>> About second point, my bad, I meant p2, sorry for having written p1.

> This paragraph only discusses uses of placeholders that are used to
> declare the type of a parameter or template parameter.
> The text "The placeholder type shall appear as [...] or [...]" applies
only
> in those contexts.
> The conversion function case is covered by [dcl.spec.auto.general]/4.
> You can't have a trailing return type for a conversion function, so
> "type-specifier-seq" in this paragraph refers to the *type-specifier-seq*
> in *conversion-type-id* ([class.conv.fct]).
> Notice that [class.conv.fct]/2 says that "its declared return type is the
> *conversion-type-id*", which links it to [dcl.spec.auto.general]/4 where
> we also talk about "declared return type".

I will study this answer in next days, thank you.

> CWG2476 <https://cplusplus.github.io/CWG/issues/2476.html> was very
extensively reviewed in CWG.

Well, then I don't fill that bad in considering that this clause from C++20
should have be improved.


> I don't know the answer to this one.

I guess you are referring to
https://stackoverflow.com/questions/79208177/multiple-occurrences-of-placeholder-type-auto-in-a-type

I hope you may check in next days.
In case that plural does not make sense, then p3 might be reworded
getting rid of "the occurrences of" and then making it
similarly to p4 that currently reads "T shall be the placeholder alone",
except allowing additional decl-specifiers (e.g. const auto). For sure, you
guys
expert in the standard may find the best way to rephrase.

Hope to hear from you about this topic here in next days.
I will also try to rebuild the story on that wording, but I am not expert
about the best web locations to do it.


PS: no comment on that third point ? trailing return type ->
trailing-return-type

Thank you.

>