--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/gcc/testsuite/gcc.dg/format/ext-1.c	Fri Oct 27 22:46:09 2017 +0900
@@ -0,0 +1,126 @@
+/* Test for format extensions beyond the C standard and X/Open standard.
+   Test for printf formats.
+*/
+/* Origin: Joseph Myers <jsm28@cam.ac.uk> */
+/* { dg-do compile } */
+/* { dg-options "-std=gnu99 -Wformat" } */
+
+#include "format.h"
+
+void
+foo (quad_t q, u_quad_t uq, quad_t *qn, size_t z, size_t *zn, long long int ll,
+     unsigned long long int ull, int i, unsigned int u, double d,
+     char *s, void *p, wchar_t *ls, wint_t lc, int *n, long int l)
+{
+  /* As an extension, GCC allows the BSD length "q" for integer formats.
+     This is largely obsoleted in C99 by %j, %ll and PRId64.
+  */
+  printf ("%qd%qi%qo%qu%qx%qX%qn", q, q, uq, uq, uq, uq, qn);
+  printf ("%qf", d); /* { dg-warning "length" "bad use of %q" } */
+  printf ("%qF", d); /* { dg-warning "length" "bad use of %q" } */
+  printf ("%qe", d); /* { dg-warning "length" "bad use of %q" } */
+  printf ("%qE", d); /* { dg-warning "length" "bad use of %q" } */
+  printf ("%qg", d); /* { dg-warning "length" "bad use of %q" } */
+  printf ("%qG", d); /* { dg-warning "length" "bad use of %q" } */
+  printf ("%qa", d); /* { dg-warning "length" "bad use of %q" } */
+  printf ("%qA", d); /* { dg-warning "length" "bad use of %q" } */
+  printf ("%qc", i); /* { dg-warning "length" "bad use of %q" } */
+  printf ("%qs", s); /* { dg-warning "length" "bad use of %q" } */
+  printf ("%qp", p); /* { dg-warning "length" "bad use of %q" } */
+  printf ("%qC", lc); /* { dg-warning "length" "bad use of %q" } */
+  printf ("%qS", ls); /* { dg-warning "length" "bad use of %q" } */
+  /* With a bad length GCC wants some argument, any argument,
+     to devour with the format conversion, as a synchronisation heuristic.
+     This may get improved later.
+  */
+  printf ("%qm", i); /* { dg-warning "length" "bad use of %q" } */
+  /* As an extension, GCC allows the length "Z" as a synonym for "z".
+     This was an extension predating C99 which should now be considered
+     deprecated; use the standard "z" instead.
+  */
+  printf ("%Zd%Zi%Zo%Zu%Zx%ZX", z, z, z, z, z, z);
+  printf ("%Zn", zn);
+  printf ("%Zf", d); /* { dg-warning "length" "bad use of %Z" } */
+  printf ("%ZF", d); /* { dg-warning "length" "bad use of %Z" } */
+  printf ("%Ze", d); /* { dg-warning "length" "bad use of %Z" } */
+  printf ("%ZE", d); /* { dg-warning "length" "bad use of %Z" } */
+  printf ("%Zg", d); /* { dg-warning "length" "bad use of %Z" } */
+  printf ("%ZG", d); /* { dg-warning "length" "bad use of %Z" } */
+  printf ("%Za", d); /* { dg-warning "length" "bad use of %Z" } */
+  printf ("%ZA", d); /* { dg-warning "length" "bad use of %Z" } */
+  printf ("%Zc", i); /* { dg-warning "length" "bad use of %Z" } */
+  printf ("%Zs", s); /* { dg-warning "length" "bad use of %Z" } */
+  printf ("%Zp", p); /* { dg-warning "length" "bad use of %Z" } */
+  printf ("%ZC", lc); /* { dg-warning "length" "bad use of %Z" } */
+  printf ("%ZS", ls); /* { dg-warning "length" "bad use of %Z" } */
+  printf ("%Zm", i); /* { dg-warning "length" "bad use of %Z" } */
+  /* As an extension, GCC allows the length "L" on integer formats
+     (but not %n) as a synonym for "ll".
+     This should be considered deprecated.
