Re: [msvcrt] Implement %I types for printf

[Alexandre Julliard]

Jesse Allen  writes:
> @@ -191,11 +191,14 @@
>      } buf;
>  } pf_output;
>
> +#define INTEGERSIZE_LONG 1
> +#define INTEGERSIZE_LONGLONG 2
> -
>
> typedef struct pf_flags_t
>  {
>      char Sign, LeftAlign, Alternate, PadZero;
>      char FieldLength, Precision;
> - char IntegerLength, IntegerDouble;
> - char IntegerLength, IntegerDouble, IntegerSize;

The IntegerDouble field was supposed to be used for that already, no
need to add another one.
> @@ -384,6 +387,16 @@
>          sprintf(p, ".%d", flags->Precision);
>          p += strlen(p);
>      }
> - if( flags->IntegerLength == 'I' )
> - {
> -
>    sprintf(p, "l");
> -
>    p++;
> - }
> - if( flags->IntegerSize == INTEGERSIZE_LONGLONG )
> - {
> -
>    sprintf(p, "l");
> -
>    p++;
> - }

This assumes that the system printf supports %ll formats, that's not
portable.

[msvcrt] Implement %I types for printf

[Jesse Allen]

On 06 Jun 2005 10:44:22 +0200, Alexandre Julliard  wrote:
> Jesse Allen  writes:
> > @@ -191,11 +191,14 @@
> >      } buf;
> >  } pf_output;
> >
> > +#define INTEGERSIZE_LONG 1
> > +#define INTEGERSIZE_LONGLONG 2
> > -
> >
> > typedef struct pf_flags_t
> >  {
> >      char Sign, LeftAlign, Alternate, PadZero;
> >      char FieldLength, Precision;
> > - char IntegerLength, IntegerDouble;
> > - char IntegerLength, IntegerDouble, IntegerSize;
>
> The IntegerDouble field was supposed to be used for that already, no
> need to add another one.

OK I changed it back to that.
> > @@ -384,6 +387,16 @@
> >          sprintf(p, ".%d", flags->Precision);
> >          p += strlen(p);
> >      }
> > - if( flags->IntegerLength == 'I' )
> > - {
> > -
> >    sprintf(p, "l");
> > -
> >    p++;
> > - }
> > - if( flags->IntegerSize == INTEGERSIZE_LONGLONG )
> > - {
> > -
> >    sprintf(p, "l");
> > -
> >    p++;
> > - }
>
> This assumes that the system printf supports %ll formats, that's not
> portable.

Well it's rebuilding the format string, I thought for libc.  Is there
another %ll-like specifier?  %I64 certainly is not. =)

Jesse

[msvcrt] Implement %I types for printf

[Alexandre Julliard]

Jesse Allen  writes:
> Well it's rebuilding the format string, I thought for libc.  Is there
> another %ll-like specifier?  %I64 certainly is not. =)

No, I'm afraid there is no standard way of doing that. You won't be
able to simply forward this one to libc, you'll need to do at least
part of the formatting by hand.

[wine] Re: [msvcrt] Implement %I types for printf

[David Lee Lambert]

On Mon, Jun 06, 2005 at 05:32:52PM +0200, Alexandre Julliard wrote:
> Jesse Allen  writes:
> > Well it's rebuilding the format string, I thought for libc.  Is there
> > another %ll-like specifier?  %I64 certainly is not. =)
>
> No, I'm afraid there is no standard way of doing that. You won't be
> able to simply forward this one to libc, you'll need to do at least
> part of the formatting by hand.

The ll flag is specified by POSIX (2001, "System Interfaces", p. 404) and
is described by the Linux (dated 2000-10-16) and Solaris (SunOS 5.9)  
manpages.  I'd certainly call that a "standard way", even if it's not
supported everywhere.

Apparently 4.4 BSD and older versions of Linux libc used a "%q" flag for 
the same thing.  Would it be OK for Jesse to use a "configure" check to 
see which is supported, at compile-time?

[wine] Re: [msvcrt] Implement %I types for printf

[Jesse Allen]

On 6/7/05, David Lee Lambert  wrote:
> On Mon, Jun 06, 2005 at 05:32:52PM +0200, Alexandre Julliard wrote:
> > No, I'm afraid there is no standard way of doing that. You won't be
> > able to simply forward this one to libc, you'll need to do at least
> > part of the formatting by hand.
>
> The ll flag is specified by POSIX (2001, "System Interfaces", p. 404) and
> is described by the Linux (dated 2000-10-16) and Solaris (SunOS 5.9)
> manpages.  I'd certainly call that a "standard way", even if it's not
> supported everywhere.
>
> Apparently 4.4 BSD and older versions of Linux libc used a "%q" flag for
> the same thing.  Would it be OK for Jesse to use a "configure" check to
> see which is supported, at compile-time?

Doing a configure check will help as a temporary option as there is no
support for %I64 as there is.  If %ll or %q are not supported, then it
can be disabled and it will work the same as it is now.

[msvcrt] Implement %I types for printf

[Jesse Allen]

On 06 Jun 2005 10:44:22 +0200, Alexandre Julliard  wrote:
> > - char IntegerLength, IntegerDouble, IntegerSize;
>
> The IntegerDouble field was supposed to be used for that already, no
> need to add another one.

The reason why I added IntegerSize is some how we need to remember if
we have been requested I32, because if we don't, then it won't work
right on 64-bit systems?  Therefore, instead of IntegerDouble,
IntegerSize.  It was untested anyway so I changed it back.