[Bug c++/52035] New: internal compiler error: tree code ‘template_type_parm’ is not supported in LTO streams

[Bug c++/52035] New: internal compiler error: tree code ‘template_type_parm’ is not supported in LTO streams

[marbacz at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52035

             Bug #: 52035
           Summary: internal compiler error: tree code
                    ‘template_type_parm’ is not supported in LTO streams
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: major
          Priority: P3
         Component: c++
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: marbacz@gmail.com


Using GCC from svn (183666) I get an ICE while compiling QtCore with -flto:

~ # g++ -flto bug.ii
bug.ii:44:1: internal compiler error: tree code ‘template_type_parm’ is not
supported in LTO streams
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.


~ # g++ --version
g++ (svn 183666-20120128) 4.7.0 20120128 (experimental)

[Bug c++/52035] internal compiler error: tree code ‘template_type_parm’ is not supported in LTO streams

[marbacz at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52035

--- Comment #1 from Marcin Baczyński <marbacz at gmail dot com> 2012-01-29 00:46:58 UTC ---
Created attachment 26497
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=26497
Reduced test case.

[Bug c++/52035] internal compiler error: tree code ‘template_type_parm’ is not supported in LTO streams

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52035

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |ice-on-valid-code, lto
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2012-01-29
     Ever Confirmed|0                           |1
           Severity|major                       |normal

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> 2012-01-29 04:51:02 UTC ---
Confirmed, reduced testcase:
template <typename T> struct QVector {
    typedef T* iterator;
    static void insert(int n);
    typedef int size_type;
};
template <typename T> void QVector<T>::insert(size_type n) {}
void error()
{
    int n;
    QVector<int>::insert(n);
}

[Bug c++/52035] internal compiler error: tree code ‘template_type_parm’ is not supported in LTO streams

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52035

--- Comment #3 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-01-30 09:34:37 UTC ---
We're writing a function decl:

 <function_decl 0x7ffff5b9b200 insert
    type <function_type 0x7ffff5b8cbd0
        type <void_type 0x7ffff5a36bd0 void type_6 VOID
            align 8 symtab 0 alias set -1 canonical type 0x7ffff5a36bd0
            pointer_to_this <pointer_type 0x7ffff5a36c78>>
        QI
        size <integer_cst 0x7ffff5a39120 constant 8>
        unit size <integer_cst 0x7ffff5a39140 constant 1>
        align 8 symtab 0 alias set -1 canonical type 0x7ffff5a54000
        arg-types <tree_list 0x7ffff5b89960 value <integer_type 0x7ffff5b8c9d8
size_type>
            chain <tree_list 0x7ffff5a289b0 value <void_type 0x7ffff5a36bd0
void>>>
        pointer_to_this <pointer_type 0x7ffff5b9c888>>
    addressable used nothrow public static weak decl_1 decl_5 decl_6 QI
defer-output file t.C line 6 col 28 align 16 context <record_type
0x7ffff5b9c348 QVector> initial <block 0x7ffff5ba00f0>
    arguments <parm_decl 0x7ffff5a267f8 n
        type <integer_type 0x7ffff5b8c9d8 size_type sizes-gimplified type_6 SI
            size <integer_cst 0x7ffff5a392e0 constant 32>
            unit size <integer_cst 0x7ffff5a39300 constant 4>
            align 32 symtab 0 alias set -1 canonical type 0x7ffff5a365e8
precision 32 min <integer_cst 0x7ffff5a39280 -2147483648> max <integer_cst
0x7ffff5a392a0 2147483647>>
        SI file t.C line 6 col 57 size <integer_cst 0x7ffff5a392e0 32> unit
size <integer_cst 0x7ffff5a39300 4>
        align 32 context <function_decl 0x7ffff5b9b200 insert>
        arg-type <integer_type 0x7ffff5b8c9d8 size_type>>
    result <result_decl 0x7ffff5b9d100 D.2212 type <void_type 0x7ffff5a36bd0
void>
        ignored VOID file t.C line 6 col 28
        align 8 context <function_decl 0x7ffff5b9b200 insert>>
    full-name "static void QVector<T>::insert(QVector<T>::size_type) [with T =
int; QVector<T>::size_type = int]"
    pending-inline-info 0x7ffff5b9e1e0 template-info 0x7ffff5b97940
    struct-function 0x7ffff5a29320>

from there its type, its argument type chain, the first arguments type-name,
its context

