This is the mail archive of the
java-patches@gcc.gnu.org
mailing list for the Java project.
[PATCH/RFA] libffi: A SH fixup and a new testcase
- From: Kaz Kojima <kkojima at rr dot iij4u dot or dot jp>
- To: gcc-patches at gcc dot gnu dot org
- Cc: java-patches at gcc dot gnu dot org, green at redhat dot com, andreast at gcc dot gnu dot org
- Date: Thu, 21 Oct 2004 09:33:04 +0900 (JST)
- Subject: [PATCH/RFA] libffi: A SH fixup and a new testcase
Hi,
SH's ffi_call_SYSV wrongly assumes that the double arguments were
aligned on stack by ffi_prep_args. This bug causes a few failures
of libjava tests with gij on SH. The appended patch fixes it and
adds a new test for it because this bug couldn't be caught with
the existing libffi tests. The result of make check is:
=== libffi Summary ===
# of expected passes 216
# of unsupported tests 2
I've checked that the new test passes also on i686-pc-linux-gnu.
OK for mainline?
FYI, with this and SH custom java-signal.h patch waiting for review
http://gcc.gnu.org/ml/gcc-patches/2004-10/msg01327.html
there are no failures in libjava test on sh4-unknown-linux-gnu.
Regards,
kaz
--
2004-10-20 Kaz Kojima <kkojima@gcc.gnu.org>
* src/sh/sysv.S (ffi_call_SYSV): Don't align for double data.
* testsuite/libffi.call/float3.c: New test case.
diff -uprN ORIG/gcc/libffi/src/sh/sysv.S LOCAL/gcc/libffi/src/sh/sysv.S
--- ORIG/gcc/libffi/src/sh/sysv.S Tue Oct 19 08:05:01 2004
+++ LOCAL/gcc/libffi/src/sh/sysv.S Tue Oct 19 10:20:25 2004
@@ -117,12 +117,6 @@ L_pass_d:
bt 1f
add #1,r3
1:
- mov r15,r0
- and #7,r0
- tst r0,r0
- bt 2f
- add #4,r15
-2:
mov #12,r0
cmp/hs r0,r3
bt/s 3f
diff -uprN ORIG/gcc/libffi/testsuite/libffi.call/float3.c LOCAL/gcc/libffi/testsuite/libffi.call/float3.c
--- ORIG/gcc/libffi/testsuite/libffi.call/float3.c Thu Jan 1 09:00:00 1970
+++ LOCAL/gcc/libffi/testsuite/libffi.call/float3.c Wed Oct 20 09:33:28 2004
@@ -0,0 +1,73 @@
+/* Area: ffi_call
+ Purpose: Check float arguments with different orders.
+ Limitations: none.
+ PR: none.
+ Originator: From the original ffitest.c */
+
+/* { dg-do run } */
+/* { dg-options -mlong-double-128 { target powerpc64*-*-* } } */
+
+#include "ffitest.h"
+#include "float.h"
+
+static double floating_1(float a, double b, long double c)
+{
+ return (double) a + b + (double) c;
+}
+
+static double floating_2(long double a, double b, float c)
+{
+ return (double) a + b + (double) c;
+}
+
+int main (void)
+{
+ ffi_cif cif;
+ ffi_type *args[MAX_ARGS];
+ void *values[MAX_ARGS];
+ double rd;
+
+ float f;
+ double d;
+ long double ld;
+
+ args[0] = &ffi_type_float;
+ values[0] = &f;
+ args[1] = &ffi_type_double;
+ values[1] = &d;
+ args[2] = &ffi_type_longdouble;
+ values[2] = &ld;
+
+ /* Initialize the cif */
+ CHECK(ffi_prep_cif(&cif, FFI_DEFAULT_ABI, 3,
+ &ffi_type_double, args) == FFI_OK);
+
+ f = 3.14159;
+ d = (double)1.0/(double)3.0;
+ ld = 2.71828182846L;
+
+ floating_1 (f, d, ld);
+
+ ffi_call(&cif, FFI_FN(floating_1), &rd, values);
+
+ CHECK(rd - floating_1(f, d, ld) < DBL_EPSILON);
+
+ args[0] = &ffi_type_longdouble;
+ values[0] = &ld;
+ args[1] = &ffi_type_double;
+ values[1] = &d;
+ args[2] = &ffi_type_float;
+ values[2] = &f;
+
+ /* Initialize the cif */
+ CHECK(ffi_prep_cif(&cif, FFI_DEFAULT_ABI, 3,
+ &ffi_type_double, args) == FFI_OK);
+
+ floating_2 (ld, d, f);
+
+ ffi_call(&cif, FFI_FN(floating_2), &rd, values);
+
+ CHECK(rd - floating_2(ld, d, f) < DBL_EPSILON);
+
+ exit (0);
+}