[Bug regression/35065] New: [4.3 regression]

[Bug regression/35065] New: [4.3 regression]

[debian-gcc at lists dot debian dot org]

[forwarded from http://bugs.debian.org/461528]

rechecked with trunk 20080202, removing -funroll-loops works ok.

  Matthias

Trying to compile the VLC source tree with gcc-4.3 appears to cause an
infinite loop during compilation. At least, cc1 keeps burning the whole
CPU for over 20 minutes whereby gcc-4.2 succeeds after a few seconds.

# gcc-4.3 -c h264-precomp.i -g -O2 -ffast-math -funroll-loops -fPIC


-- 
           Summary: [4.3 regression]
           Product: gcc
           Version: 4.3.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: regression
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: debian-gcc at lists dot debian dot org
GCC target triplet: i486-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35065

[Bug regression/35065] [4.3 regression] infinite loop while compiling VLC media player

[debian-gcc at lists dot debian dot org]

------- Comment #1 from debian-gcc at lists dot debian dot org  2008-02-03 14:10 -------
Created an attachment (id=15086)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=15086&action=view)
preprocessed source


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35065

[Bug middle-end/35065] [4.3 Regression] infinite loop while compiling VLC media player

[rguenth at gcc dot gnu dot org]

------- Comment #2 from rguenth at gcc dot gnu dot org  2008-02-03 15:15 -------
We are looping in vt_find_locations ().  Probably a similar issue as PR31412.
Confirmed with r132071.  I'm trying to reduce the testcase.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rguenth at gcc dot gnu dot
                   |                            |org, matz at suse dot de
          Component|regression                  |middle-end
            Summary|[4.3 regression] infinite   |[4.3 Regression] infinite
                   |loop while compiling VLC    |loop while compiling VLC
                   |media player                |media player
   Target Milestone|---                         |4.3.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35065

[Bug middle-end/35065] [4.3 Regression] infinite loop while compiling VLC media player

[rguenth at gcc dot gnu dot org]

------- Comment #3 from rguenth at gcc dot gnu dot org  2008-02-03 17:43 -------
Created an attachment (id=15088)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=15088&action=view)
reduced testcase

Hangs with -O2 -g -funroll-loops on i?86.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35065

[Bug middle-end/35065] [4.3 Regression] infinite loop while compiling VLC media player in vt_find_locations

[rguenth at gcc dot gnu dot org]

------- Comment #4 from rguenth at gcc dot gnu dot org  2008-02-03 17:44 -------
Confirmed.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
 GCC target triplet|i486-linux-gnu              |i?86-*-*
           Keywords|                            |compile-time-hog
   Last reconfirmed|0000-00-00 00:00:00         |2008-02-03 17:44:48
               date|                            |
            Summary|[4.3 Regression] infinite   |[4.3 Regression] infinite
                   |loop while compiling VLC    |loop while compiling VLC
                   |media player                |media player in
                   |                            |vt_find_locations


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35065

[Bug debug/35065] [4.3 Regression] infinite loop while compiling VLC media player in vt_find_locations

[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=35065

[Bug debug/35065] [4.3 Regression] infinite loop while compiling VLC media player in vt_find_locations

[zadeck at naturalbridge dot com]

------- Comment #5 from zadeck at naturalbridge dot com  2008-02-12 14:56 -------
Richi,

I looked at this code once but I really do not know this code at all and really
do not want to learn it.  It will take a fair amount of time to try to figure
out what the underlying dataflow problem is and then figure out how to modify
it so that it is strictly monotonic (i.e. guaranteed to converge).

However, given that this is just improving debugging information, it should be
possible to just hack a count of the number of iterations and just stop after
20, i.e the way that flow used to do it.  

If this is a blocker, this is what I would do to get the release out of the
door.  

