[Patch, fortran] PR108663 - Accepts invalid bug with pdtXXX

[Paul Richard Thomas]

Hi Jerry,

Thanks - pushed as r16-7520 with a rather heavily corrected ChangeLog.

Cheers

Paul

On Sat, 14 Feb 2026 at 19:44, Jerry D <jvdelisle2@gmail.com> wrote:
>
> On 2/14/26 9:28 AM, Paul Richard Thomas wrote:
> > Hello All,
> >
> > This bug allowed direct references by name to PDT instances. It took
> > me a long time to have the necessary light bulb moment: The names of
> > PDT instances never have to go through the scanner and so are not
> > subject to the ministrations of gfc_wide_tolower. Therefore, rendering
> > all the letters of the PDT prefix in upper case prevents pdt_types
> > from being directly referenced in code.
> >
> > I took the opportunity to define PDT_TEMPLATE and PDT_TEMPLATE_LEN in
> > gfortran.h.
> >
> > The chunk in resolve.cc is a funny one. I had previously encountered
> > the test for assumed LEN parameters in typebound procedure dummies
> > (pdt_37.f03) failing due to apparently unrelated changes in the
> > frontend. It seems that it is the order of appearance of pdt_templates
> > and pdt_types that mattered. The problem is fixed by testing for both,
> > as is necessary for this patch to pass regression testing.
> >
> > Regtests OK with FC43/x86_64. OK for mainline? I will be offline for a
> > few weeks from Monday evening and so I will push the patch then if it
> > hasn't been reviewed. In the words of Hitchhikers Guide to the Galaxy,
> > it is "mostly harmless".
> >
> > Paul
>
> OK for mainline Paul, thanks.
>
> Jerry
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Change.Logs
Type: application/octet-stream
Size: 854 bytes
Desc: not available
URL: <https://gcc.gnu.org/pipermail/fortran/attachments/20260215/87f3abc9/attachment.obj>