Question regarding declaration of RTL_DRIVE_LETTER_CURDIR in include\ndk\rtltypes.h

[Sven Barth]

Hello together!

I'm currently poking around in the NDK headers very much and I noted a 
small inconsistency in the declaration of one type.

The declaration is the following in include\ndk\rtltypes.h:

==== source begin ====

typedef struct RTL_DRIVE_LETTER_CURDIR
{
     USHORT Flags;
     USHORT Length;
     ULONG TimeStamp;
     UNICODE_STRING DosPath;
} RTL_DRIVE_LETTER_CURDIR, *PRTL_DRIVE_LETTER_CURDIR;

==== source end ====

Shouldn't it be "_RTL_DRIVE_LETTER_CURDIR" instead of 
"RTL_DRIVE_LETTER_CURDIR"?

I don't know what influences this missing underscore has on compilation 
and such, cause I can read C without big problems and write a bit as 
well, but I don't understand every single bit a C compiler does ^^

Regards,
Sven

[Colin Finck]

Sven Barth [e-mail] wrote:
> Shouldn't it be "_RTL_DRIVE_LETTER_CURDIR" instead of
> "RTL_DRIVE_LETTER_CURDIR"?

It doesn't matter if you just define a variable using this structure 
like "RTL_DRIVE_LETTER_CURDIR MyVariable;".

It just matters if you somewhere declare it as "struct 
RTL_DRIVE_LETTER_CURDIR MyVariable;".
If this isn't done anywhere, we could even create an anonymous structure 
(Microsoft extension to the C language) by defining the structure as 
"typedef struct {...} RTL_DRIVE_LETTER_CURDIR, *PRTL_DRIVE_LETTER_CURDIR;"

But as all headers out there seem to define it as "typedef struct 
RTL_DRIVE_LETTER_CURDIR", we should do as well to ensure compatibility 
with all applications using the structure :-)

Cheers,

Colin

[Sven Barth]

On 13.11.2010 19:22, Colin Finck wrote:
> Sven Barth [e-mail] wrote:
> > Shouldn't it be "_RTL_DRIVE_LETTER_CURDIR" instead of
> > "RTL_DRIVE_LETTER_CURDIR"?
>
> It doesn't matter if you just define a variable using this structure
> like "RTL_DRIVE_LETTER_CURDIR MyVariable;".
>
> It just matters if you somewhere declare it as "struct
> RTL_DRIVE_LETTER_CURDIR MyVariable;".
> If this isn't done anywhere, we could even create an anonymous structure
> (Microsoft extension to the C language) by defining the structure as
> "typedef struct {...} RTL_DRIVE_LETTER_CURDIR, *PRTL_DRIVE_LETTER_CURDIR;"
>
> But as all headers out there seem to define it as "typedef struct
> RTL_DRIVE_LETTER_CURDIR", we should do as well to ensure compatibility
> with all applications using the structure :-)

It just caught my eye cause every other struct definition in the ReactOS 
headers (that I've seen so far) is named with a leading underscore...

But thank you for your explanation why this isn't a problem.

Regarding the anonymous structure: wouldn't this break compatibility 
with gcc? (I don't know what features gcc supports ^^)

Regards,
Sven

[Colin Finck]

Sven Barth [e-mail] wrote:
> Regarding the anonymous structure: wouldn't this break compatibility
> with gcc? (I don't know what features gcc supports ^^)

Anonymous structures are fully supported by GCC when using the flag 
-fms-extensions. We have it enabled throughout the whole codebase.

Colin

[Javier Agustìn Fernàndez Arroyo]

...and what about cmake?
´cos we are moving from gcc to cmake....

On Sat, Nov 13, 2010 at 7:44 PM, Colin Finck [e-mail] wrote:
> Sven Barth [e-mail] wrote:
> > Regarding the anonymous structure: wouldn't this break compatibility
> > with gcc? (I don't know what features gcc supports ^^)
>
> Anonymous structures are fully supported by GCC when using the flag
> -fms-extensions. We have it enabled throughout the whole codebase.
>
> Colin
>
> Ros-dev mailing list
> [e-mail]
> http://www.reactos.org/mailman/listinfo/ros-dev

[Pierre Schweitzer]

Hi,

GCC is a compiler. CMAKE is a build system we use with... GCC (for the moment). So no point in the question.

