Re: Support a wildcard in backtrace_functions

> On 28 Feb 2024, at 19:50, Jelte Fennema-Nio <me(at)jeltef(dot)nl> wrote:
>
> On Wed, 28 Feb 2024 at 19:04, Daniel Gustafsson <daniel(at)yesql(dot)se> wrote:
>> This should be "equal to or higher" right?
>
> Correct, nicely spotted. Fixed that. I also updated the docs for the
> new backtrace_functions_min_level GUC itself too, as well as creating
> a dedicated options array for the GUC. Because when updating its docs
> I realized that none of the existing level arrays matched what we
> wanted.

Looks good, I'm marking this ready for committer for now. I just have a few
small comments:

+ A single <literal>*</literal> character is interpreted as a wildcard and
+ will cause all errors in the log to contain backtraces.
This should mention that it's all error matching the level (or higher) of the
newly introduced GUC.

+ gettext_noop("Sets the message levels that create backtraces when backtrace_functions is configured"),
I think we should add the same "Each level includes.." long_desc, and the
short_desc should end with period.

+ <para>
+ Backtrace support is not available on all platforms, and the quality
+ of the backtraces depends on compilation options.
+ </para>
I don't think we need to duplicate this para here, having it on
backtrace_functions suffice.

--
Daniel Gustafsson