dmd2.025, Windows

Some standard functions receive 'string' arguments.
This is not a bug, but not good specs.

void[] read( in string name );
void[] write( in string name, const void[] buffer );
...

They should be

void[] read( in char[] name );
void[] write( in char[] name, const void[] buffer );
...

Former cannot receive char[]-arguments, and latter can. This change has backward compatibility.

CLXX wrote:
> dmd2.025, Windows
> 
> Some standard functions receive 'string' arguments.
> This is not a bug, but not good specs.
> 
> void[] read( in string name );
> void[] write( in string name, const void[] buffer );
> ....
> 
> They should be
> 
> void[] read( in char[] name );
> void[] write( in char[] name, const void[] buffer );
> ....
> 
> Former cannot receive char[]-arguments, and latter can. This change has backward compatibility.

I... wait, what?

*blinks*

That change will just mean that the functions can't take string arguments, and if you WANT to pass a string, you have to heap-allocate a new one.

Why don't you just pass a string?

  -- Daniel

On 28.02.2009 15:47, Daniel Keep wrote:
>
> CLXX wrote:
>>  dmd2.025, Windows
>>
>>  Some standard functions receive 'string' arguments.
>>  This is not a bug, but not good specs.
>>
>>  void[] read( in string name );
>>  void[] write( in string name, const void[] buffer );
>>  ....
>>
>>  They should be
>>
>>  void[] read( in char[] name );
>>  void[] write( in char[] name, const void[] buffer );
>>  ....
>>
>>  Former cannot receive char[]-arguments, and latter can.
>>  This change has backward compatibility.
>
> I... wait, what?
>
> *blinks*
>
> That change will just mean that the functions can't take string
> arguments, and if you WANT to pass a string, you have to heap-allocate a
> new one.
>
> Why don't you just pass a string?

Hehe, someone had to little coffee today.  'in' implies const, and invariant is implicitly convertible to const. So Mr. 'CLXX' is right. :P

CLXX wrote:
> dmd2.025, Windows
> 
> Some standard functions receive 'string' arguments.
> This is not a bug, but not good specs.
> 
> void[] read( in string name );
> void[] write( in string name, const void[] buffer );
> ...
> 
> They should be
> 
> void[] read( in char[] name );
> void[] write( in char[] name, const void[] buffer );
> ...
> 
> Former cannot receive char[]-arguments, and latter can.
> This change has backward compatibility.

Yah, many of them are already changed in my tree. Phobos will have to sit this release out, though.

Andrei

torhu wrote:
> On 28.02.2009 15:47, Daniel Keep wrote:
>>
>> CLXX wrote:
>>>  dmd2.025, Windows
>>>
>>>  Some standard functions receive 'string' arguments.
>>>  This is not a bug, but not good specs.
>>>
>>>  void[] read( in string name );
>>>  void[] write( in string name, const void[] buffer );
>>>  ....
>>>
>>>  They should be
>>>
>>>  void[] read( in char[] name );
>>>  void[] write( in char[] name, const void[] buffer );
>>>  ....
>>>
>>>  Former cannot receive char[]-arguments, and latter can.
>>>  This change has backward compatibility.
>>
>> I... wait, what?
>>
>> *blinks*
>>
>> That change will just mean that the functions can't take string arguments, and if you WANT to pass a string, you have to heap-allocate a new one.
>>
>> Why don't you just pass a string?
> 
> Hehe, someone had to little coffee today.  'in' implies const, and invariant is implicitly convertible to const. So Mr. 'CLXX' is right. :P

*stares at it*

Well, that's what I get for posting at 2 AM.

  -- Daniel