On 7/11/23 12:52 AM, Michael Paquier wrote:
> On Mon, Jul 10, 2023 at 09:11:36AM +0200, Alvaro Herrera wrote:
>> I don't like this bit, because it means the .txt file is now ungreppable
>> as source of the enum name. Things become mysterious and people have to
>> track down the event name by reading the the Perl generating script.
>> It's annoying. I'd rather have the extra column, even if it means a
>> little duplicity.
>
> Hmm. I can see your point that we'd lose the direct relationship
> between the enum and string when running a single `git grep` from the
> tree, still attempting to do that does not actually lead to much
> information gained? Personally, I usually grep for code when looking
> for consistent information across various paths in the tree. Wait
> events are very different: each enum is used in a single place in the
> tree making their grep search the equivalent of looking at
> wait_event_names.txt anyway?
>
Before commit fa88928470 one could find the relationship between the enum and the name
in wait_event.c (a simple git grep would provide it).
With commit fa88928470 in place, one could find the relationship between the enum and the name
in wait_event_names.txt (a simple git grep would provide it).
With the proposal we are discussing here, once the build is done and so the pgstat_wait_event.c
file is generated then we have the same "grep" capability than pre commit fa88928470 (except that
"git grep" can't be used and one would need things like
find . -name "*.c" -exec grep -il "WAIT_EVENT_CHECKPOINTER_MAIN" {} \;)
I agree that it is less "obvious" than pre fa88928470 but still doable though.
Regards,
--
Bertrand Drouvot
PostgreSQL Contributors Team
RDS Open Source Databases
Amazon Web Services: https://aws.amazon.com
