[Bug c++/22139] New: [4.1 regression] Segfault

[Bug c++/22139] New: [4.1 regression] Segfault

[bangerth at dealii dot org]

With attached file, I get the following segfault with mainline:  
 
examples/step-18> /home/bangerth/bin/gcc-4.1-pre/bin/c++ -c step-18.ii 
step-18.ii:7102: warning: ‘__malloc__’ attribute ignored 
step-18.ii: In constructor &#8216;std::_Vector_base<_Tp, 
_Alloc>::_Vector_base(size_t, const _Alloc&) [with _Tp = Vector<double>, 
_Alloc = std::allocator<Vector<double> >]&#8217;: 
step-18.ii:85305: internal compiler error: Segmentation fault 
Please submit a full bug report, 
with preprocessed source if appropriate. 
See <URL:http://gcc.gnu.org/bugs.html> for instructions. 
 
W.

-- 
           Summary: [4.1 regression] Segfault
           Product: gcc
           Version: 4.1.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P2
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: bangerth at dealii dot org
                CC: gcc-bugs at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.1 regression] Segfault

[bangerth at dealii dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.1.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.1 regression] Segfault

[bangerth at dealii dot org]

------- Additional Comments From bangerth at dealii dot org  2005-06-21 17:58 -------
Created an attachment (id=9125)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=9125&action=view)
Preprocessed sources


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.1 regression] Segfault

[bangerth at dealii dot org]

------- Additional Comments From bangerth at dealii dot org  2005-06-21 18:00 -------
This looks like a memory problem -- the backtrace is this: 
 
(gdb) r -quiet step-18.ii -o /dev/null 
Starting 
program: /home/bangerth/bin/gcc-4.1-pre/libexec/gcc/i686-pc-linux-gnu/4.1.0/cc1plus 
-quiet step-18.ii -o /dev/null 
step-18.ii:7102: warning: ‘__malloc__’ attribute ignored 
 
Program received signal SIGSEGV, Segmentation fault. 
0x084da5ef in ggc_set_mark (p=0x14) at ggc-page.c:594 
594       return base[L1][L2]; 
(gdb) bt 
#0  0x084da5ef in ggc_set_mark (p=0x14) at ggc-page.c:594 
#1  0x0814cb1f in gt_ggc_mx_lang_tree_node (x_p=0x43683dec) at gt-cp-tree.h:69 
#2  0x0814cb88 in gt_ggc_mx_lang_tree_node (x_p=0x4305fd80) at 
gt-cp-tree.h:349 
#3  0x0814cf0d in gt_ggc_mx_lang_tree_node (x_p=0x4362c000) at 
gt-cp-tree.h:121 
#4  0x0814ce36 in gt_ggc_mx_lang_tree_node (x_p=0x42f96930) at 
gt-cp-tree.h:188 
#5  0x0814ca86 in gt_ggc_mx_lang_decl (x_p=0x40982a40) at gt-cp-tree.h:367 
#6  0x0814cfff in gt_ggc_mx_lang_tree_node (x_p=0x4362af30) at 
gt-cp-tree.h:151 
#7  0x0814cb88 in gt_ggc_mx_lang_tree_node (x_p=0x42f8ea50) at 
gt-cp-tree.h:349 
#8  0x0814d757 in gt_ggc_mx_lang_type (x_p=0x4361a1b0) at gt-cp-tree.h:441 
#9  0x0814d310 in gt_ggc_mx_lang_tree_node (x_p=0x4362557c) at 
gt-cp-tree.h:182 
#10 0x0814d214 in gt_ggc_mx_lang_tree_node (x_p=0x436255e8) at 
gt-cp-tree.h:155 
#11 0x0814cb88 in gt_ggc_mx_lang_tree_node (x_p=0x42ff3618) at 
gt-cp-tree.h:349 
#12 0x0814d232 in gt_ggc_mx_lang_tree_node (x_p=0x4364e288) at 
gt-cp-tree.h:157 
#13 0x0814ceef in gt_ggc_mx_lang_tree_node (x_p=0x436253cc) at 
gt-cp-tree.h:119 
#14 0x0814cf58 in gt_ggc_mx_lang_tree_node (x_p=0x436254a4) at 
gt-cp-tree.h:126 
#15 0x0814ce36 in gt_ggc_mx_lang_tree_node (x_p=0x42f9d738) at 
gt-cp-tree.h:188 
 
and so on, for exactly 2500 frames. 
 
W. 

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.1 regression] Segfault

[bangerth at dealii dot org]

------- Additional Comments From bangerth at dealii dot org  2005-06-21 18:03 -------
I also get essentially the same backtrace (with the call to  
  ggc_set_mark (p=0x14) 
at the top) from the 4.0.1pre CVS as of 2005-05-31, although this 
happens at a different place in the source code. I'm pretty sure the 
problem is the same, though. 
 
