Re: [AD] mzscheme bindings in svn

[Jon Rafkind <workmin@xxxxxxxxxx>]

Jon Rafkind wrote:
> Peter Wang wrote:
>   
> > On 2007-04-28, Jon Rafkind <workmin@xxxxxxxxxx> wrote:
> >   
> >     
> > > > > >      
> > > > > >           
> > > > > >             
> > > > > That would be fine (but call it __al_xwin_dont_install_signal_handlers
> > > > > or something).  A dirtier solution would be to save the MzScheme signal
> > > > > handlers before calling allegro_init(), then restore them immediately
> > > > > afterwards.  This would work with current and older versions of Allegro.
> > > > >
> > > > >  
> > > > >    
> > > > >         
> > > > >           
> > > > I think messing with the invariants about the garbage collector is 
> > > > tricky business and might lead to very strange results.
> > > >       
> > > >         
> > I don't think it's that tricky.  During the short time that Allegro is
> > initialising we wouldn't be writing to any pages which MzScheme has
> > marked as read-only (I'm assuming that's how it works) and we'd restore
> > it straight after.  Disabling *all* of Allegro's emergency signal
> > handlers seems overkill.
> >
> >   
> >     
> I was thinking a problem might occur when allegro is initialized via a 
> scheme thread so other scheme things are going on and a write-barrier is 
> hit while Allegro is being initialized. I'm sure its incredibly unlikely 
> but if the situation ever arose the programmer would go bonkers trying 
> to debug it.
>   
I talked to a developer of mzscheme who said it should be ok to 
save/restore the signals used by mzscheme during Allegro initilization 
as well as blocking/unblocking SIGINT, SIGCHLD, and SIGPROF. I can 
probably do this without any modification to Allegro.