[Bug middle-end/51998] New: compiler hangs on self-recursive alias attribute

[Bug middle-end/51998] New: compiler hangs on self-recursive alias attribute

[vries at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51998

             Bug #: 51998
           Summary: compiler hangs on self-recursive alias attribute
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: vries@gcc.gnu.org


inline-2.c:
...
static void f (void) __attribute__((alias("f")));

void g ()
{
  f ();
}
...

hangs (with compiler build with r183325):
...
$ gcc inline-2.c -O2 -S
...

a mutually recursive version has the same problem:
...
static void f (void) __attribute__((alias("g")));
static void g (void) __attribute__((alias("f")));

void h ()
{
  f ();
}
...

The compiler is stuck in this loop in cgraph_function_or_thunk_node:
...
1042      while (node)
(gdb) 
1044          if (node->alias && node->analyzed)
(gdb) 
1045        node = cgraph_alias_aliased_node (node);
(gdb) 
1042      while (node)
...

The following tentative patch allows the compiler to abort compilation:
...
Index: cgraph.h
===================================================================
--- cgraph.h (revision 183325)
+++ cgraph.h (working copy)
@@ -27,6 +27,7 @@ along with GCC; see the file COPYING3.
 #include "tree.h"
 #include "basic-block.h"
 #include "function.h"
+#include "diagnostic-core.h"
 #include "ipa-ref.h"    /* FIXME: inappropriate dependency of cgraph on IPA. 
*/

 enum availability
@@ -1037,12 +1038,17 @@ cgraph_function_node (struct cgraph_node
 static inline struct cgraph_node *
 cgraph_function_or_thunk_node (struct cgraph_node *node, enum availability
*availability)
 {
+  struct cgraph_node *start = node;
   if (availability)
     *availability = cgraph_function_body_availability (node);
   while (node)
     {
       if (node->alias && node->analyzed)
-    node = cgraph_alias_aliased_node (node);
+    {
+      node = cgraph_alias_aliased_node (node);
+      if (start == node)
+        fatal_error ("function %q+D part of alias cycle", start->decl);
+    }
       else
     return node;
       if (node && availability)
...

with an error message:
...
$ gcc inline-2.c inline-3.c -O2 -S
inline-2.c:1:13: fatal error: function ‘f’ part of alias cycle
compilation terminated.
inline-3.c:1:13: fatal error: function ‘f’ part of alias cycle
compilation terminated.
...

[Bug middle-end/51998] compiler hangs on self-recursive alias attribute

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51998

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2012-01-25
                 CC|                            |hubicka at gcc dot gnu.org
     Ever Confirmed|0                           |1

--- Comment #1 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-01-25 14:27:09 UTC ---
Honza?

[Bug middle-end/51998] compiler hangs on self-recursive alias attribute

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51998

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-01-25 14:43:46 UTC ---
I think fatal_error is undesirable, you should error on it somewhere and just
drop the alias attribute.

[Bug middle-end/51998] compiler hangs on self-recursive alias attribute

[vries at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51998

--- Comment #3 from vries at gcc dot gnu.org 2012-01-25 16:33:12 UTC ---
(In reply to comment #2)
> I think fatal_error is undesirable, you should error on it somewhere and just
> drop the alias attribute.

Jakub, 

like this? :
...
Index: cgraph.h
===================================================================
--- cgraph.h (revision 183325)
+++ cgraph.h (working copy)
@@ -27,6 +27,7 @@ along with GCC; see the file COPYING3.
 #include "tree.h"
 #include "basic-block.h"
 #include "function.h"
+#include "diagnostic-core.h"
 #include "ipa-ref.h"    /* FIXME: inappropriate dependency of cgraph on IPA. 
*/

 enum availability
@@ -1037,12 +1038,21 @@ cgraph_function_node (struct cgraph_node
 static inline struct cgraph_node *
 cgraph_function_or_thunk_node (struct cgraph_node *node, enum availability
*availability)
 {
+  struct cgraph_node *start = node;
   if (availability)
     *availability = cgraph_function_body_availability (node);
   while (node)
     {
       if (node->alias && node->analyzed)
-    node = cgraph_alias_aliased_node (node);
+    {
+      node = cgraph_alias_aliased_node (node);
+      if (start == node)
+        {
+          node->alias = false;
+          error ("function %q+D part of alias cycle", start->decl);
+          return node;
+        }
+    }
       else
     return node;
       if (node && availability)
...

now normal rather than fatal errors:
...
$ gcc inline-2.c inline-3.c -O2 -S
inline-2.c:1:13: error: function ‘f’ part of alias cycle
inline-3.c:1:13: error: function ‘f’ part of alias cycle
...

[Bug middle-end/51998] compiler hangs on self-recursive alias attribute

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51998

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-01-25 16:38:05 UTC ---
I think remove_attribute would be desirable too.  But I wonder if it can't be
detected earlier than here.  In any case, I'd like to hear Honza on this.

[Bug middle-end/51998] compiler hangs on self-recursive alias attribute

[hubicka at ucw dot cz]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51998

--- Comment #5 from Jan Hubicka <hubicka at ucw dot cz> 2012-01-26 14:23:21 UTC ---
> I think remove_attribute would be desirable too.  But I wonder if it can't be
> detected earlier than here.  In any case, I'd like to hear Honza on this.

I was under impression that varasm code on aliases should already detect
cycles.
I guess best time to detect cycle is when alias node is created, I will prepare
patch
for this.

Honza

[Bug middle-end/51998] compiler hangs on self-recursive alias attribute

[hubicka at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51998

--- Comment #6 from Jan Hubicka <hubicka at gcc dot gnu.org> 2012-01-31 13:56:41 UTC ---
Created attachment 26534
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=26534
Patch in testing

Hi,
I am testing the attached patch. It also plugs symmetric problem on varpool.
The cycle testing code in varasm I was mentioning is used for weakrefs only.
Porbably that checks could go, but I would preffer to wait for 4.8. There are
too many changes in 4.7 concerning aliases already.

Honza

[Bug middle-end/51998] compiler hangs on self-recursive alias attribute

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51998

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-01-31 14:50:56 UTC ---
I guess in varpool.c you shouldn't talk about function in the message.

[Bug middle-end/51998] compiler hangs on self-recursive alias attribute

[hubicka at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51998

--- Comment #8 from Jan Hubicka <hubicka at gcc dot gnu.org> 2012-02-02 13:29:34 UTC ---
Author: hubicka
Date: Thu Feb  2 13:29:31 2012
New Revision: 183836

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=183836
Log:

    PR middle-end/51998
    * cgraphunit.c (cgraph_analyze_function): Break cyclic aliases.
    * varpool.c (varpool_analyze_pending_decls): Likewise.

    * testsuite/gcc.dg/alias-12.c: New testcase.
    * testsuite/gcc.dg/alias-13.c: New testcase.

Added:
    trunk/gcc/testsuite/gcc.dg/alias-12.c
    trunk/gcc/testsuite/gcc.dg/alias-13.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/cgraphunit.c
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/varpool.c

[Bug middle-end/51998] compiler hangs on self-recursive alias attribute

[hubicka at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51998

Jan Hubicka <hubicka at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED

--- Comment #9 from Jan Hubicka <hubicka at gcc dot gnu.org> 2012-02-02 13:32:20 UTC ---
Updated the error message (thanks, Jakub) and comitted.