Next step, full error message: 
ibtool: compile:  gfortran -I../basics -I../utilities -I../testing -I../system -I../combinatorics -I../rng -I../physics -I../fastjet -I../qft -I../types -I../particles -I../../circe1/src -I../../circe2/src -I../pdf_builtin -I../lhapdf -I../qed_pdf -I../variables -I../expr_base -I../parsing -g -O2 -c sf_lhapdf.f90  -fPIC -o .libs/sf_lhapdf.o

in gfc_format_decoder, at fortran/error.c:970
0x5f46ce gfc_format_decoder
	../../gcc/fortran/error.c:970
0x18b722c pp_format(pretty_printer*, text_info*)
	../../gcc/pretty-print.c:1475
0x189c41e diagnostic_report_diagnostic(diagnostic_context*, diagnostic_info*)
	../../gcc/diagnostic.c:1159
0x807a05 gfc_report_diagnostic
	../../gcc/fortran/error.c:782
0x807a05 gfc_warning
	../../gcc/fortran/error.c:815
0x808016 gfc_warning(int, char const*, ...)
	../../gcc/fortran/error.c:846
0x9572bf check_against_globals
	../../gcc/fortran/frontend-passes.c:5528
0x8a8b02 do_traverse_symtree
	../../gcc/fortran/symbol.c:4170
0x95f5d1 gfc_check_externals(gfc_namespace*)
	../../gcc/fortran/frontend-passes.c:5563
0x86e554 gfc_parse_file()
	../../gcc/fortran/parse.c:6499
0x8bf9bf gfc_be_parse_file
	../../gcc/fortran/f95-lang.c:212
Please submit a full bug report,


Code will follow in a minute.

Ok, this is triggered by a stupid thing in our own code, but I think the compiler should not stumble over it. The error appears when there is a routine that is defined as external, but there is also an explicit interface for it:
module m
  implicit none
  private

  interface
     subroutine GetXminM (set, xmin)
       integer, intent(in) :: set
       real, intent(out) :: xmin
     end subroutine GetXminM
  end interface

contains

  subroutine s ()
    real :: xmin
    integer :: set
    external :: GetXminM
    call GetXminM (set, xmin)
  end subroutine s

end module m

From the backtrace, that was very probably me with
https://gcc.gnu.org/git/gitweb.cgi?p=gcc.git;h=cc9a9229285a26ac12bc8de53237ce9c4d42f867 .

(In reply to Thomas Koenig from comment #3)
> From the backtrace, that was very probably me with
> https://gcc.gnu.org/git/gitweb.cgi?p=gcc.git;
> h=cc9a9229285a26ac12bc8de53237ce9c4d42f867 .

Yes, it started with this revision.

I see nothing in the standard which prohibits this, so ice-on-valid.

What we have here is, in gfc_check_externals0,

(gdb) call debug(def_sym->formal) 

|| symbol: '_formal_0'    
  type spec : (INTEGER 4)
  attributes: (VARIABLE  ARTIFICIAL DUMMY)
|| symbol: '_formal_1'    
  type spec : (REAL 4)
  attributes: (VARIABLE  ARTIFICIAL DUMMY)
(gdb) call debug(sym->formal)

|| symbol: 'set'          
  type spec : (INTEGER 4)
  attributes: (VARIABLE  DUMMY(IN))
|| symbol: 'xmin'         
  type spec : (REAL 4)
  attributes: (VARIABLE  DUMMY(OUT))

so the inferred argument list from the call does not match the
interface.  And, since the location is not set, we get an ICE
on top of it.

I think resolution should have found the interface in addition
to the EXTERNAL statement.

Two things:  We should not warn about INTENT mismatches when
we artificially generate the prototypes, and we should set a
valid gfc_locus.

Both done with the attached patch.

diff --git a/gcc/fortran/frontend-passes.c b/gcc/fortran/frontend-passes.c
index 69f9ca64c97..7768fdc25ca 100644
--- a/gcc/fortran/frontend-passes.c
+++ b/gcc/fortran/frontend-passes.c
@@ -5441,6 +5441,7 @@ check_externals_procedure (gfc_symbol *sym, locus *loc,
   gfc_current_ns = gsym->ns;
 
   gfc_get_formal_from_actual_arglist (new_sym, actual);
+  new_sym->declared_at = *loc;
   gfc_current_ns = save_ns;
 
   return 0;
diff --git a/gcc/fortran/interface.c b/gcc/fortran/interface.c
index 0cc504f4e04..e51820918b8 100644
--- a/gcc/fortran/interface.c
+++ b/gcc/fortran/interface.c
@@ -1343,7 +1343,8 @@ gfc_check_dummy_characteristics (gfc_symbol *s1, gfc_symbol *s2,
     }
 
   /* Check INTENT.  */
-  if (s1->attr.intent != s2->attr.intent)
+  if (s1->attr.intent != s2->attr.intent && !s1->attr.artificial
+      && !s2->attr.artificial)
     {
       snprintf (errmsg, err_len, "INTENT mismatch in argument '%s'",
                s1->name);

The master branch has been updated by Thomas Kथà¤nig <tkoenig@gcc.gnu.org>:

https://gcc.gnu.org/g:90efb2512d40f0a59b063aa0416a6d28849ee424

commit r11-2036-g90efb2512d40f0a59b063aa0416a6d28849ee424
Author: Thomas Koenig <tkoenig@gcc.gnu.org>
Date:   Sat Jul 11 19:16:16 2020 +0200

    Fix ICE on warning with new interface check.
    
    In the test case, there was a warning about INTENT where an EXTERNAL
    masked an interface in an outer scope, when the location of the
    symbol was not set, leading to an ICE.
    
    Two problems, two-part solution: It makes no sense to warn about
    INTENT for artificially generated formal argument lists, and the
    location should be set.
    
    gcc/fortran/ChangeLog:
    
            PR fortran/96073
            * frontend-passes.c (check_externals_procedure):  Add locus
            information for new_sym.
            * interface.c (gfc_check_dummy_characteristics): Do not warn
            about INTENT for artificially generated variables.
    
    gcc/testsuite/ChangeLog:
    
            PR fortran/96073
            * gfortran.dg/interface_48.f90: New test.

Fixed on master, closing.

... and thanks for the timely bug report!