| From: | Kirill Reshke <reshkekirill(at)gmail(dot)com> |
|---|---|
| To: | pgsql-hackers <pgsql-hackers(at)postgresql(dot)org> |
| Subject: | Do not emit WAL record for FSM truncaton when unneeded. |
| Date: | 2025-11-22 06:12:05 |
| Message-ID: | CALdSSPhhWpufHQAFT8bZXR_qQNmYNDEVUfJtKYXSNSUDOyeCgw@mail.gmail.com |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Hi hackers!
While working on other matters, I noticed that the
`fsm_truncate_avail` function returns a bool parameter which is never
used. This makes me think we can make an enhancement here. There are
basically two options:
1) make return type `void`
2) make use of return value.
I tried the second option. In FreeSpaceMapPrepareTruncateRel (the only
user of fsm_truncate_avail), we call fsm_truncate_avail and then
WAL-log FSM page changes.
Instead, can use `fsm_truncate_avail ` return value to check if
WAL-logging can be skipped.
PFA with these changes.
Thoughts?
P.S. I did not make a POC of real-life scenarios when the FSM page is
unaltered by fsm_truncate_avail (yet). I wonder if there are such
scenarios. If there is none, option 1 should proceed IMHO.
--
Best regards,
Kirill Reshke
| Attachment | Content-Type | Size |
|---|---|---|
| v1-0001-Do-not-emit-WAL-record-for-FSM-truncaton-when-unn.patch | application/octet-stream | 2.2 KB |
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Nazir Bilal Yavuz | 2025-11-22 08:09:30 | Re: meson and check-tests |
| Previous Message | Nico Williams | 2025-11-22 05:24:11 | Re: [oauth] SASL mechanisms |