[Spice-devel] [PATCH 1/2] define SPICE_CONSTRUCTOR_FUNC and SPICE_DESTRUCTOR_FUNC macros

[Victor Toso]

Hi,

On Wed, Mar 16, 2016 at 11:52:55AM -0400, Frediano Ziglio wrote:
> > 
> > Hi,
> > 
> > On Mon, Feb 29, 2016 at 10:36:53AM +0000, Frediano Ziglio wrote:
> > > Allow to define functions executed at program/shared object initialization
> > > or close.
> > > 
> > > Signed-off-by: Frediano Ziglio <fziglio at redhat.com>
> > > ---
> > >  common/macros.h | 22 ++++++++++++++++++++++
> > >  1 file changed, 22 insertions(+)
> > > 
> > > diff --git a/common/macros.h b/common/macros.h
> > > index 47289be..fe36929 100644
> > > --- a/common/macros.h
> > > +++ b/common/macros.h
> > > @@ -29,5 +29,27 @@
> > >  #define SPICE_ATTR_PRINTF
> > >  #endif /* __GNUC__ */
> > >  
> > > +#ifdef __GNUC__
> > > +#define SPICE_CONSTRUCTOR_FUNC(func_name) \
> > > +    static void __attribute__((constructor)) func_name(void)
> > > +#define SPICE_DESTRUCTOR_FUNC(func_name) \
> > > +    static void __attribute__((destructor)) func_name(void)
> > > +#elif defined(_MSC_VER)
> > > +#define SPICE_CONSTRUCTOR_FUNC(func_name) \
> > > +    static void func_name(void); \
> > > +    static int func_name ## _wrapper(void) { func_name(); return 0; } \
> > > +    __pragma(section(".CRT$XCU",read)) \
> > > +    __declspec(allocate(".CRT$XCU")) static int (* _array ##
> > > func_name)(void) = func_name ## _wrapper; \
> > > +    static void func_name(void)
> > > +#define SPICE_DESTRUCTOR_FUNC(func_name) \
> > > +    static void func_name(void); \
> > > +    static int func_name ## _wrapper(void) { func_name(); return 0; } \
> > > +    __pragma(section(".CRT$XPU",read)) \
> > > +    __declspec(allocate(".CRT$XPU")) static int (* _array ##
> > > func_name)(void) = func_name ## _wrapper; \
> > > +    static void func_name(void)
> > > +#else
> > > +#error Please implement SPICE_CONSTRUCTOR_FUNC and SPICE_DESTRUCTOR_FUNC
> > > for this compiler
> > > +#endif
> > > +
> > >  
> > >  #endif /* __MACROS_H */
> > > --
> > > 2.5.0
> > 
> > Testes, works without issues.
> > 
> > I'm wondering why not to include the whole header from glib [0] and
> > maybe file a bug to ask them to export in the future.
> > 
> > [0] https://git.gnome.org/browse/glib/tree/glib/gconstructor.h
> > 
> > Then, in case G_HAS_CONSTRUCTORS is not defined we can include different
> > arch constructors. What do you think?
> > 
> > PS: while looking for an open bug:
> > aix: https://bugzilla.gnome.org/show_bug.cgi?id=763560
> > hp-ux/ia: https://bugzilla.gnome.org/show_bug.cgi?id=763466
> > 
> > cheers,
> >   toso
> > 
> 
> Do we want to support HP-UX or AIX? I did work on HP-UX and from what I
> can see not even HP is supporting that much this OS.

Not really. I just thought it was interesting to share :)

> Personally I don't like Gnome implementation, particularly the atexit call
> on Windows.
> I don't understand the g_slist_find call either.
>
> Frediano

Your version is fine. I doubt that there is any benefit to use the glib
version at the moment. My point was more about the future. If glib exposes
those macros we could avoid duplicating it here.

Acked-by: Victor Toso <victortoso at redhat.com>