Merge lp://qastaging/~compiz-team/compiz/compiz.ccs_general_cleanup into lp://qastaging/compiz/0.9.8

> compizconfig/gconf/src/gconf.c

Why change array to be "const char**" ? Also why cast it in free?

> compizconfig/gsettings/src/gsettings.c

Casting "const gchar**" to "char**"? Surely that should be "gchar**" if anything.

Why add the "const" if it is only causing pain?

I personally dislike TRUE being defined to ~0, but that's just me (other places define it to be 1).

Shouldn't !FALSE == TRUE? Pretty sure !0 is defined to be 1.

> compizconfig/tests/compizconfig_ccs_mocked_allocator.h

Header files should stand alone. here you are using stuff from <gmock/gmock.h>

free(void*) and g_free(void*) means those functions can accept any pointer type EXCEPT a pointer to const data.

The reason why they don't accept "const void *" is to honour the const properly. Freeing memory means the system is taking over it again and hence will write to it. free() would give the wrong message if it accepted "const void *" and hence any type of pointer.

The two solutions are:
  1. Dynamically allocated memory should never be assigned to const pointers; or
  2. Cast all pointers to non-const on free: free((void*)p)

I don't see either option being much better than the other. Although #1 (removing the const) makes for slightly simpler code.

> > compizconfig/gconf/src/gconf.c
>
> Why change array to be "const char**" ? Also why cast it in free?
>
> > compizconfig/gsettings/src/gsettings.c
>
> Casting "const gchar**" to "char**"? Surely that should be "gchar**" if
> anything.
>
> Why add the "const" if it is only causing pain?

Yep, you're completely right, I am an idiot. ccsSetValueListFromStringArray takes const char ** now, and I just changed the declaration to get around the error. A better way of doing this:

gchar **array = calloc (1, sizeof (gchar *));

ccsSetValueListFromStringArray ((const gchar **) array);

g_strfreev (array);
>
> I personally dislike TRUE being defined to ~0, but that's just me (other
> places define it to be 1).
>
> Shouldn't !FALSE == TRUE? Pretty sure !0 is defined to be 1.

Yep it does suck, I don't know what kind of subtle bugs we might encounter though if we change the definition right now, so I'd prefer to leave it as is for the time being.

>
> > compizconfig/tests/compizconfig_ccs_mocked_allocator.h
>
> Header files should stand alone. here you are using stuff from
> <gmock/gmock.h>

Ack, fixed.

Tested downstream in the proposal for lp:~compiz-team/compiz/compiz.ccs_gnome_integration_gsettings_writer

No commit message specified.