Re: Propagate XLogFindNextRecord error to callers

On Mon, 09 Feb 2026 at 10:21, Anthonin Bonnefoy <anthonin(dot)bonnefoy(at)datadoghq(dot)com> wrote:
> On Tue, Feb 3, 2026 at 5:05 PM Mircea Cadariu <cadariu(dot)mircea(at)gmail(dot)com> wrote:
>> I've picked up the review for your patch.
>
> Thanks for picking it!
>
>> Attached is a failing test that reproduces the issue. Have I got it
>> right? We can consider using it to validate your patch then.
>
> Yeah, that's the gist of it. However, the test you've written will
> only work on little endian architectures. Also, I think the xlog page
> header size won't have the 4 bytes padding on 32 bits systems.
>
> I've added a similar test in 001_basic.pl, but it relies on copying an
> existing WAL file and setting the WAL magic to 0000. This way, the
> result will be the same independent of the endianness and memory
> padding.
>

+ {
+ if (errormsg)
+ pg_fatal("could not find a valid record after %X/%X: %s",
+ LSN_FORMAT_ARGS(private.startptr), errormsg);
+ else
+ pg_fatal("could not find a valid record after %X/%X",
+ LSN_FORMAT_ARGS(private.startptr));

For consistency, this should use the %X/%08X format as elsewhere.

> Regards,
> Anthonin Bonnefoy
>
> [2. text/x-diff; v2-0001-Propage-errormsg-to-XLogFindNextRecord-caller.patch]...

--
Regards,
Japin Li
ChengDu WenWu Information Technology Co., Ltd.