The following code incorrectly compiles, and worse returns incorrect result. on i686-pc-linux-gnu, both 4.1.0 and 4.1.1 -- start snip #include <stddef.h> // (incomplete) template class that encapsulates an array of chars template<int i> class CharArray { public: char d[i]; char& operator [] ( int indx ) { return d[indx]; } }; struct Foo { int i; CharArray<44> caCharArray; }; int main( int argc, char* argv[] ) { Foo foo; int i; // Shouldn't compile, and worse results in incorrect offset. i = offsetof(Foo, caCharArray[0]); } -- end snip variables as an array index correctly give an error, but unfortunatly not everything that looks like an constant array index is.
For some reason, this also passes icc. Interestingly, it isn't even the case that offsetof manages to look through operator[] -- removing the body of operator[] doesn't affect the result. W.
This gives an ICE-on-invalid. template<int i> struct A { char d[i]; char &operator [] ( int indx ) { return d[indx]; } }; struct B { A<44> a; }; int main() { return __builtin_offsetof(B, a[0]); } Here, fold_offsetof_1 does not expect a CALL_EXPR to be there and dies. The reason for the reporter's bug is similar but we do not ICE because we have another INDIRECT_REF because of dereferencing the "char&" reference, and the tree is like INDIRECT_REF( CALL_EXPR ( operator[], INDIRECT_REF(null), 0 ) ) fold_offsetof_1 does not expect anything inside the INDIRECT_REF, and blindly returns 0. Testing a patch.
Subject: Bug number PR28573 A patch for this bug has been added to the patch tracker. The mailing list url for the patch is http://gcc.gnu.org/ml/gcc-patches/2006-08/msg00170.html
Subject: Bug 28573 Author: bonzini Date: Thu Aug 17 07:02:55 2006 New Revision: 116208 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=116208 Log: 2006-08-17 Paolo Bonzini <bonzini@gnu.org> PR c++/28573 * c-common.c (fold_offsetof_1): Add an argument and recurse down to it or the INTEGER_CST. Fail on a CALL_EXPR. (fold_offsetof): Pass new argument to fold_offsetof_1. * c-parser.c (c_parser_postfix_expression): Don't include a NULL operand into an INDIRECT_REF. * c-typeck.c (build_unary_op): Adjust call to fold_offsetof. cp: 2006-08-17 Paolo Bonzini <bonzini@gnu.org> PR c++/28573 * semantics.c (finish_offsetof): Add new argument to fold_offsetof. testsuite: 2006-08-17 Paolo Bonzini <bonzini@gnu.org> PR c++/28573 * g++.dg/parse/offsetof6.C: New test. * g++.dg/parse/offsetof7.C: New test. Added: trunk/gcc/testsuite/g++.dg/parse/offsetof6.C trunk/gcc/testsuite/g++.dg/parse/offsetof7.C Modified: trunk/gcc/ChangeLog trunk/gcc/c-common.c trunk/gcc/c-common.h trunk/gcc/c-parser.c trunk/gcc/c-typeck.c trunk/gcc/cp/ChangeLog trunk/gcc/cp/semantics.c trunk/gcc/testsuite/ChangeLog
won't fix in GCC-4.0.x. Adjustine milestone.
Closing 4.1 branch.