RFC: gfc_internal_warning

[Fritz Reese]

On Mon, Nov 7, 2016 at 8:27 AM, Thomas Koenig <tkoenig@netcologne.de> wrote:
> Hello world,
>
> I am currently working on PR 78826, to make sure that everything
> in our AST has a location.
>
> Once that is the case, I plan to submit a version of the patch I use
> to detect the issue for trunk. Basically, it uses the expression walker
> to walk across the whole AST and complain if there is an issue.  This
> would catch the case of somebody putting in a new piece of code and
> forgetting to update the location information (which is easy to do,
> because there currently is no immediate feedback from the compiler).
>
> Question: What short of complaint should it make?  Code generation
> is not affected by lack of location information, so a warning would
> be more appropriate. The warning would also make sure there are
> testsuite failures for any new, wrong code, and users would not
> be affected as they would be with an ICE.
>
> I therefore propose to use a new function, gfc_internal_warning, for
> this kind of thing.
>
> comments?
>
>         Thomas

Sounds like a good idea to me.
---
Fritz Reese