Apparently Intel C++ 10.025 doesn't consider a operator= to be a
copy-assignment operator if it is const, so the test fails to compile when
the compiler attempts to generate its own definition. According to 12.8 p9
and p10, this code _should_ be legal. The problem does not occur with 10.026
on Windows [I don't have access to 10.025 on Windows or 10.026 on Linux], so
I have not filed an issue with Intel.
Here is a simple testcase and the compile result...
[EMAIL PROTECTED] tests]$ cat t.cpp
struct S {
const S* const self;
S () : self (this) { }
S (const S &s) : self (this) { }
void operator=(const S &s) const { }
};
int main () {
S().operator=(S());
return 0;
}
[EMAIL PROTECTED] tests]$ icc -V
Intel(R) C Compiler for applications running on Intel(R) 64, Version
10.0 Build 20070613 Package ID: l_cc_c_10.0.025
Copyright (C) 1985-2007 Intel Corporation. All rights reserved.
[EMAIL PROTECTED] tests]$ icc -c t.cpp
t.cpp(2): error: implicitly generated assignment operator cannot copy:
const member "S::self"
struct S {
^
detected during implicit generation of "S &S::operator=(const
S &)" at line 10
compilation aborted for t.cpp (code 2)
[EMAIL PROTECTED] tests]$
Here is a simple patch to avoid the issue.
Index: 26.valarray.transcend.stdcxx-315.cpp
===================================================================
--- 26.valarray.transcend.stdcxx-315.cpp (revision 589749)
+++ 26.valarray.transcend.stdcxx-315.cpp (working copy)
@@ -48,7 +48,7 @@
~S () { pass = pass && this == self; }
operator double () const { pass = pass && this == self; return 1.0; }
- void operator=(const S &s) const {
+ void operator=(const S &s) {
pass = pass && &s == s.self && this == self;
}
S operator- () const { pass = pass && this == self; return *this; }
--
View this message in context:
http://www.nabble.com/-PATCH--26.valarray.transcend.stdcxx-315.cpp-fails-to-compile-on-Intel-C%2B%2B-10.025-tf4714891.html#a13477604
Sent from the stdcxx-dev mailing list archive at Nabble.com.
