The same is true for -Werror.
Warnings still give an exit status code of zero.

> The same is true for -Werror.

I have to correct myself:
-Werror gives a non-zero exit status, but still writes "Warning:". I think gfortran should follow gcc by changing also the label from "Warning:" to "Error:" for -Werror.

-pedantic-errors seems to be difficult as it is hard to divide -Wall warnings from -pedantic warnings. The easiest way is probably to imply -Wall by -pedantic_errors (defined in flags.h as flag_pedantic_errors).

I am not sure if gfortran diagnostics are different, but... are you sure that particular warning is a pedantic warning and not simply an unconditional warning? AFAIK, pedantic-errors will turn only pedantic warnings into errors, other warnings will still be warnings unless -Werror.

> I am not sure if gfortran diagnostics are different,

I think gfortran handles the warnings quite different, not that I know much about the details of the C frontend.

> but... are you sure that particular warning is a pedantic warning and not 
> simply an unconditional warning?

primary.c:  if (x_hex && pedantic
primary.c-      && (gfc_notify_std (GFC_STD_GNU, "Extension: Hexadecimal "

This gives only an error with pedantic set (and -std=f95 or -std=f2003).

Otherwise, -pedantic is quite interwoven with the rest: options.c, e.g.:

  if (pedantic)
    {
      gfc_option.warn_ampersand = 1;
      gfc_option.warn_tabs = 0;
    }

No idea how to untangle -pedantic from -Wtabs or -Wampersand if -pedantic-errors has been given, but -Werror has not.

Silently accepting and ignoring it, seems not to be the right way. The simple solution is not to accept -pedantic-errors (and to point to -Werror) or to turn on -Werror for -pedantic-errors. Both are rather easy solutions - and feel a bit clumsy. It needs presumably quite a lot of work to support -pedantic-errors properly.

First and simpler step should be to change "Warning:" into "Error:" for -Werror to be in line with the C front end.

(In reply to comment #4)
> > I am not sure if gfortran diagnostics are different,
> 
> I think gfortran handles the warnings quite different, not that I know much
> about the details of the C frontend.
> 

Ah, OK. Then forget anything that I said.

(In reply to comment #4)
> 
> No idea how to untangle -pedantic from -Wtabs or -Wampersand if
> -pedantic-errors has been given, but -Werror has not.
>

What gfortran should do is that if pedantic enables Wtabs, then the warnings should be of the form:

if (pedantic && warn_tabs)
 pedantic("whatever");
else if (warn_tabs)
 warning("whatever");

pedantic() emits errors if -pedantic-errors, otherwise it emits warnings.
warning() emits errors if -Werror, otherwise it emits warnings.

I guess there would be similar functions in gfortran. (It would be great to integrate the diagnostics machinery but making things work in a similar way is already a step forward).

An update given the new diagnostics machinery:

* -Werror works as expected (comment #2)

* -pedantic-errors (-Werror=pedantic) should give errors for anything marked with -Wpedantic such as:

gfc_warning (OPT_Wpedantic, "Extension: backslash character at %C");

However, Fortran does not seem to use OPT_Wpedantic. You could use LangEnabledBy to enable with it whatever option you wish.

The following patch produces the desired error for the example
of comment #0:

--- gcc/fortran/primary.c       (revision 232904)
+++ gcc/fortran/primary.c       (working copy)
@@ -1698,7 +1698,9 @@
        }
     }
 
-  if (!gfc_notify_std (GFC_STD_GNU, "argument list function at %C"))
+  if (!gfc_notify_std (GFC_STD_GNU, "argument list function at %C") ||
+      (pedantic &&
+       !gfc_warning (OPT_Wpedantic, "Extension: argument list function at %C")))
     {
       m = MATCH_ERROR;
       goto cleanup;


% gfc-trunk -c -Werror=pedantic gcc/testsuite/gfortran.dg/c_by_val_4.f
gcc/testsuite/gfortran.dg/c_by_val_4.f:14:22:

       CALL DOIT( %VAL( P ) ) ! { dg-warning "Extension: argument list function" }
                      1
Error: Extension: argument list function at (1) [-Werror=pedantic]
gcc/testsuite/gfortran.dg/c_by_val_4.f:16:22:

       CALL DOIT( %VAL( P ) ) ! { dg-warning "Extension: argument list function" }
                      1
Error: Extension: argument list function at (1) [-Werror=pedantic]


A more general solution might be to extend gfc_notify_std to generally
handle extensions.

Since at least 4.8 '-pedantic -Werror' gives an error;

% gfortran /opt/gcc/work/gcc/testsuite/gfortran.dg/c_by_val_4.f -c -pedantic -Werror
/opt/gcc/work/gcc/testsuite/gfortran.dg/c_by_val_4.f:14:6:

       CALL DOIT( %VAL( P ) ) ! { dg-warning "Extension: argument list function" }
      1
Error: Unclassifiable statement at (1)
/opt/gcc/work/gcc/testsuite/gfortran.dg/c_by_val_4.f:16:6:

       CALL DOIT( %VAL( P ) ) ! { dg-warning "Extension: argument list function" }
      1
Error: Unclassifiable statement at (1)

but '-pedantic-errors' does not

% gfortran /opt/gcc/work/gcc/testsuite/gfortran.dg/c_by_val_4.f -c -pedantic-errors
/opt/gcc/work/gcc/testsuite/gfortran.dg/c_by_val_4.f:14:22:

       CALL DOIT( %VAL( P ) ) ! { dg-warning "Extension: argument list function" }
                      1
Warning: GNU Extension: argument list function at (1)
/opt/gcc/work/gcc/testsuite/gfortran.dg/c_by_val_4.f:16:22:

       CALL DOIT( %VAL( P ) ) ! { dg-warning "Extension: argument list function" }
                      1
Warning: GNU Extension: argument list function at (1)

AFAICT '-pedantic-errors' is not handled by the gfortran FE.

Note that '-pedantic' may change a warning to an error:

% gfortran /opt/gcc/work/gcc/testsuite/gfortran.dg/hollerith8.f90 -pedantic
/opt/gcc/work/gcc/testsuite/gfortran.dg/hollerith8.f90:6:17:

   call wrtout (9hHELLO YOU, 9)
                 1
Warning: Legacy Extension: Hollerith constant at (1)
/opt/gcc/work/gcc/testsuite/gfortran.dg/hollerith8.f90:6:15:

   call wrtout (9hHELLO YOU, 9)
               1
Error: Rank mismatch in argument 'iarray' at (1) (rank-1 and scalar)