W. 

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.1 regression] Segfault

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-06-21 18:12 -------
Hmm,
pr22139.ii:7996: internal compiler error: in ggc_set_mark, at ggc-page.c:1259
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault

[bangerth at dealii dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[4.1 regression] Segfault   |[4.0/4.1 regression]
                   |                            |Segfault


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-06-21 18:36 -------
This worked with "3.5.0 20040909" and "4.0.0 20041124" but not with "4.0.0 20050225".

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |critical
      Known to fail|                            |4.0.0 4.1.0
      Known to work|                            |3.4.0
   Target Milestone|4.1.0                       |4.0.1


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-06-21 18:41 -------
I am going to try to reduce this with "--param ggc-min-expand=0 --param ggc-min-heapsize=0" 
which takes a long time on my poor machine.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |pinskia at gcc dot gnu dot
                   |                            |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-06-21 18:43 -------
I am starting to think this is just a stack overflow and a defect in how the GC works (or someone forgot 
chain_next which should have reduced the stack usage).

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault

[bangerth at dealii dot org]

------- Additional Comments From bangerth at dealii dot org  2005-06-21 18:48 -------
Created an attachment (id=9126)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=9126&action=view)
Smaller testcase

Attached is another testcase that has only half as many lines (~40k) and
that may be simpler to reduce...

W.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
Attachment #9125 is|0                           |1
           obsolete|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-06-21 18:55 -------
> Attached is another testcase that has only half as many lines (~40k) and
> that may be simpler to reduce...

Well it takes a long to reduce because I am also running the Ada/ACATS testsuite in the background.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-06-21 19:11 -------
Actually it is not a stack overflow but I real bug in the C++ front-end.
Hmm, we are chaning the TREE_CHAIN of error_mark node, wtf.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-06-21 21:57 -------
(In reply to comment #10)
> Actually it is not a stack overflow but I real bug in the C++ front-end.
> Hmm, we are chaning the TREE_CHAIN of error_mark node, wtf.
"I" should "a", for some reason I missed typed it.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-06-21 23:04 -------
This was most likely caused by:
2004-12-30  Mark Mitchell  <mark@codesourcery.com>

        * decl.c (duplicate_decls): Call ggc_free on declarations we will
        not be needing any longer.

The FUNCTION_DECL is still referenced for some reason after the ggc_free.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-06-21 23:07 -------
(In reply to comment #12)
> This was most likely caused by:
> 2004-12-30  Mark Mitchell  <mark@codesourcery.com>
> 
>         * decl.c (duplicate_decls): Call ggc_free on declarations we will
>         not be needing any longer.
> 
> The FUNCTION_DECL is still referenced for some reason after the ggc_free.

In fact commenting out the ggc_free, this works again so it has to be this patch as this is the same as 
reverting the patch.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mmitchel at gcc dot gnu dot
                   |                            |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-06-21 23:15 -------
We still reference the old decl in DECL_TEMPLATE_SPECIALIZATIONS of the template_decl determinant in 
this case.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-06-21 23:43 -------
This is the reason why ggc_free is considered a bad idea, because if this was really dead, it would have 
been GC'd already but it is not dead.  And isn't the reason why we moved alway from what 2.95.3 did to 
the GC is so we don't have hard to debug problems like this?

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
OtherBugsDependingO|                            |21687
              nThis|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault

[mmitchel at gcc dot gnu dot org]

------- Additional Comments From mmitchel at gcc dot gnu dot org  2005-07-06 16:57 -------
Mine.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |mark at codesourcery dot com
                   |dot org                     |
             Status|UNCONFIRMED                 |ASSIGNED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault

[mmitchel at gcc dot gnu dot org]

------- Additional Comments From mmitchel at gcc dot gnu dot org  2005-07-06 17:02 -------
Postponed until 4.0.2.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.0.1                       |4.0.2


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault

[mmitchel at gcc dot gnu dot org]

------- Additional Comments From mmitchel at gcc dot gnu dot org  2005-07-15 23:07 -------
While calling ggc_free may be a bad idea in general, it's certainly a good idea
in this case; empirically, that change resulted in a significantly faster
compiler, as the comment in the code indicates.  Furthermore, anything using
NEWDECL after we returned was broken anyhow; this is just showing us the problem.

Unfortunately, I'm not able to reproduce this problem on my machine.  Is there
any additional information about how we're using DECL_TEMPLATE_SPECIALIZATIONS?
 Or, information about your exact GC parameters to I can try to reproduce it
that way?

