Eric Lemings wrote:
-----Original Message-----
From: Martin Sebor [mailto:[EMAIL PROTECTED]
Sent: Monday, June 02, 2008 2:44 PM
To: [email protected]
Subject: Re: [STDCXX-550] Please peer review this change.
Eric Lemings wrote:
Would like to get another set of eyes on this before I submit.
$ svn diff include/deque
Index: include/deque
===================================================================
--- include/deque (revision 662487)
+++ include/deque (working copy)
@@ -749,7 +749,7 @@
void _C_insert (const iterator &__it,
_IntType __n, _IntType __x, int) {
// see 23.1.1, p9 and DR 438
- _C_insert_n (__it, __n, __x);
+ _C_insert_n (__it, __n, const_reference (__x));
I suspect this is not correct. Suppose the type of __x is
a 4 byte int and const_reference is an 8 byte long. The cast
will make _C_insert_n() to read 4 bytes past the end of __x.
A better fix might be to cast __x to value_type, as long as
doing so doesn't violate [sequence.reqmts].
Yep. Good call. Will change to value_type() cast.
Do read DR 438 before applying the cast. There are some obscure
corner cases here, e.g., IntType being a user-defined "integer-
like" type. I don't know if we support this case now, or if we
do, if it's being tested, but suffice it to say that there are
subtle differences between direct-initialization either via
a function-style cast or static_cast, or copy-initialization
(passing arguments to functions).
Martin
Brad.
See also LWG DR 438 for some background:
http://www.open-std.org/jtc1/sc22/wg21/docs/lwg-defects.html#438
Martin