Kenny


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35065

[Bug debug/35065] [4.3 Regression] infinite loop while compiling VLC media player in vt_find_locations

[rguenther at suse dot de]

------- Comment #6 from rguenther at suse dot de  2008-02-12 15:03 -------
Subject: Re:  [4.3 Regression] infinite loop while compiling
 VLC media player in vt_find_locations

On Tue, 12 Feb 2008, zadeck at naturalbridge dot com wrote:

> 
> 
> ------- Comment #5 from zadeck at naturalbridge dot com  2008-02-12 14:56 -------
> Richi,
> 
> I looked at this code once but I really do not know this code at all and really
> do not want to learn it.  It will take a fair amount of time to try to figure
> out what the underlying dataflow problem is and then figure out how to modify
> it so that it is strictly monotonic (i.e. guaranteed to converge).
> 
> However, given that this is just improving debugging information, it should be
> possible to just hack a count of the number of iterations and just stop after
> 20, i.e the way that flow used to do it.  
> 
> If this is a blocker, this is what I would do to get the release out of the
> door.  

Matz has analyzed this bug already and a fix is in his mind.

Richard.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35065

[Bug debug/35065] [4.3 Regression] infinite loop while compiling VLC media player in vt_find_locations

[matz at gcc dot gnu dot org]

------- Comment #7 from matz at gcc dot gnu dot org  2008-02-12 15:36 -------
The immediate problem is that the solution iterates between two orders.
On my machine (i686) it all happen around basic block 231.  The even
iterations (after 128 iterations) the var-track info for bb 231 starts with:

Basic block 231:
IN:
Stack adjustment: 0
Reg 0: i+0 i_cycle+0
Reg 1: i+0

the odd iterations it's:

Basic block 231:
IN:
Stack adjustment: 0
Reg 0: i_cycle+0 i+0
Reg 1: i+0

so the problem is in the hardreg->expr association.  One time hardreg 0
contains i_cycle and i, the other time it contains i and i_cycle.  That's
of course completely equivalent, but leads to subtle differences in the 
the OUT set (for some yet unknown reason, I think probably _there_ the
actual problem lies).

The cause for these different IN sets is that the reg0 info isn't the join
(i.e. intersection) but the union of information from the predecessors.
Predecessor A has "reg0: i" in OUT, predecessor B has "reg0: i_cycle" in OUT.
(Note: reg0 never contains both i and i_cycle in any OUT set).  The code
in vt_find_locations() does a _union_ of the information of the predecessors,
i.e. it joins not only the common parts of the info but also the differences.
The joining happens via linked lists, hence it is inherently dependend on
the order the elements are seen.

This and the problem that equivalent but differently ordered INs lead
to different OUT sets, together with the fact that 231 is a loop basic block
(i.e. feeds itself), leads to the oscillation.

The fix I had in mind is to perhaps stabilize all this by not unioning
but intersecting the predecessors OUT sets.  This actually makes more sense,
as surely if reg0 contain "i" at the end on block A, and "i_cycle" at the
end of block B, it clearly has indeterminate value at the join point of
both blocks, except if "i" and "i_cycle" somehow are equivalent.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35065

[Bug debug/35065] [4.3 Regression] infinite loop while compiling VLC media player in vt_find_locations

[matz at gcc dot gnu dot org]

------- Comment #8 from matz at gcc dot gnu dot org  2008-02-12 16:06 -------
Actually the code in var-tracking.c does handle the situation that a reg
contains multiple decls.  Or better it tries to.  If there weren't an
obvious bug in clobber_variable_part().  With that fixed the looping doesn't
occur anymore, as the OUT set now doesn't depend on the order of IN anymore,
as it should be:

@@ -2595,6 +2655,8 @@ clobber_variable_part (dataflow_set *set
                              pool_free (attrs_pool, anode);
                              *anextp = anext;
                            }
+                         else
+                           anextp = &anode->next;
                        }
                    }


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35065

[Bug debug/35065] [4.3 Regression] infinite loop while compiling VLC media player in vt_find_locations

[matz at gcc dot gnu dot org]

------- Comment #9 from matz at gcc dot gnu dot org  2008-02-13 14:54 -------
Subject: Bug 35065

Author: matz
Date: Wed Feb 13 14:53:59 2008
New Revision: 132286

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=132286
Log:
        PR debug/35065
        * var-tracking.c (clobber_variable_part): Correctly traverse the
        list.