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-07-15 23:24 -------
I was able to reproduce the failure with the last testcase with "--param ggc-min-expand=0 --param 
ggc-min-heapsize=0".

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-07-16 02:20 -------
(In reply to comment #18)
>  Or, information about your exact GC parameters to I can try to reproduce it
> that way?
I think I know why the default GC parameters cannot reproduce this on the mainline is because the decl 
memory reduction caused us not to collect as much as often.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault

[reichelt at gcc dot gnu dot org]

------- Additional Comments From reichelt at gcc dot gnu dot org  2005-07-16 04:01 -------
Crashes with "g++ --param ggc-min-expand=1 -c PR22139.ii".


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-07-16 04:31 -------
Here is a reduced testcase, reproducible with  --param ggc-min-expand=0 --param ggc-min-
heapsize=0:
   template <int rank, int dim> class Tensor;
    template <int rank, int dim> struct SymmetricTensor {     SymmetricTensor (const Tensor<2,dim> 
&t);     friend void foo(); };
    template <> SymmetricTensor<2,2>::SymmetricTensor (const Tensor<2,2> &t) {}
    template <> SymmetricTensor<2,3>::SymmetricTensor (const Tensor<2,3> &t) {}


(Thanks to delta for this one)

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault with templates and friend functions

[mmitchel at gcc dot gnu dot org]

------- Additional Comments From mmitchel at gcc dot gnu dot org  2005-07-17 20:55 -------
Thank you for reducing the test case!  

I now understand the bug, and am working on a fix.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault with templates and friend functions

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-07-17 22:02 -------
Subject: Bug 22139

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	mmitchel@gcc.gnu.org	2005-07-17 22:02:10

Modified files:
	gcc/testsuite  : ChangeLog 
	gcc/cp         : ChangeLog cp-tree.h decl.c 
Added files:
	gcc/testsuite/g++.dg/template: friend36.C 

Log message:
	PR c++/22139
	* cp-tree.h (DECL_TEMPLATE_INFO): Improve documentation.
	* decl.c (duplicate_decls): Re-register template specializations
	for functions that have DECL_TEMLPLATE_INFO, even if they do not
	have DECL_TEMPLATE_INSTANTIATION set.
	
	PR c++/22139
	* g++.dg/template/friend36.C: New test.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&r1=1.5777&r2=1.5778
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/g++.dg/template/friend36.C.diff?cvsroot=gcc&r1=NONE&r2=1.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/ChangeLog.diff?cvsroot=gcc&r1=1.4823&r2=1.4824
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/cp-tree.h.diff?cvsroot=gcc&r1=1.1154&r2=1.1155
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/decl.c.diff?cvsroot=gcc&r1=1.1413&r2=1.1414



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault with templates and friend functions

[mmitchel at gcc dot gnu dot org]

------- Additional Comments From mmitchel at gcc dot gnu dot org  2005-07-17 22:33 -------
Fixed in 4.0.2.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault with templates and friend functions

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-07-17 22:31 -------
Subject: Bug 22139

CVSROOT:	/cvs/gcc
Module name:	gcc
Branch: 	gcc-4_0-branch
Changes by:	mmitchel@gcc.gnu.org	2005-07-17 22:30:58

Modified files:
	gcc/testsuite  : ChangeLog 
	gcc/cp         : ChangeLog cp-tree.h decl.c 
Added files:
	gcc/testsuite/g++.dg/template: friend36.C 

Log message:
	PR c++/22139
	* cp-tree.h (DECL_TEMPLATE_INFO): Improve documentation.
	* decl.c (duplicate_decls): Re-register template specializations
	for functions that have DECL_TEMLPLATE_INFO, even if they do not
	have DECL_TEMPLATE_INSTANTIATION set.
	
	PR c++/22139
	* g++.dg/template/friend36.C: New test.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-4_0-branch&r1=1.5084.2.284&r2=1.5084.2.285
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/g++.dg/template/friend36.C.diff?cvsroot=gcc&only_with_tag=gcc-4_0-branch&r1=NONE&r2=1.1.2.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-4_0-branch&r1=1.4648.2.68&r2=1.4648.2.69
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/cp-tree.h.diff?cvsroot=gcc&only_with_tag=gcc-4_0-branch&r1=1.1106.2.10&r2=1.1106.2.11
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/decl.c.diff?cvsroot=gcc&only_with_tag=gcc-4_0-branch&r1=1.1371.2.15&r2=1.1371.2.16



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139

[Bug c++/22139] [4.0/4.1 regression] Segfault with templates and friend functions

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-07-31 04:58 -------
*** Bug 19776 has been marked as a duplicate of this bug. ***

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |peter at syncad dot com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22139