Re: Persist injection points across server restarts

Hi,

Thank you for addressing the review comments.

> It appears that Github CI is reporting failures with
> > injection_points/002_data_persist
> > failing across all OSes.
>
> I am not sure to see what you are referring to here, based on the
> following reports:
> https://commitfest.postgresql.org/patch/5731/
> https://cirrus-ci.com/github/postgresql-cfbot/postgresql/cf%2F5731
> http://cfbot.cputube.org/highlights/all.html
>
> The last failure reported that I know of was due to the addition of
> the runtime arguments to the macro INJECTION_POINT().
>
>
After applying the v3-patches, I see failure like these:
macOS - Sonoma - Meson - Cirrus CI
<https://cirrus-ci.com/task/4589192849653760>
Windows - Server 2019, VS 2019 - Meson & ninja - Cirrus CI
<https://cirrus-ci.com/task/5715092756496384>
Linux - Debian Bookworm - Meson - Cirrus CI
<https://cirrus-ci.com/task/6629886430806016>



> > I wonder if it would be better to include a check for the return value of
> > `durable_rename` to
> > manage potential failure scenarios.
>
> Or we can be smarter and make this call an ERROR.
>

One issue with this approach is that if a failure occurs during rename, we
cannot
unlink(INJ_DUMP_FILE_TMP), because durable_rename does not handle this
before throwing an error.

Should we also consider unlinking the target file INJ_DUMP_FILE in case of
an
error during renaming?
Since target files are already synced at the start in the durable_rename()
function.

Thank you,
Rahila Syed