[Bug tree-optimization/38977] New: [4.4 Regression] bash no longer builds with profile-feedback

[Bug tree-optimization/38977] New: [4.4 Regression] bash no longer builds with profile-feedback

[rguenth at gcc dot gnu dot org]

I see

jobs.c: In function 'make_child':
jobs.c:4121: error: corrupted profile info: profile data is not flow-consistent
jobs.c:4121: error: corrupted profile info: number of executions for edge 10-11
thought to be -1726
jobs.c:4121: error: corrupted profile info: number of executions for edge 10-12
thought to be 1745
jobs.c:4121: error: corrupted profile info: number of iterations for basic
block 11 thought to be -120

this error did not exist in GCC 4.3.  In 4.4 -fprofile-correction was
introduced (not enabled by default) guarded with a new check in profile.c:

  /* Check for inconsistent basic block counts */
  inconsistent = is_inconsistent ();

  if (inconsistent)
   {
     if (flag_profile_correction)
       {
...
       }
     else
       error ("corrupted profile info: profile data is not flow-consistent");
   }

so we now unconditionally throw an error on something we ignored before.

IMHO we should either enable flag_profile_correction by default or
remove this error again.


-- 
           Summary: [4.4 Regression] bash no longer builds with profile-
                    feedback
           Product: gcc
           Version: 4.4.0
            Status: UNCONFIRMED
          Keywords: rejects-valid
          Severity: major
          Priority: P3
         Component: tree-optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: rguenth at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[rguenth at gcc dot gnu dot org]

-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.4.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[jakub at gcc dot gnu dot org]

------- Comment #1 from jakub at gcc dot gnu dot org  2009-01-27 07:02 -------
I saw this once even during gcc profiledbootstrap, so I agree the error should
go for 4.4.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[rguenth at gcc dot gnu dot org]

------- Comment #2 from rguenth at gcc dot gnu dot org  2009-01-27 16:25 -------
Mine.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |rguenth at gcc dot gnu dot
                   |dot org                     |org
             Status|UNCONFIRMED                 |ASSIGNED
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2009-01-27 16:25:02
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[rguenth at gcc dot gnu dot org]

------- Comment #3 from rguenth at gcc dot gnu dot org  2009-01-27 17:01 -------
Hm, unfortunately simply disabling that error doesn't fix my bash build.
Instead we now error on

jobs.c:4121: error: corrupted profile info: number of executions for edge 10-11
thought to be -1726
jobs.c:4121: error: corrupted profile info: number of executions for edge 10-12
thought to be 1745
jobs.c:4121: error: corrupted profile info: number of iterations for basic
block 11 thought to be -120

which is an error that is present also in 4.3.  So I guess there is some real
bug wrt profiling in gcc 4.4.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|rguenth at gcc dot gnu dot  |unassigned at gcc dot gnu
                   |org                         |dot org
             Status|ASSIGNED                    |NEW


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[rguenth at gcc dot gnu dot org]

------- Comment #4 from rguenth at gcc dot gnu dot org  2009-01-28 12:47 -------
Back in time I traced the failure back to possibly

2008-04-05  Ben Elliston  <bje@au.ibm.com>

        * tree-cfg.c (need_fake_edge_p): Return false for calls to
        builtins that return exactly once and do not throw. Cache call to
        call_expr_flags.

(last known good rev. is 133889, first known fail is 133929)


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[doko at ubuntu dot com]

------- Comment #5 from doko at ubuntu dot com  2009-01-28 17:08 -------
PR38292 shows similiar error messages


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[rguenth at gcc dot gnu dot org]

-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[rguenth at gcc dot gnu dot org]

------- Comment #6 from rguenth at gcc dot gnu dot org  2009-01-30 16:35 -------
I think this may be due to the use of setjmp and

DEF_GCC_BUILTIN        (BUILT_IN_SETJMP, "setjmp", BT_FN_INT_PTR, ATTR_NULL)

missing the fact that setjmp returns twice.

