Re: More msvcrt scanf tests

[Alexandre Julliard]

Uwe Bonnes  writes:
> - todo_wine 	ok(strstr(buffer,"e+008") != 0,"Sprintf different "%s"\n",buffer);
> - sprintf(buffer,I64x,(ULONGLONG)0xffffffffffffffff);

Long long constants are not portable, you need to compute them from
long constants.

More msvcrt scanf tests

[Boaz Harrosh]

Alexandre Julliard wrote:
> Uwe Bonnes  writes:
> > - todo_wine 	ok(strstr(buffer,"e+008") != 0,"Sprintf different "%s"\n",buffer);
> > - sprintf(buffer,I64x,(ULONGLONG)0xffffffffffffffff);
>
> Long long constants are not portable, you need to compute them from
> long constants.

The official way to do it in GCC (any gcc) is 0x1234567812345678LL Just 
like we used to do 0x12345678L in 16 bit
In VC++ it is 0x1234567812345678i64. Intel can do LL as well. Not sure 
about borland or watcom. What do OS X use?

I guess you can do:
#ifdef _GCC_
#define i64 LL
#endif

or the opposite way (#define LL i64 )

More msvcrt scanf tests

[Dmitry Timoshkov]

"Boaz Harrosh"  wrote:
> > Long long constants are not portable, you need to compute them from
> > long constants.
>
> The official way to do it in GCC (any gcc) is 0x1234567812345678LL Just 
> like we used to do 0x12345678L in 16 bit
> In VC++ it is 0x1234567812345678i64. Intel can do LL as well. Not sure 
> about borland or watcom. What do OS X use?
>
> I guess you can do:
> #ifdef _GCC_
> #define i64 LL
> #endif
>
> or the opposite way (#define LL i64 )

We have to use an approach implemented in include/wine/debug.h,
wine_dbgstr_longlong(). Any modifier for LL in the printf format
string is not portable.

More msvcrt scanf tests

[Uwe Bonnes]

> > > > > "Dmitry" == Dmitry Timoshkov  writes:

Dmitry> "Boaz Harrosh"  wrote:
    >> >Long long constants are not portable, you need to compute them from
    >> >long constants.
    >> >
    >> >  
    >> >
    >> 
    >> The official way to do it in GCC (any gcc) is 0x1234567812345678LL
    >> Just like we used to do 0x12345678L in 16 bit In VC++ it is
    >> 0x1234567812345678i64. Intel can do LL as well. Not sure about
    >> borland or watcom. What do OS X use?
    >> 
    >> I guess you can do: #ifdef _GCC_ #define i64 LL #endif
    >> 
    >> or the opposite way (#define LL i64 )

Dmitry> We have to use an approach implemented in include/wine/debug.h,
    Dmitry> wine_dbgstr_longlong(). Any modifier for LL in the printf format
    Dmitry> string is not portable.

But wine_dbgstr_longlong doesn't take a field length, doesn't know how to
print decimal or octal and so on...

More msvcrt scanf tests

[Uwe Bonnes]

> > > > > "Alexandre" == Alexandre Julliard  writes:

Alexandre> Uwe Bonnes  writes:
    >> + todo_wine ok(strstr(buffer,"e+008") != 0,"Sprintf different
    >> "%s"\n",buffer); +
    >> sprintf(buffer,I64x,(ULONGLONG)0xffffffffffffffff);

Alexandre> Long long constants are not portable, you need to compute
    Alexandre> them from long constants.

The wine code is sprankled with LONGLONG constants, e.g.:

dlls/oleaut32/variant.c:        LONGLONG     lVal = -1;

Where is the code in this test suite supposed to differ  from the code in
the rest of wine? It's supposed to be compiled against the wine windows
headers. 
#include "wine/test.h"
-> #include <windef.h>
-> #include <winnt.h>
and there
#ifndef _ULONGLONG_
#define _ULONGLONG_
typedef signed __int64   LONGLONG,   *PLONGLONG;
typedef unsigned __int64 ULONGLONG,  *PULONGLONG;
#endif

Bye

More msvcrt scanf tests

[Robert Shearman]

Uwe Bonnes wrote:
> Alexandre> Long long constants are not portable, you need to compute
> Alexandre> them from long constants.
>
> The wine code is sprankled with LONGLONG constants, e.g.:
>
> dlls/oleaut32/variant.c:        LONGLONG     lVal = -1;

He means constants like 0x12345689LL. You have to do this instead:
(LONGLONG)0x1 << 32 | (LONGLONG)0x23456789.

Rob

More msvcrt scanf tests

[Uwe Bonnes]

> > > > > "Robert" == Robert Shearman  writes:

Robert> Uwe Bonnes wrote:
    >>
    Alexandre> Long long constants are not portable, you need to compute
    Alexandre> them from long constants.
    >>  The wine code is sprankled with LONGLONG constants, e.g.:
    >> 
    >> dlls/oleaut32/variant.c: LONGLONG lVal = -1;
    >>

Robert> He means constants like 0x12345689LL. You have to do this
    Robert> instead: (LONGLONG)0x1 << 32 | (LONGLONG)0x23456789.

Okay!.

Any hint how to do for this for decimal numbers?

More msvcrt scanf tests

[Boaz Harrosh]

Kills me why we have to work extra time in the position of a compiler.
GCC does that fine a macro can fix the msvc way. What are those 
supported compilers that choke?

Uwe Bonnes wrote:
> > > > > > "Robert" == Robert Shearman  writes:
>
> Robert> He means constants like 0x12345689LL. You have to do this
>    Robert> instead: (LONGLONG)0x1 << 32 | (LONGLONG)0x23456789.
>
> Okay!.
>
> Any hint how to do for this for decimal numbers?

More msvcrt scanf tests

[Robert Shearman]

Boaz Harrosh wrote:
> Kills me why we have to work extra time in the position of a compiler.
> GCC does that fine a macro can fix the msvc way. What are those
> supported compilers that choke?

I don't know, but any C99-compatible compiler should support the LL suffix
and the ll width format specifier (in the C library). MSVC doesn't support
these and AFAIK never will.

Rob

More msvcrt scanf tests

[Boaz Harrosh]

Robert Shearman wrote:
> I don't know, but any C99-compatible compiler should support the LL suffix
> and the ll width format specifier (in the C library). MSVC doesn't support
> these and AFAIK never will.
>
> Rob

Never say never. It does if you:
#define LL i64
#define ll i64
So what is exactly the problem ?