        * gcc.dg/pr35065.c: New testcase.

Added:
    trunk/gcc/testsuite/gcc.dg/pr35065.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/var-tracking.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35065

[Bug debug/35065] [4.3 Regression] infinite loop while compiling VLC media player in vt_find_locations

[matz at gcc dot gnu dot org]

------- Comment #10 from matz at gcc dot gnu dot org  2008-02-13 15:00 -------
This is fixed now.


-- 

matz at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35065

[Bug debug/35065] [4.3 Regression] infinite loop while compiling VLC media player in vt_find_locations

[dominiq at lps dot ens dot fr]

------- Comment #11 from dominiq at lps dot ens dot fr  2008-02-13 16:00 -------
The test gcc.dg/pr35065.c fails on i686-apple-darwin9 with:

FAIL: gcc.dg/pr35065.c (test for excess errors)
Excess errors:
/opt/gcc/gcc-4.3-work/gcc/testsuite/gcc.dg/pr35065.c:33: warning:
initialization makes pointer from integer without a cast
/opt/gcc/gcc-4.3-work/gcc/testsuite/gcc.dg/pr35065.c:129: warning: incompatible
implicit declaration of built-in function 'malloc'
/opt/gcc/gcc-4.3-work/gcc/testsuite/gcc.dg/pr35065.c:138: warning:
initialization makes pointer from integer without a cast
/opt/gcc/gcc-4.3-work/gcc/testsuite/gcc.dg/pr35065.c:185: warning: assignment
makes pointer from integer without a cast
/opt/gcc/gcc-4.3-work/gcc/testsuite/gcc.dg/pr35065.c:214: warning: assignment
makes pointer from integer without a cast
/opt/gcc/gcc-4.3-work/gcc/testsuite/gcc.dg/pr35065.c:92: warning: 'Close' used
but never defined

Not that compiled with -c instead of -S, I have the additional errors:

/var/folders/iU/iUj3xngxGYe3MPCc0TZUcE+++TI/-Tmp-//ccnJAmUR.s:21:non-relocatable
subtraction expression, "_Close" minus "L00000000001$pb"
/var/folders/iU/iUj3xngxGYe3MPCc0TZUcE+++TI/-Tmp-//ccnJAmUR.s:21:symbol:
"_Close" can't be undefined in a subtraction expression


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35065

[Bug debug/35065] [4.3 Regression] infinite loop while compiling VLC media player in vt_find_locations

[matz at gcc dot gnu dot org]

------- Comment #12 from matz at gcc dot gnu dot org  2008-02-13 16:18 -------
Whoops, sorry.  I fixed the warnings in a different copy of the testcase
than the one I committed :-(  Committed the right version now.  (Yes, it still
tests what it should test, without the var-tracking patch it runs into an
endless loop).


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35065

[Bug debug/35065] [4.3 Regression] infinite loop while compiling VLC media player in vt_find_locations

[jakub at gcc dot gnu dot org]

------- Comment #13 from jakub at gcc dot gnu dot org  2008-02-13 22:24 -------
This is not sufficient.
block_New, nal_get_annexeb, block_ChainGather still don't have prototypes and
thus:
warning: cast to pointer from integer of different size
is reported on each of the casts of these function calls to (block_t *).


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35065

[Bug debug/35065] [4.3 Regression] infinite loop while compiling VLC media player in vt_find_locations

[matz at gcc dot gnu dot org]

------- Comment #14 from matz at gcc dot gnu dot org  2008-02-13 23:03 -------
Sigh.  I've tested the changed testcase only on 32bit :-(  Update to r132300.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35065