Re: Improve docs syntax checking and enable it in the meson build

Hi,

On Mon, 16 Feb 2026 at 03:47, Hayato Kuroda (Fujitsu)
<kuroda(dot)hayato(at)fujitsu(dot)com> wrote:
>
> Dear Nazir,
>
> I created a similar patch with yours because I did not recognize this thread.
> Not sure you are still interested in, but here are my comments based on my
> experience.

Thank you for looking into this!

> 00.
> Needs rebase.

Done.

> 01.
> I found that postgres core was built when I ran `meson test --suite doc`
> command, but ideally it's not needed. Per my experiment, we must clarify
> dependencies for the test, something like below.
>
> ```
> --- a/doc/src/sgml/meson.build
> +++ b/doc/src/sgml/meson.build
> @@ -337,7 +337,8 @@ test(
> 'doc' / 'sgml_syntax_check',
> meson_bin,
> args: ['compile', postgres_full_xml_target],
> - suite: 'doc'
> + suite: 'doc',
> + depends: doc_generated,
> )
> ```

Sorry, I didn't understand what you meant, could you please elaborate?
'postgres_full_xml_target' target already depends on 'doc_generated'
and 'sgml_check_stamp' targets.

> 02.
> Missing update of copyright.

Done.

> 03.
> I found that for the autoconf build system, syntax checking cannot work again
> if nothing is changed. This behavior is different from the HEAD. Per my
> experiment, below can fix the issue.
>
> ```
> --- a/doc/src/sgml/Makefile
> +++ b/doc/src/sgml/Makefile
> @@ -201,6 +201,7 @@ MAKEINFO = makeinfo
> ##
>
> # Quick syntax check without style processing
> +.PHONY: postgres-full.xml
> check: postgres-full.xml
> ```

Yes, there is a behavior change but I don't think this is an issue. If
nothing is changed, syntax check just reports it as:

```
$ make check
make: Nothing to be done for 'check'.
```

> 04.
> `make check` is slightly slower than HEAD. I think it is OK but others can also
> confirm and can test on their env.
>
> HEAD:
> ```
> $ time make check
> ...
>
> real 0m0.592s
> user 0m0.507s
> sys 0m0.084s
> ```
>
> patched:
> ```
> $ time make check
> ...
>
> real 0m1.217s
> user 0m0.973s
> sys 0m0.237s
> ```

It is slower on my end too:

HEAD:
real 0m0.344s
user 0m0.301s
sys 0m0.041s

Patched:
real 0m0.451s
user 0m0.373s
sys 0m0.076s

The difference is not that big on my end, though. I think this is
expected and acceptable but I would like to hear other people's
thoughts too.

--
Regards,
Nazir Bilal Yavuz
Microsoft