Re: gboolean definition and size

[Owen Taylor <otaylor redhat com>]

Dan Nicolaescu <dann godzilla ICS UCI EDU> writes:

> Hi!
> 
> gboolean is currently defined as an int. 
> 
> The C99 standard contains now a standard boolean size, and it's _not_
> int size. 
> 
> gcc-3.x defines a magic type _Bool that is used to define the bool
> type. It's size is 1 byte.
> 
> Is there any reason for gboolean to have a different size? 
> 
> IMHO it should be defined in such a way that it makes interoperability
> with C99 code easy. 
> 
> Probably gboolean should be defined as _Bool when using gcc-3.x 
> Another reason is that in the future compilers might have
> special optimizations for bool types... 
> 
> As changing the definition of gboolean is a binary incompatible
> change, it's probably a good idea to do it before releasing gtk-2.0 ...
> 
> 
> I tried to search the mailing list archive, and didn't find any
> discussion of this, I hope it has not been beaten to death in the
> past... 

Yes, it was beaten to death in the past. Unfortunately, the archive
search is not working currently.

The basic problem with the bool definition without compiler support
for a real bool type (no stdbool.h doesn't give that) is things like:

 gboolean foo = flags & MY_FLAG;

And other such less obviously wrong expressions. Making it a guint
to help with bitfield expressions is more of a possible compatible
change, but I sort of like keeping it as simple as possible.

Regards,
                                        Owen