Terminology in Fortran error messages and options

[Richard Biener]

On Sat, Jun 6, 2026 at 10:55 AM Thomas Koenig <tkoenig@netcologne.de> wrote:
>
> Hello world,
>
> gfortran error messages are sometimes at odds with Fortran
> standard terminology.  For example, -Wuninitialized warns
> about automatic-storage variables which do not have a value
> assigned, and the error messages also are "used uninitialized".
>
> In Fortran standard terminology, on the other hand, a variable
> becomes "defined" or "undefined", so a Fortran compiler should
> strive to use that terminology.
>
> How to deal with this?  I would propose to
>
> a) only use Fortran standard terminology for new options
>     and error messages
>
> b) Eventually go through Fortran error messages and
>     change them if possible
>
> c) Check if it is possible to modify middle-end error messages
>     to switch to Fortran terminology
>
> d) Think about what to do for options which use the
>     wrong (for Fortran) terminology
>
> Comments? Should I open a PR?

I think the status quo is least confusing with respect to

 a) diagnostic option naming and its documentation

 b) when you consider multi-language projects where you'd
   get a mix of 'undefined' and 'uninitialized' diagnostics

>
> Best regards
>
>         Thomas
>