+  */
+  printf ("%Ld%Li%Lo%Lu%Lx%LX", ll, ll, ull, ull, ull, ull);
+  /* As an extension, derived from syslog, GCC allows the conversion
+     specifier "m" for formatting strerror(errno).  This may be used
+     with width, precision and the "-" flag, the same as %s.
+  */
+  printf ("%m%3m%.4m%5.6m");
+  printf ("%*m", i);
+  printf ("%.*m", i);
+  printf ("%*.*m", i, i);
+  printf ("%3.*m", i);
+  printf ("%*.4m", i);
+  printf ("%-m");
+  printf ("%+m"); /* { dg-warning "flag" "bad %+m" } */
+  printf ("% m"); /* { dg-warning "flag" "bad % m" } */
+  printf ("%#m"); /* { dg-warning "flag" "bad %#m" } */
+  printf ("%0m"); /* { dg-warning "flag" "bad %0m" } */
+  printf ("%'m"); /* { dg-warning "flag" "bad %'m" } */
+  printf ("%hm", i); /* { dg-warning "length" "bad %hm" } */
+  printf ("%hhm", i); /* { dg-warning "length" "bad %hhm" } */
+  printf ("%lm", i); /* { dg-warning "length" "bad %lm" } */
+  printf ("%llm", i); /* { dg-warning "length" "bad %llm" } */
+  printf ("%jm", i); /* { dg-warning "length" "bad %jm" } */
+  printf ("%zm", i); /* { dg-warning "length" "bad %zm" } */
+  printf ("%tm", i); /* { dg-warning "length" "bad %tm" } */
+  printf ("%Lm", i); /* { dg-warning "length" "bad %Lm" } */
+  printf ("%qm", i); /* { dg-warning "length" "bad %qm" } */
+  printf ("%Zm", i); /* { dg-warning "length" "bad %Zm" } */
+  /* It should be OK to mix %m formats with $ operand number formats.  */
+  printf ("%2$ld%m%1$d", i, l);
+  /* Likewise, %m formats with width and precision should not have an
+     operand number for the %m itself.
+  */
+  printf ("%*2$.*1$m", i, i);
+  printf ("%1$*2$.*1$m", i, i); /* { dg-warning "no argument" "printf %1\$m" } */
+  /* As an extension, glibc includes the "I" flag for decimal
+     formats, to output using the locale's digits (e.g. in Arabic).
+     In GCC, we require this to be in the standard place for flags, though
+     glibc allows it also after width or precision.
+  */
+  printf ("%Id%Ii%Iu", i, i, u);
+  printf ("%Io", u); /* { dg-warning "flag" "bad use of I flag" } */
+  printf ("%Ix", u); /* { dg-warning "flag" "bad use of I flag" } */
+  printf ("%IX", u); /* { dg-warning "flag" "bad use of I flag" } */
+  printf ("%In", n); /* { dg-warning "flag" "bad use of I flag" } */
+  printf ("%If", d);
+  printf ("%IF", d);
+  printf ("%Ie", d);
+  printf ("%IE", d);
+  printf ("%Ig", d);
+  printf ("%IG", d);
+  printf ("%Ia", d); /* { dg-warning "flag" "bad use of I flag" } */
+  printf ("%IA", d); /* { dg-warning "flag" "bad use of I flag" } */
+  printf ("%Ic", i); /* { dg-warning "flag" "bad use of I flag" } */
+  printf ("%Is", s); /* { dg-warning "flag" "bad use of I flag" } */
+  printf ("%Ip", p); /* { dg-warning "flag" "bad use of I flag" } */
+  printf ("%IC", lc); /* { dg-warning "flag" "bad use of I flag" } */
+  printf ("%IS", ls); /* { dg-warning "flag" "bad use of I flag" } */
+  printf ("%Im"); /* { dg-warning "flag" "bad use of I flag" } */
+
+  /* As an extension, GCC does format checking on "unlocked"
+     i.e. thread unsafe versions of these functions.  */
+  fprintf_unlocked (stdout, "%d", i);
+  fprintf_unlocked (stdout, "%ld", i); /* { dg-warning "format" "fprintf_unlocked" } */
+  printf_unlocked ("%d", i);
+  printf_unlocked ("%ld", i); /* { dg-warning "format" "printf_unlocked" } */
+}