Re: Seeking patch for ambiguous conversion

[VÃclav Haisman <V dot Haisman at sh dot cvut dot cz>]

Earl Chew wrote:
> Is there a recent (3.4.2 or later) patch for the following problem?
> 
> Consider the short program fragment below. There are two failures, and
> one success under gcc.
> 
> The entire fragment compiles ok with the latest Microsoft Visual Studio
> and even one version before that.
> 
> With gcc 3.4.1 (Cygwin) and gcc 3.4.2, I obtain error messages like:
> 
> ambig.cpp: In function `bool fail1(const X&)':
> ambig.cpp:16: error: ambiguous overload for 'operator==' in 'x == 4660'
> ambig.cpp:16: note: candidates are: operator==(int, int) <built-in>
> ambig.cpp:5: note:  bool operator==(short unsigned int, const C&)
> ambig.cpp: In function `bool fail2(const X&)':
> ambig.cpp:21: error: ambiguous overload for 'operator==' in 'x == 4660u'
> ambig.cpp:21: note: candidates are: operator==(int, int) <built-in>
> ambig.cpp:5: note:  bool operator==(short unsigned int, const C&)
> 
> It seems the compiler will not prefer the conversion from X to
> unsigned short in order to compile the (subsequent unsigned short to
> unsigned short) comparison and considers the competing conversions
> to C as viable.
> 
> Earl Chew
> 
> -------------------------------------------------------------------------------
> 
> 
> class C
> {
> public:
>     C(unsigned short);
>     friend bool operator==(unsigned short, const C&);
> };
> 
> class X
> {
> public:
>     operator unsigned short() const;
> };
> 
> bool fail1(const X& x)
> {
>     return x == 0x1234;
> }
> 
> bool fail2(const X& x)
> {
>     return x == (unsigned short) 0x1234;
> }
> 
> bool succeeds(const X& x)
> {
>     return (unsigned short) x == 0x1234;
> }
> 
GCC 4.1 compiles it. Cygwin's GCC is a bit old.

--
VH

Attachment: signature.asc
Description: OpenPGP digital signature