Re: GUC flags

On Sun, Nov 28, 2021 at 09:08:33PM -0600, Justin Pryzby wrote:
> This isn't flagged with GUC_EXPLAIN:
> enable_incremental_sort

Yeah, that's inconsistent.

> This isn't marked GUC_NOT_IN_SAMPLE, like all other DEVELOPER_OPTIONS:
> trace_recovery_messages

Indeed.

> I'm not sure jit_tuple_deforming should be marked GUC_NOT_IN_SAMPLE.
> I disable this one because it's slow for tables with many attributes.
> Same for jit_expressions ?

That would be consistent. Both are not in postgresql.conf.sample.

> bgwriter_lru_maxpages should have GUC_UNIT_BLOCKS
>
> max_identifier_length should have BYTES (like log_parameter_max_length and
> track_activity_query_size).
>
> block_size and wal_block_size should say BYTES (like wal_segment_size)
> And all three descriptions should say ".. in [prefix]bytes" (but see below).

Okay for these.

> Maybe these should have COMPAT_OPTIONS_PREVIOUS:
> integer_datetimes
> ssl_renegotiation_limit

Hmm. Okay as well for integer_datetimes.

> autovacuum_freeze_max_age has a comment about pg_resetwal which is obsolete
> since 74cf7d46a.
>
> checkpoint_warning refers to "checkpoint segments", which is obsolete since
> 88e982302.

That's part of 0002. That's a bit weird to use now, so I'd agree with
your suggestion to use "WAL segments" instead.

0001, to adjust the units, and 0003, to make the GUC descriptions less
unit-dependent, are good ideas.

- gettext_noop("Use of huge pages on Linux or Windows."),
+ gettext_noop("Enable use of huge pages on Linux or Windows."),
This should be "Enables use of".

{"compute_query_id", PGC_SUSET, STATS_MONITORING,
- gettext_noop("Compute query identifiers."),
+ gettext_noop("Enables in-core computation of a query identifier."),
This could just be "Computes"?

I am not completely sure that all the contents of 0002 are
improvements, but the suggestions done for huge_pages,
ssl_passphrase_command_supports_reload, checkpoint_warning and
commit_siblings seem fine to me.
--
Michael