Best regards,
P. Schweitzer
> ...and what about cmake?
> ´cos we are moving from gcc to cmake....
>
> On Sat, Nov 13, 2010 at 7:44 PM, Colin Finck [e-mail] wrote:
> > Sven Barth [e-mail] wrote:
> > > Regarding the anonymous structure: wouldn't this break compatibility
> > > with gcc? (I don't know what features gcc supports ^^)
> >
> > Anonymous structures are fully supported by GCC when using the flag
> > -fms-extensions. We have it enabled throughout the whole codebase.
> >
> > Colin
> >
> > Ros-dev mailing list
> > [e-mail]
> > http://www.reactos.org/mailman/listinfo/ros-dev
>
> Ros-dev mailing list
> [e-mail]
> http://www.reactos.org/mailman/listinfo/ros-dev

[Javier Agustìn Fernàndez Arroyo]

ah.. ok sorry

On Sat, Nov 13, 2010 at 8:45 PM, Pierre Schweitzer <
[e-mail]> wrote:
> Hi,
>
> GCC is a compiler. CMAKE is a build system we use with... GCC (for the
> moment). So no point in the question.
>
> Best regards,
> P. Schweitzer
> > ...and what about cmake?
> > ´cos we are moving from gcc to cmake....
> >
> > On Sat, Nov 13, 2010 at 7:44 PM, Colin Finck [e-mail] wrote:
> > > Sven Barth [e-mail] wrote:
> > > > Regarding the anonymous structure: wouldn't this break compatibility
> > > > with gcc? (I don't know what features gcc supports ^^)
> > >
> > > Anonymous structures are fully supported by GCC when using the flag
> > > -fms-extensions. We have it enabled throughout the whole codebase.
> > >
> > > Colin
> > >
> > > Ros-dev mailing list
> > > [e-mail]
> > > http://www.reactos.org/mailman/listinfo/ros-dev
> >
> > Ros-dev mailing list
> > [e-mail]
> > http://www.reactos.org/mailman/listinfo/ros-dev
>
> Ros-dev mailing list
> [e-mail]
> http://www.reactos.org/mailman/listinfo/ros-dev

[Sven Barth]

On 13.11.2010 19:44, Colin Finck wrote:
> Sven Barth [e-mail] wrote:
> > Regarding the anonymous structure: wouldn't this break compatibility
> > with gcc? (I don't know what features gcc supports ^^)
>
> Anonymous structures are fully supported by GCC when using the flag
> -fms-extensions. We have it enabled throughout the whole codebase.

Ok, thanks again :)

Regards,
Sven

[Ged Murphy]

Colin Finck wrote:
> Sven Barth [e-mail] wrote:
> > Shouldn't it be "_RTL_DRIVE_LETTER_CURDIR" instead of
> > "RTL_DRIVE_LETTER_CURDIR"?
>
> It doesn't matter if you just define a variable using this structure 
> like "RTL_DRIVE_LETTER_CURDIR MyVariable;".

It matters from a debugging perspective. The symbols store the struct name,
not the typedef.
Anyone expecting to dump data based on the data type with an underscore
would get an error on reactos.

I've just had a look at win7:

0: kd> dt ntdll!RTL_DRIVE_LETTER_CURDIR
Symbol ntdll!RTL_DRIVE_LETTER_CURDIR not found.
0: kd> dt ntdll!_RTL_DRIVE_LETTER_CURDIR
   +0x000 Flags            : Uint2B
   +0x002 Length           : Uint2B
   +0x004 TimeStamp        : Uint4B
   +0x008 DosPath          : _STRING

So we actually break compatibility with Windows by not copying struct names
exactly.
This is indeed a typo that needs fixing.

Ged.