Re: [Patch, fortran PR89645/99065 No IMPLICIT type error with: ASSOCIATE( X => function() )

[Paul Richard Thomas <paul.richard.thomas@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 2413 bytes --]

Hi Harald,

Roger that about the comments. The major part of my recent efforts has been
to maximise comments - apparently not always successfully!

The main reason that I want to "fix everything" is that this is it; I will
not work on this approach anymore. The gfortran/g95 founder's approach was
very clever but has found it's limit with the associate construct. The sad
thing is that this is the only blocker that I know of.

Thanks

Paul


On Tue, 12 Mar 2024 at 21:07, Harald Anlauf <anlauf@gmx.de> wrote:

> Hi Paul,
>
> On 3/12/24 15:54, Paul Richard Thomas wrote:
> > Hi All,
> >
> > This is the last posting of this patch before I push it. Harald is OK
> with
> > it on the grounds that the inferred_type flag guards the whole lot,
> > except for the chunks in trans-stmt.cc.
> >
> > In spite of Harald's off-list admonition not to try to fix everything at
> > once, this version fixes most of the inquiry reference bugs
> > (associate_68.f90) with the exception of character(kind=4) function
> > selectors. The reason for this is that I have some housekeeping to do
> > before release on finalization and then I want to replace this patch in
> > 15-branch with two pass parsing. My first attempts at the latter were a
> > partial success.
>
> you wouldn't stop trying to fix everything, would you?  ;-)
>
> > It regtests OK on x86_64. Unless there are objections, I will commit on
> > Thursday evening.
>
> No objections, just one wish: could you improve the text of the
> following comments so that mere mortals understand them?
>
> diff --git a/gcc/fortran/primary.cc b/gcc/fortran/primary.cc
> index 12e7bf3c873..0ab69bb9dce 100644
> --- a/gcc/fortran/primary.cc
> +++ b/gcc/fortran/primary.cc
> [...]
> +      /* If there is a usable inquiry reference not there are no matching
> +        derived types, force the inquiry reference by setting unknown the
> +        type of the primary expression.  */
>
>
> I have a hard time parsing the first part of that sentence.
>
> diff --git a/gcc/fortran/symbol.cc b/gcc/fortran/symbol.cc
> index 5d9852c79e0..16adb2a7efb 100644
> --- a/gcc/fortran/symbol.cc
> +++ b/gcc/fortran/symbol.cc
> [...]
> +/* Find all derived types in the uppermost namespace that have a component
> +   a component called name and stash them in the assoc field of an
> +   associate name variable.
>
>
> "a component" too much?
>
> Thanks,
> Harald
>
> > Cheers
> >
> > Paul
>
>