[Patch, fortran] PR31867 - function result with character LEN computed at run time

[Paul Richard Thomas]

FX,

> I somehow lost track of this patch, sorry. This is OK for mainline,
> with added testcases from PR31994.

Thanks - will do on the testcases.
>
> Do you think this should go into 4.2? It doesn't appear to be a
> regression, but it sure is a nasty wrong-code bug. In the end, I guess
> it depends on how safe you think the patch is.

I think that we wait a week or two to check if it causes any
regressions and then commit it to 4.2.  It is right to mark this bug
as critical in PR31994.

Paul