 <record_type 0x7ffff5b8c690 QVector type_0 type_5 type_6 VOID
    size <integer_cst 0x7ffff5a39000 type <integer_type 0x7ffff5a360a8
bitsizetype> constant 0>
    unit size <integer_cst 0x7ffff5a1dfa0 type <integer_type 0x7ffff5a36000
sizetype> constant 0>
    align 8 symtab 0 alias set -1 canonical type 0x7ffff5b8c690
    fields <type_decl 0x7ffff5b85b80 QVector
        type <record_type 0x7ffff5b8c738 QVector type_5 VOID
            align 8 symtab 0 alias set -1 canonical type 0x7ffff5b8c690
            full-name "struct QVector<T>"
            n_parents=0 use_template=0 interface-unknown
            chain <type_decl 0x7ffff5b85a10 QVector>>
        nonlocal decl_4 VOID file t.C line 1 col 38
        align 1 context <record_type 0x7ffff5b8c690 QVector> result
<record_type 0x7ffff5b8c690 QVector>

        chain <type_decl 0x7ffff5b85cf0 iterator type <pointer_type
0x7ffff5b8c888 iterator>
            nonlocal VOID file t.C line 2 col 20
            align 1 context <record_type 0x7ffff5b8c690 QVector> result
<pointer_type 0x7ffff5b8c7e0>
           >>
    full-name "struct QVector<T>"
    n_parents=0 use_template=0 interface-unknown
    pointer_to_this <pointer_type 0x7ffff5b8cdc8> chain <type_decl
0x7ffff5b85a10 QVector>>

its type fields and then one of its types:

 <pointer_type 0x7ffff5b8c888 iterator
    type <template_type_parm 0x7ffff5b8c5e8 T type_0 type_6 VOID
        align 8 symtab 0 alias set -1 canonical type 0x7ffff5b8c5e8
       index 0 level 1 orig_level 1
        pointer_to_this <pointer_type 0x7ffff5b8c7e0> chain <type_decl
0x7ffff5b85958 T>>
    unsigned DI
    size <integer_cst 0x7ffff5a1df60 type <integer_type 0x7ffff5a360a8
bitsizetype> constant 64>
    unit size <integer_cst 0x7ffff5a1df80 type <integer_type 0x7ffff5a36000
sizetype> constant 8>
    align 64 symtab 0 alias set -1 canonical type 0x7ffff5b8c7e0>

Not sure if it's a but that the DECL_CONTEXT ends up in an uninstantiated
template?

[Bug c++/52035] internal compiler error: tree code ‘template_type_parm’ is not supported in LTO streams

[marbacz at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52035

Marcin Baczyński <marbacz at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|                            |4.6.2

--- Comment #4 from Marcin Baczyński <marbacz at gmail dot com> 2012-02-01 23:39:38 UTC ---
I think this should be marked as a 4.7 regression. I just tested 4.6.2 and it
works.

gcc version 4.6.2 (Gentoo 4.6.2 p1.3, pie-0.5.0)

[Bug c++/52035] [4.7 regression] internal compiler error: tree code ‘template_type_parm’ is not supported in LTO streams

[jason at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52035

Jason Merrill <jason at gcc dot gnu.org> changed:

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

--- Comment #5 from Jason Merrill <jason at gcc dot gnu.org> 2012-02-06 21:28:32 UTC ---
The problem here is that when we go to instantiate QVector<int>, we step
through the members in order.  When we instantiate the declaration of insert,
we're using the declaration from the out-of-class definition, which uses the
size_type typedef.  But we haven't instantiated size_type yet, so tsubst thinks
that we're instantiating it now, and we end up just returning the version from
the template.

[Bug c++/52035] [4.7 regression] internal compiler error: tree code ‘template_type_parm’ is not supported in LTO streams

[jason at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52035

Jason Merrill <jason at gcc dot gnu.org> changed:

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

[Bug c++/52035] [4.7 regression] internal compiler error: tree code ‘template_type_parm’ is not supported in LTO streams

[jason at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52035

--- Comment #6 from Jason Merrill <jason at gcc dot gnu.org> 2012-02-08 09:52:15 UTC ---
Author: jason
Date: Wed Feb  8 09:52:11 2012
New Revision: 184000

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=184000
Log:
    PR c++/52035
    * pt.c (tsubst): Strip uninstantiated typedef.

Added:
    trunk/gcc/testsuite/g++.dg/lto/pr52035_0.C
Modified:
    trunk/gcc/cp/ChangeLog
    trunk/gcc/cp/pt.c
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/testsuite/g++.dg/lto/README

[Bug c++/52035] [4.7 regression] internal compiler error: tree code ‘template_type_parm’ is not supported in LTO streams

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52035

Paolo Carlini <paolo.carlini at oracle dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|---                         |4.7.0

--- Comment #7 from Paolo Carlini <paolo.carlini at oracle dot com> 2012-02-08 22:47:12 UTC ---
Confirmed fixed.