[PATCH] testsuite: endianness in gnu_vector.exp

[PATCH] testsuite: endianness in gnu_vector.exp

[Jose E. Marchesi]

Hi.

Small patch making gnu_vector.exp to work properly in big-endian
machines.

Tested in sparc64-*-linux-gnu.

2013-10-07  Jose E. Marchesi  <jose.marchesi@oracle.com>

	* gdb.base/gnu_vector.exp: Care about endianness when casting
	scalars to vectors.

Index: testsuite/gdb.base/gnu_vector.exp
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.base/gnu_vector.exp,v
retrieving revision 1.13
diff -u -r1.13 gnu_vector.exp
--- testsuite/gdb.base/gnu_vector.exp	27 Jun 2013 18:50:30 -0000	1.13
+++ testsuite/gdb.base/gnu_vector.exp	7 Oct 2013 14:25:35 -0000
@@ -40,6 +40,14 @@
     return -1
 }
 
+# Get endianess for the scalar->vector casts
+gdb_test_multiple "show endian" "show endian" {
+    -re ".* (big|little) endian.*$gdb_prompt $" { 
+	set endian $expect_out(1,string) 
+	pass "endianness: $endian"
+    }
+}
+
 # Test printing of character vector types
 gdb_test "print c4" "\\\$$decimal = \\{1, 2, 3, 4\\}"
 gdb_test "print c4\[2\]" "\\\$$decimal = 3"
@@ -83,9 +91,14 @@
 # When casting to vector the input type must have the same length as
 # the total length of the vector.
 gdb_test "print (char4) 0x01010101" "\\\$$decimal = \\{1, 1, 1, 1\\}"
-gdb_test "print (char4) ia" "\\\$$decimal = \\{2, 0, 0, 0\\}"
 gdb_test "print (int2) lla" "\\\$$decimal = \\{1, 1\\}"
 
+if { ![string compare $endian big] } then {
+    gdb_test "print (char4) ia" "\\\$$decimal = \\{0, 0, 0, 2\\}"
+} else {
+    gdb_test "print (char4) ia" "\\\$$decimal = \\{2, 0, 0, 0\\}"
+}
+
 gdb_test "print (int2) 1" "can only cast scalar to vector of same size"
 gdb_test "print (longlong2) 2" "can only cast scalar to vector of same size"
 gdb_test "print (float2) 3" "can only cast scalar to vector of same size"

Re: [PATCH] testsuite: endianness in gnu_vector.exp

[Tom Tromey]

>>>>> "Jose" == Jose E Marchesi <jose.marchesi@oracle.com> writes:

Jose> Small patch making gnu_vector.exp to work properly in big-endian
Jose> machines.

The patch looks reasonable enough, but is this how casting to vector
types works?  That is, is the gdb result here consistent with what gcc
does?  This isn't clear from the existing test.

Tom

Re: [PATCH] testsuite: endianness in gnu_vector.exp

[Jose E. Marchesi]

    Jose> Small patch making gnu_vector.exp to work properly in big-endian
    Jose> machines.
    
    The patch looks reasonable enough, but is this how casting to vector
    types works?  That is, is the gdb result here consistent with what gcc
    does?  This isn't clear from the existing test.
    
Well, this is all I can find on the gcc manual on casting vectors:

"It is possible to cast from one vector type to another, provided they
 are of the same size (in fact, you can also cast vectors to and from
 other datatypes of the same size)."

No explicit mention to the endianness issue.  But then the behavior of
gcc indicates that it stores the elements of the vectors in memory
ordering: the following program generates different results in x86_64 (2
0) and SPARC/BE (0 2).

typedef int __attribute__((vector_size(8))) vector_t;

void
main()
{
  long i = 2;
  vector_t v = (vector_t) i;
  printf ("%d %d\n", v[0], v[1]);
}

So it looks like the test is now consistent with what gcc does...

Re: [PATCH] testsuite: endianness in gnu_vector.exp

[Tom Tromey]

>>>>> "Jose" == Jose E Marchesi <jose.marchesi@oracle.com> writes:

Jose> So it looks like the test is now consistent with what gcc does...

Thanks, I appreciate your looking.

The patch is ok.

Tom

Re: [PATCH] testsuite: endianness in gnu_vector.exp

[Jose E. Marchesi]

    >>>>> "Jose" == Jose E Marchesi <jose.marchesi@oracle.com> writes:
    
    Jose> So it looks like the test is now consistent with what gcc does...
    
    Thanks, I appreciate your looking.
    
    The patch is ok.

Committed as posted in
http://sourceware.org/ml/gdb-patches/2013-10/msg00188.html