[Bug c++/13438] New: [3.4 Regression] internal compiler error: tree check: expected identifier_node, have integer_cst in c_parse_error at c-common.c:5921

[Bug c++/13438] New: [3.4 Regression] internal compiler error: tree check: expected identifier_node, have integer_cst in c_parse_error at c-common.c:5921

[pinskia at gcc dot gnu dot org]

struct A { friend typename struct B; }; 

This is a very recent regression as Phil's regression hunter only goes to 20031209.
And it is most likely caused by:
2003-12-17  Mark Mitchell  <mark@codesourcery.com>

        PR c++/10603
        PR c++/12827
        * parser.c (cp_parser_error): Help c_parse_error print good
        messages if the next token is a keyword.
        (cp_parser_parameter_declaration_list): When resynchronizing after
        a bad parameter declaration, stop if a comma is found.
        (cp_parser_parameter_declaration): Avoid backtracking.

-- 
           Summary: [3.4 Regression] internal compiler error: tree check:
                    expected identifier_node, have integer_cst in
                    c_parse_error at c-common.c:5921
           Product: gcc
           Version: 3.4.0
            Status: UNCONFIRMED
          Keywords: ice-on-invalid-code, error-recovery
          Severity: minor
          Priority: P2
         Component: c++
        AssignedTo: mark at codesourcery dot com
        ReportedBy: pinskia at gcc dot gnu dot org
                CC: gcc-bugs at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13438

[Bug c++/13438] [3.4 Regression] internal compiler error: tree check: expected identifier_node, have integer_cst in c_parse_error at c-common.c:5921

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2003-12-19 05:01 -------
This is most likely will ICE without checking also.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2003-12-19 05:01:01
               date|                            |
   Target Milestone|---                         |3.4.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13438

[Bug c++/13438] [3.4 Regression] internal compiler error: tree check: expected identifier_node, have integer_cst in c_parse_error at c-common.c:5921

[steven at gcc dot gnu dot org]

------- Additional Comments From steven at gcc dot gnu dot org  2003-12-20 14:27 -------
We call c_parse_error with a wrong "token->value" tree: 
 
1769      /* Output the MESSAGE -- unless we're parsing tentatively.  */ 
1770      if (!cp_parser_simulate_error (parser)) 
1771        { 
1772          cp_token *token; 
1773          token = cp_lexer_peek_token (parser->lexer); 
1774          c_parse_error (message, 
1775                         /* Because c_parser_error does not understand 
1776                            CPP_KEYWORD, keywords are treated like 
1777                            identifiers.  */ 
1778                         (token->type == CPP_KEYWORD ? CPP_NAME : 
token->type), 
1779                         token->value); 
1780        } 
(gdb) step 
c_parse_error (msgid=0x8392760 "expected nested-name-specifier", 
token=CPP_NAME, value=0x4001e35c) 
    at ../../mainline/gcc/c-common.c:5904 
5904      if (token == CPP_EOF) 
(...) 
5921        error ("%s before \"%s\"", string, IDENTIFIER_POINTER (value)); 
(gdb) p debug_tree (value) 
 <integer_cst 0x4001e35c type <integer_cst 0x4001e35c 1> constant 1> 
$19 = void 
 
So even though (token->type == CPP_KEYWORD), token->value is an integer 
constant tree... 

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13438

[Bug c++/13438] [3.4 Regression] internal compiler error: tree check: expected identifier_node, have integer_cst in c_parse_error at c-common.c:5921

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2003-12-20 20:34 -------
Remember there is no difference between a latency bug that gets exposed by the patch or a bug in 
the patch in GCC's mind.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13438

[Bug c++/13438] [3.4 Regression] internal compiler error: tree check: expected identifier_node, have integer_cst in c_parse_error at c-common.c:5921

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2003-12-20 23:17 -------
Here is another testcase which ICE now (with the same ICE) but this one is not error-
recovery bug:
 template<typename TYPE> struct B { bool f(); };
template<typename TYPE> struct A : public struct B<TYPE> { void g(); };

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|minor                       |normal
   Last reconfirmed|2003-12-19 05:01:01         |2003-12-20 23:17:24
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13438

[Bug c++/13438] [3.4 Regression] internal compiler error: tree check: expected identifier_node, have integer_cst in c_parse_error at c-common.c:5921

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2003-12-22 08:03 -------
Subject: Bug 13438

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	mmitchel@gcc.gnu.org	2003-12-22 08:03:15

Modified files:
	gcc/cp         : ChangeLog cp-tree.h decl.c lex.c 
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/g++.dg/parse: error8.C 

Log message:
	PR c++/13438
	* cp-tree.h (cp_tree_index): Remove CPTI_RECORD_TYPE,
	CPTI_UNION_TYPE, CPTI_ENUM_TYPE.
	(record_type_node): Remove.
	(union_type_node): Likewise.
	(enum_type_node): Likewise.
	* decl.c: Remove mention of above tree nodes in comment.
	* lex.c (cxx_init): Do not assign to record_type_node,
	union_type_node, or enum_type_node.  Simplify handling of
	class_type_node.
	
	PR c++/13438
	* g++.dg/parse/error8.C: New test.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/ChangeLog.diff?cvsroot=gcc&r1=1.3827&r2=1.3828
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/cp-tree.h.diff?cvsroot=gcc&r1=1.941&r2=1.942
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/decl.c.diff?cvsroot=gcc&r1=1.1165&r2=1.1166
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/lex.c.diff?cvsroot=gcc&r1=1.320&r2=1.321
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&r1=1.3295&r2=1.3296
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/g++.dg/parse/error8.C.diff?cvsroot=gcc&r1=NONE&r2=1.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13438

[Bug c++/13438] [3.4 Regression] internal compiler error: tree check: expected identifier_node, have integer_cst in c_parse_error at c-common.c:5921

[mmitchel at gcc dot gnu dot org]

------- Additional Comments From mmitchel at gcc dot gnu dot org  2003-12-22 08:04 -------
Fixed in GCC 3.4.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13438