[PATCH, testsuit] Add tests for PR33009

[PATCH, testsuit] Add tests for PR33009

[Gabor Loki]

[-- Attachment #1: Type: text/plain, Size: 288 bytes --]

Hi,

This patch adds some trivial checks for -frtl-abstract-sequences.
Inspired by PR33009.

Ok for trunk?

--Gabor


2007-03-03  Gabor Loki  <loki@gcc.gnu.org>

	PR 33009
	* gcc.c-torture/compile/pr11832.c: Check -frtl-abstract-sequences.
	* gcc.c-torture/compile/pr33009.c: Likewise.



[-- Attachment #2: PR33009-test.patch --]
[-- Type: text/x-patch, Size: 1722 bytes --]

Index: gcc/testsuite/gcc.c-torture/compile/pr33009.c
===================================================================
--- gcc/testsuite/gcc.c-torture/compile/pr33009.c	(revision 0)
+++ gcc/testsuite/gcc.c-torture/compile/pr33009.c	(revision 0)
@@ -0,0 +1,36 @@
+/* { dg-do compile } */
+/* { dg-options "-frtl-abstract-sequences" } */
+
+char *progName;
+int bar0 (char *, ...);
+void bar1 (char *);
+void exit (int);
+
+
+#define SAME \
+ bar0 ("%s: Bad flag `%s'\n", argv[i], argv[i] );\
+ bar1 ( progName ); \
+ exit ( 1 );
+
+
+int foo ( int argc, char *argv[] )
+{
+    int i;
+    for (i = 0; i < argc; i++) {
+	switch (argv[i][0]) {
+	case 'c':
+	    break;
+	default: 
+	
+	    SAME
+	    
+	    break;
+	}
+    }
+    for (i = 0; i < argc; i++) {
+    
+	SAME
+	
+    }
+    return 0;
+}
Index: gcc/testsuite/gcc.c-torture/compile/pr11832.c
===================================================================
--- gcc/testsuite/gcc.c-torture/compile/pr11832.c	(revision 0)
+++ gcc/testsuite/gcc.c-torture/compile/pr11832.c	(revision 0)
@@ -0,0 +1,30 @@
+/* { dg-do compile } */
+/* { dg-options "-frtl-abstract-sequences" } */
+
+int a, b, e;
+unsigned char *c;
+void foo()
+{
+  int d = 13;
+  b = -1;   
+  switch (e) {
+    case 1:
+      b++; c[b] = (unsigned char)d;
+      break;
+    case 2:
+      b++; c[b] = (unsigned char)d;
+      b++; c[b] = (unsigned char)d;
+      break;
+    case 3:
+      b++; c[b] = (unsigned char)d;
+      b++; c[b] = (unsigned char)d;
+      b++; c[b] = (unsigned char)d;
+      break;
+    default:
+      a = 1;
+      b++; c[b] = (unsigned char)d;
+      b++; c[b] = (unsigned char)d;
+      b++; c[b] = (unsigned char)d;
+      b++; c[b] = (unsigned char)d;
+  }
+}

Re: [PATCH, testsuit] Add tests for PR33009

[Ian Lance Taylor]

Gabor Loki <loki@inf.u-szeged.hu> writes:

> 2007-03-03  Gabor Loki  <loki@gcc.gnu.org>
> 
> 	PR 33009
> 	* gcc.c-torture/compile/pr11832.c: Check -frtl-abstract-sequences.
> 	* gcc.c-torture/compile/pr33009.c: Likewise.

This is OK.

Thanks.

Ian

Re: [PATCH, testsuit] Add tests for PR33009

[David Daney]

Gabor Loki wrote:
> Hi,
> 
> This patch adds some trivial checks for -frtl-abstract-sequences.
> Inspired by PR33009.
> 
> Ok for trunk?
> 
> --Gabor
> 
> 
> 2007-03-03  Gabor Loki  <loki@gcc.gnu.org>
> 
>     PR 33009
>     * gcc.c-torture/compile/pr11832.c: Check -frtl-abstract-sequences.
>     * gcc.c-torture/compile/pr33009.c: Likewise.
> 
> 

Both of these are ICEing on mipsel-linux.

C test results were clean before your patch:

http://gcc.gnu.org/ml/gcc-testresults/2008-03/msg00521.html
http://gcc.gnu.org/ml/gcc-testresults/2008-03/msg00302.html


David Daney

Re: [PATCH, testsuit] Add tests for PR33009

[David Daney]

Ian Lance Taylor wrote:
> Gabor Loki <loki@inf.u-szeged.hu> writes:
>
>   
>> 2007-03-03  Gabor Loki  <loki@gcc.gnu.org>
>>
>> 	PR 33009
>> 	* gcc.c-torture/compile/pr11832.c: Check -frtl-abstract-sequences.
>> 	* gcc.c-torture/compile/pr33009.c: Likewise.
>>     
>
> This is OK.
>   
These tests are failing on almost all targets except i686 and x86_64.

Should we just xfail them, or is someone likely to fix rtl-factoring?

On mips the failure is in compute_init_costs where the result of
make_jump_insn_raw doesn't match any insns so  computing its length
ICEs.  I have not looked an any other targets.

David Daney

Re: [PATCH, testsuit] Add tests for PR33009

[Gabor Loki]

David Daney wrote:
>>> 	* gcc.c-torture/compile/pr11832.c: Check -frtl-abstract-sequences.
>>> 	* gcc.c-torture/compile/pr33009.c: Likewise.
>>>     
> These tests are failing on almost all targets except i686 and x86_64.
> 
> Should we just xfail them, or is someone likely to fix rtl-factoring?

Maybe, but I am working on this topic.
More info and a possible fix can be found at PR33642.

> On mips the failure is in compute_init_costs where the result of
> make_jump_insn_raw doesn't match any insns so  computing its length
> ICEs.  I have not looked an any other targets.

Thanks your report.
As I see on mips-elf, the last patch from PR33642 fixed this problem.

-- Gabor Loki