Re: Incorrect logic in XLogNeedsFlush()

> On Sep 18, 2025, at 07:20, Michael Paquier <michael(at)paquier(dot)xyz> wrote:
>
> On Tue, Sep 16, 2025 at 09:40:50AM +0900, Michael Paquier wrote:
>> As a whole, the patch looks like a good balance, able to satisfy the
>> new case you want to handle, Melanie. I am guessing that you'd want
>> to tweak it and apply it yourself, so please feel free.
>
> Hearing nothing, I'd like to move ahead with this improvement. I have
> tweaked a bit the comments, as suggested. If one switches the check
> of XLogNeedsFlush() from XLogInsertAllowed() to RecoveryInProgress(),
> the recovery test 015 blows up as expected.
>
> Any opinions or more word-smithing required?
> --
> Michael
> <v3-0001-Make-XLogFlush-and-XLogNeedsFlush-decision-more-c.patch>

My 5 cents:

```
+ * XLogInsertAllowed() is used as an end-of-recovery checkpoint is
+ * launched while recovery is still in progress, RecoveryInProgress()
```

“XLogInsertAllowed() is used as an end-of-recovery checkpoint is launched” sounds like XLogInsertAllowed() is being “used as a checkpoint,” how about rephrase as:

XLogInsertAllowed() is needed because an end-of-recovery checkpoint can be launched while recovery is still in progress,

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