Note also the inconsistency in

  if (is_gimple_call (t)
      && fndecl
      && DECL_BUILT_IN (fndecl)
      && (call_flags & ECF_NOTHROW)
      && !(call_flags & ECF_NORETURN)
      && !(call_flags & ECF_RETURNS_TWICE))
   return false;

  if (is_gimple_call (t)
      && !(call_flags & ECF_NORETURN))
    return true;

wrt ECF_NORETURN.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[jakub at gcc dot gnu dot org]

------- Comment #7 from jakub at gcc dot gnu dot org  2009-01-30 16:54 -------
ECF_RETURNS_TWICE is set by special_function_p, doesn't need anything special
in
builtins.def.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[matz at gcc dot gnu dot org]

------- Comment #8 from matz at gcc dot gnu dot org  2009-01-30 17:12 -------
But special_function_p looks at the name of the function, which is
__builtin_setjmp in case the builtin is used explicitely:

% cat x.c
#include <setjmp.h>
jmp_buf env;
int f(void){  return __builtin_setjmp(env);}
% gdb --args cc1 x.c
(gdb) b special_function_p
...
474       if (fndecl && DECL_NAME (fndecl)
(gdb)
487           const char *name = IDENTIFIER_POINTER (DECL_NAME (fndecl));
(gdb)
488           const char *tname = name;
(gdb) p name
$3 = 0x7ffff7f60460 "__builtin_setjmp"

Up to now, it's only a theory that it might be setjmp related, purely because
it's jobs.c of bash which exhibits the error, and presumably jobs.c deals
with, well, jobs and forks, and possibly setjmps around.

But what we do know is, that its indeed Ben's patch.  Reverting it lets
bash build again.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[jakub at gcc dot gnu dot org]

------- Comment #9 from jakub at gcc dot gnu dot org  2009-01-30 17:23 -------
Ah, I thought special_function_p would strip off __builtin_ prefix from tname,
but apparently it doesn't.  I'd say it should, something like:

--- gcc/calls.c2009-01-28 12:57:50.000000000 +0100
+++ gcc/calls.c2009-01-30 18:22:26.000000000 +0100
@@ -498,10 +498,14 @@ special_function_p (const_tree fndecl, i
        && ! strcmp (name, "__builtin_alloca"))))
 flags |= ECF_MAY_BE_ALLOCA;

-      /* Disregard prefix _, __ or __x.  */
+      /* Disregard prefix _, __, __x or __builtin_.  */
       if (name[0] == '_')
 {
-  if (name[1] == '_' && name[2] == 'x')
+  if (name[1] == '_'
+      && name[2] == 'b'
+      && !strncmp (name + 3, "uiltin_", 7))
+    tname += 10;
+  else if (name[1] == '_' && name[2] == 'x')
     tname += 3;
   else if (name[1] == '_')
     tname += 2;


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[bonzini at gnu dot org]

------- Comment #10 from bonzini at gnu dot org  2009-01-31 14:38 -------
> IMHO we should either enable flag_profile_correction by default or
> remove this error again.

Agreed.


-- 

bonzini at gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |bonzini at gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[sje at cup dot hp dot com]

------- Comment #11 from sje at cup dot hp dot com  2009-02-02 16:27 -------
Has anyone tested the patch in comment #9.  I didn't test it with bash but I
did test it with Python (PR 38292) because I was looking at that bug and it
seemed to fix the problem.


-- 

sje at cup dot hp dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |sje at cup dot hp dot com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[rguenth at gcc dot gnu dot org]

------- Comment #12 from rguenth at gcc dot gnu dot org  2009-02-02 16:46 -------
I'll see if it helps bash.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[jakub at gcc dot gnu dot org]

------- Comment #13 from jakub at gcc dot gnu dot org  2009-02-04 08:18 -------
Even if it fixes just PR38292, I'd say it would be worth applying.
I'll bootstrap/regtest and submit it then.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[jakub at gcc dot gnu dot org]

------- Comment #14 from jakub at gcc dot gnu dot org  2009-02-04 19:51 -------
Subject: Bug 38977

Author: jakub
Date: Wed Feb  4 19:50:58 2009
New Revision: 143937

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=143937
Log:
        PR tree-optimization/38977
        PR gcov-profile/38292
        * calls.c (special_function_p): Disregard __builtin_
        prefix.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/calls.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[rguenth at gcc dot gnu dot org]

------- Comment #15 from rguenth at gcc dot gnu dot org  2009-02-04 22:51 -------
It appearanlty does not fix my bash build.  I'm double-checking now.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[rguenth at gcc dot gnu dot org]

------- Comment #16 from rguenth at gcc dot gnu dot org  2009-02-04 23:16 -------
jobs.c calls fork(), but special_function_p seems to miss that.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[rguenth at gcc dot gnu dot org]

------- Comment #17 from rguenth at gcc dot gnu dot org  2009-02-05 13:37 -------
Testing for fork as well doesn't fix it either.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[rguenth at gcc dot gnu dot org]

