Re: [Patch, fortran] PR20880 - Failure to detect procedures encompassing own interface

[Paul Thomas <paulthomas2 at wanadoo dot fr>]

Brooks,
> > I believe that the test should include && sym->referenced.  I will
> > check this tonight and will commit under the obvious rule if it's OK.
The original PR no longer produces an error if I add the check on 
attr.referenced.  I am just about to scour the standard to see whether I 
should care or not.  It does regtest OK, as long as I adjust the new 
testcase.
> What does this do in the case where whether the symbol is referenced 
> is unknown at compile-time?  For instance, in Grigory's example module 
> the symbol isn't referenced, but it's exported and could be referenced 
> in some program that includes the module and is compiled later.
It carries the ambiguous attribute with it through the .mod file and the 
moment that it is used and referenced some place, the error will pop up.

Thanks for the contribution to the discussion.

Paul