Re: ntdll / kernel32 #39-2

[Dmitry Timoshkov]

"Eric Pouech"  wrote:
> +DWORD WINAPI GetFullPathNameW( LPCWSTR name, DWORD len, LPWSTR buffer,
> -
>                           LPWSTR *lastpart )
>
> +{
> - LPWSTR      dst = HeapAlloc(GetProcessHeap(), 0, len * sizeof(buffer));
> - DWORD       ret;
> -
> - if (!dst) return 0;
> - ret = RtlGetFullPathName_U(name, len * sizeof(WCHAR), dst, lastpart) / sizeof(WCHAR);

I don't quite understand the above bit of code:

1. HeapAlloc(GetProcessHeap(), 0, len * sizeof(buffer)) when buffer is
a pointer
2. Why you don't pass buffer directly to RtlGetFullPathName_U without 
allocating an intermediate storage.

ntdll / kernel32 #39-2

[Eric Pouech]

Dmitry Timoshkov a e'crit :
> "Eric Pouech"  wrote:
> > +DWORD WINAPI GetFullPathNameW( LPCWSTR name, DWORD len, LPWSTR buffer,
> > -
> >                           LPWSTR *lastpart )
> >
> > +{
> > - LPWSTR      dst = HeapAlloc(GetProcessHeap(), 0, len * sizeof(buffer));
> > - DWORD       ret;
> > -
> > - if (!dst) return 0;
> > - ret = RtlGetFullPathName_U(name, len * sizeof(WCHAR), dst, lastpart) / sizeof(WCHAR);
>
> I don't quite understand the above bit of code:
> - HeapAlloc(GetProcessHeap(), 0, len * sizeof(buffer)) when buffer is
>
> a pointer

good catch
> - Why you don't pass buffer directly to RtlGetFullPathName_U without
>
> allocating an intermediate storage.

because RtlGetFullPathName_U is buggy when name = buffer (it would be 
better anyway to fix RtlGetFullPathName_U to support that case).

A+