------- Comment #18 from rguenth at gcc dot gnu dot org  2009-02-05 15:31 -------
Some more details:

jobs.c: In function 'make_child':
jobs.c:4121: error: corrupted profile info: profile data is not flow-consistent
jobs.c:4121: error: corrupted profile info: number of executions for edge 10-11
thought to be -1726
jobs.c:4121: error: corrupted profile info: number of executions for edge 10-12
thought to be 1745
jobs.c:4121: error: corrupted profile info: number of iterations for basic
block 11 thought to be -120

;; basic block 8, loop depth 0, count 1745
;; prev block 7, next block 9
;; pred:       7 [100.0%]  count:1745 (fallthru)
;; succ:       9 count:1625 (true) 12 count:120 (false)
<bb 8>:
default_buffered_input.139 = default_buffered_input;
if (default_buffered_input.139 != -1)
  goto <bb 9>;
else
  goto <bb 12>;

;; basic block 9, loop depth 0, count 1625
;; prev block 8, next block 10
;; pred:       8 count:1625 (true)
;; succ:       11 count:1606 (true) 10 count:19 (false)
<bb 9>:
if (async_p == 0)
  goto <bb 11>;
else
  goto <bb 10>;

;; basic block 10, loop depth 0, count 19
;; prev block 9, next block 11
;; pred:       9 count:19 (false)
;; succ:       11 count:-1726 (true) 12 count:1745 (false)
<bb 10>:
default_buffered_input.139 = default_buffered_input;
if (default_buffered_input.139 > 0)
  goto <bb 11>;
else
  goto <bb 12>;

;; basic block 11, loop depth 0, count -120
;; prev block 10, next block 12
;; pred:       9 count:1606 (true) 10 count:-1726 (true)
;; succ:       12 count:1625 (fallthru) EXIT count:-1745 (fake)
<bb 11>:
default_buffered_input.139 = default_buffered_input;
sync_buffered_stream (default_buffered_input.139);

;; basic block 12, loop depth 0, count 3490
;; prev block 11, next block 13
;; pred:       8 count:120 (false) 10 count:1745 (false) 11 count:1625
(fallthru)
;; succ:       13 (true) 18 count:3490 (false)
<bb 12>:
pid.140 = fork ();
pid = pid.140;
if (pid < 0)
  goto <bb 13>;
else
  goto <bb 18>;

;; basic block 13, loop depth 0, count 0
;; prev block 12, next block 14
;; pred:       12 (true)
;; succ:       14 [100.0%]  (fallthru) EXIT (fake)
<bb 13>:
sys_error (&"fork"[0]);

;; basic block 18, loop depth 0, count 3490
;; prev block 17, next block 19
;; pred:       12 count:3490 (false) 17 (fallthru)
;; succ:       19 count:1745 (true) 48 count:1745 (false)
<bb 18>:
if (pid == 0)
  goto <bb 19>;
else
  goto <bb 48>;


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[rguenth at gcc dot gnu dot org]

------- Comment #19 from rguenth at gcc dot gnu dot org  2009-02-05 15:36 -------
Relevant source snippet:

  if (default_buffered_input != -1 &&
      (!async_p || default_buffered_input > 0))
    sync_buffered_stream (default_buffered_input);

  if ((pid = fork ()) < 0)
    {
      sys_error ("fork");
...
      throw_to_top_level ();
    }

  if (pid == 0)
    {


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[rguenth at gcc dot gnu dot org]

------- Comment #20 from rguenth at gcc dot gnu dot org  2009-02-05 22:13 -------
I have a patch that works.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |rguenth at gcc dot gnu dot
                   |dot org                     |org
             Status|NEW                         |ASSIGNED
   Last reconfirmed|2009-01-27 16:25:02         |2009-02-05 22:13:54
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[rguenth at gcc dot gnu dot org]

------- Comment #21 from rguenth at gcc dot gnu dot org  2009-02-06 09:06 -------
Subject: Bug 38977

Author: rguenth
Date: Fri Feb  6 09:06:39 2009
New Revision: 143982

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=143982
Log:
2009-02-06  Richard Guenther  <rguenther@suse.de>

        PR middle-end/38977
        * tree-cfg.c (need_fake_edge_p): Force a fake edge for
        fork because we may expand it as __gcov_fork.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/tree-cfg.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[rguenth at gcc dot gnu dot org]

------- Comment #22 from rguenth at gcc dot gnu dot org  2009-02-06 09:08 -------
Fixed.


-- 

rguenth at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977

[Bug tree-optimization/38977] [4.4 Regression] bash no longer builds with profile-feedback

[doko at ubuntu dot com]

------- Comment #23 from doko at ubuntu dot com  2009-02-12 13:40 -------
PR38292 is still seen with trunk 20090211 on amd64, works for i486.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38977