another .2003 issue to be verified

Thread view:

Enable EMail Alerts

Start New Thread

Thread rating:

Stefan Slapeta - 18 Mar 2005 12:05 GMT

VC 2003 calls f(char const*) for the following code, but it should call
the function template as f(char const*) is no perfect match (because of
decay).

Please could someone analyze the behaviour in VC .2005?

Thanks, Stefan

template <typename T>
void f(T const&)
{
// this one should be called!
}

void f(const char *)
{
}

int main()
{
char buf[] = "abc";
f(buf);
}

Reply to this Message

Stefan Slapeta - 18 Mar 2005 12:07 GMT

[...]

BTW, seems to no be no open issue for this; I'm going to submit one if
VC 2005 is still wrong.

Stefan

Reply to this Message

David Lowndes - 18 Mar 2005 12:31 GMT

>VC 2003 calls f(char const*) for the following code

VS2005 B1 does the same.

Dave

Reply to this Message

Stefan Slapeta - 18 Mar 2005 13:09 GMT

>>VC 2003 calls f(char const*) for the following code
>
> VS2005 B1 does the same.

Thanks. I've entered Bug FDBK22888.

Stefan

Reply to this Message

Tom Widmer - 18 Mar 2005 13:40 GMT

> VC 2003 calls f(char const*) for the following code, but it should call
> the function template as f(char const*) is no perfect match (because of
> decay).

I think it's because of the qualification adjustment actually - decay
alone (a form of Lvalue Transformation) would mean that the two were
ambiguous apart from the second being a non-template, so that would win
(change to void f(char*) to see this effect on a conforming compiler).
See 13.3.3.2/3:

"S1 is a proper subsequence of S2 (comparing the conversion sequences in
the canonical form defined by 13.3.3.1.1, excluding any Lvalue
Transformation; the identity conversion sequence is considered to be a
subsequence of any non-identity conversion sequence)"

> Please could someone analyze the behaviour in VC .2005?

Pass, sorry about the unnecessary correction!

Tom

Reply to this Message

Stefan Slapeta - 21 Mar 2005 12:08 GMT

> I think it's because of the qualification adjustment actually - decay
> alone (a form of Lvalue Transformation) would mean that the two were
[quoted text clipped - 6 lines]
> Transformation; the identity conversion sequence is considered to be a
> subsequence of any non-identity conversion sequence)"

you are perfectly right - very interesting: if you replace char[] by
char*, the template is called! (that's what confused me when I wrote this).

Stefan

Reply to this Message

another .2003 issue to be verified

Free Magazines