Re: [Qemu-devel] [PATCH 1/4] memory: make memory API parsable by gtkdoc-scan

[Stefan Weil <sw@weilnetz.de>]

Am 14.12.2011 17:34, schrieb malc:
> On Wed, 14 Dec 2011, Anthony Liguori wrote:
>
>> GTK/glib uses a convenient of:
>>
>> typedef struct _CamelCase CamelCase;
>>
>> The reason that they use a separate struct name is that in C++, the 
>> struct
>> namespace not a separate namespace from the type namespace. This is 
>> actually a
>> reasonable policy for QEMU to adopt as we eventually start exporting 
>> C libraries
>> that may be consumed by C++ programs.
>>
>> I think the use of _ does not violate the C specification as the struct
>> namespace is not the same as the type namespace which is what the C 
>> spec refers
>> to if I understand it correctly.
>
> It does violate the standard _ followed by upper case letter is reserved
> in all contexts.

sCamelCase instead of _CamelCase seems to work, too.

I just finished a first test with gtk-doc and had no problems.

So it's possible to support gtk-doc (which is a good thing) _and_
keep the standard (which is very important, too).

The new rule for structure declarations in QEMU could be like this:

/* forward declaration */
struct sCamelCase;

/* struct definition */
typedef struct sCamelCase {
   /* values follow here ... */
} CamelCase;

Structures which don't need a forward declaration can use
simplified definitions:

typedef struct {
   /* values follow here ... */
} CamelCaseWithoutForwardDeclaration;

Regards,

Stefan Weil