evan)
>>     ---------------------
>>     HighGo Software Co., Ltd.
>>     https://www.highgo.com/ <https://www.highgo.com/>
>>
>>     <v2-0001-Mark-ItemPointer-arguments-as-const-thoughoutly.patch>
> 
>     Here is the CF patch https://commitfest.postgresql.org/patch/6046/
>     <https://commitfest.postgresql.org/patch/6046/>, and the CI tests
>     passed.

I have committed most of this patch.

I didn't like the few places in itemptr.h where you changed an existing 
ItemPointerData* back to ItemPointer, so I left those out.

More importantly, some of the proposed changes change the signatures of 
callback functions in the index or table AM APIs.  (And also the 
documentation wasn't updated.)  This would break source code 
compatibility with existing extensions that use those APIs.  There are 
have been previous proposals like this in [0], [1] where the changes to 
those APIs were not committed.  And your other patch 'Mark function 
arguments of type "Datum *" as "const Datum *" where possible' might 
have similar problems (although it looks like it's touching different 
places than [0] and [1]).  So I left those changes out of what I committed.

That doesn't mean we can never change these APIs.  We certainly change 
many backend APIs all the time, and extensions providing table or index 
AMs are sophisticated and will need to make adjustments anyway.  But it 
would be better if we did any changes in a deliberate way with explicit 
notice and advice for extensions (like commit 76acf4b722f for example).

[0]: 
https://www.postgresql.org/message-id/flat/14c31f4a-0347-0805-dce8-93a9072c05a5%40eisentraut.org
[1]: 
https://www.postgresql.org/message-id/flat/f8c739d9-f48d-4187-b214-df3391ba41ab%40eisentraut.org

Hi Peter,

> On Oct 30, 2025, at 21:30, Peter Eisentraut <peter@eisentraut.org> wrote:
>
> I have committed most of this patch.

Thank you so much for committing my patch and for your guidance. I’m still ramping up on PostgreSQL development, and
yourguidance has been super helpful to me. 

>
> I didn't like the few places in itemptr.h where you changed an existing ItemPointerData* back to ItemPointer, so I
leftthose out. 
>
> More importantly, some of the proposed changes change the signatures of callback functions in the index or table AM
APIs. (And also the documentation wasn't updated.)  This would break source code compatibility with existing extensions
thatuse those APIs. There are have been previous proposals like this in [0], [1] where the changes to those APIs were
notcommitted.  And your other patch 'Mark function arguments of type "Datum *" as "const Datum *" where possible' might
havesimilar problems (although it looks like it's touching different places than [0] and [1]).  So I left those changes
outof what I committed. 

I will revisit that patch once I am back from vacation.

Best regards,
--
Chao Li (Evan)
HighGo Software Co., Ltd.
https://www.highgo.com/