Constant assignments to wchar allow values greater than wchar.max. http://svn.kuehne.cn/dstress/nocompile/wchar_04.d

Constant assignments to wchar allow values smaller than wchar.min. http://svn.kuehne.cn/dstress/nocompile/wchar_06.d

Thomas

Thomas Kuehne schrieb am Wed, 17 Nov 2004 09:45:12 +0100:
> Constant assignments to wchar allow values greater than wchar.max. http://svn.kuehne.cn/dstress/nocompile/wchar_04.d
>
> Constant assignments to wchar allow values smaller than wchar.min. http://svn.kuehne.cn/dstress/nocompile/wchar_06.d

dchar has the same problems.

http://svn.kuehne.cn/dstress/nocompile/dchar_04.d http://svn.kuehne.cn/dstress/nocompile/dchar_06.d

Thomas

I don't believe this is a bug. It is analogous to:

    byte b = 1000;

which is allowed as well.

"Thomas Kuehne" <thomas-dloop@kuehne.thisisspam.cn> wrote in message news:n9ur62-jt9.ln1@kuehne.cn...
>
> Thomas Kuehne schrieb am Wed, 17 Nov 2004 09:45:12 +0100:
> > Constant assignments to wchar allow values greater than wchar.max. http://svn.kuehne.cn/dstress/nocompile/wchar_04.d
> >
> > Constant assignments to wchar allow values smaller than wchar.min. http://svn.kuehne.cn/dstress/nocompile/wchar_06.d
>
> dchar has the same problems.
>
> http://svn.kuehne.cn/dstress/nocompile/dchar_04.d http://svn.kuehne.cn/dstress/nocompile/dchar_06.d
>
> Thomas

Can you please comment on following findings?

fail:
	short A = 32768;	// short.max+1
	byte  B = 128;		// byte.max+1
	char  C = 256;		// char.max+1
pass:
	short a = -32769;	// short.min-1
	byte  b = -129;		// byte.min-1
	char  c = -1;		// char.min-1
	wchar D = 655536;	// wchar.max+1
	wchar d = -1;		// wchar.min-1
	dchar E = 1114112;	// dchar.max+1
	dchar e = -1;		// dchar.min-1

The only documentation I found for this was "lex.html".

In fact some funny consequences occur. According to the integer
resolution table "-1" is not considered to be an integer literal
- e.g. ints have a range from 0 to int.max. This explains why no lower
borders are checked. But why are the upper borders of short, byte and
char checked where as neither (d/w)char's upper nor lower border is
checked?

Just for clarification: I am only talking about constant assignments known at compile time, not about any overflow issues during the execution.

Thomas

Walter schrieb am Fri, 19 Nov 2004 14:18:28 -0800:
> I don't believe this is a bug. It is analogous to:
>
>     byte b = 1000;
>
> which is allowed as well.
>
>> Thomas Kuehne schrieb am Wed, 17 Nov 2004 09:45:12 +0100:
>> > Constant assignments to wchar allow values greater than wchar.max. http://svn.kuehne.cn/dstress/nocompile/wchar_04.d
>> >
>> > Constant assignments to wchar allow values smaller than wchar.min. http://svn.kuehne.cn/dstress/nocompile/wchar_06.d
>>
>> dchar has the same problems.
>>
>> http://svn.kuehne.cn/dstress/nocompile/dchar_04.d http://svn.kuehne.cn/dstress/nocompile/dchar_06.d

Walter wrote:

> I don't believe this is a bug. It is analogous to:
> 
>     byte b = 1000;
> 
> which is allowed as well.

gdc 0.8 doesn't like that code:
"cannot implicitly convert expression 1000 of type int to byte"

--anders

I obviously need to clean this up.

"Thomas Kuehne" <thomas-dloop@kuehne.thisisspam.cn> wrote in message news:rra272-427.ln1@kuehne.cn...
>
>
> Can you please comment on following findings?
>
> fail:
> short A = 32768; // short.max+1
> byte  B = 128; // byte.max+1
> char  C = 256; // char.max+1
> pass:
> short a = -32769; // short.min-1
> byte  b = -129; // byte.min-1
> char  c = -1; // char.min-1
> wchar D = 655536; // wchar.max+1
> wchar d = -1; // wchar.min-1
> dchar E = 1114112; // dchar.max+1
> dchar e = -1; // dchar.min-1
>
> The only documentation I found for this was "lex.html".
>
> In fact some funny consequences occur. According to the integer
> resolution table "-1" is not considered to be an integer literal
> - e.g. ints have a range from 0 to int.max. This explains why no lower
> borders are checked. But why are the upper borders of short, byte and
> char checked where as neither (d/w)char's upper nor lower border is
> checked?
>
> Just for clarification: I am only talking about constant assignments known at compile time, not about any overflow issues during the execution.
>
> Thomas
>
> Walter schrieb am Fri, 19 Nov 2004 14:18:28 -0800:
> > I don't believe this is a bug. It is analogous to:
> >
> >     byte b = 1000;
> >
> > which is allowed as well.
> >
> >> Thomas Kuehne schrieb am Wed, 17 Nov 2004 09:45:12 +0100:
> >> > Constant assignments to wchar allow values greater than wchar.max. http://svn.kuehne.cn/dstress/nocompile/wchar_04.d
> >> >
> >> > Constant assignments to wchar allow values smaller than wchar.min. http://svn.kuehne.cn/dstress/nocompile/wchar_06.d
> >>
> >> dchar has the same problems.
> >>
> >> http://svn.kuehne.cn/dstress/nocompile/dchar_04.d http://svn.kuehne.cn/dstress/nocompile/dchar_06.d