[vlc-devel] [RFCv2] Thread cancellation on WinCE

[Pierre Ynard]

> > +#  define VLC_CANCEL_INIT { NULL, true, false, NULL }
> 
> NULL is the error case for HANDLE?

Yes.

> > +static DWORD vlc_cancelable_wait (DWORD count, const HANDLE *handles,
> > +                                  DWORD delay)
> > +{
> > +    vlc_cancel_t *nfo = vlc_threadvar_get (&cancel_key);
> > +    if (nfo == NULL)
> > +    {
> > +        /* Main thread - cannot be cancelled anyway */
> > +        return WaitForMultipleObjects (count, handles, FALSE, delay);
> > +    }
> 
> If you collapsed vlc_cancelable_wait into WaitForMultipleObjectsEx,
> you could factor that, and the non alertable case, right?

Yes, I could factor the call to WaitForMultipleObjects with a goto.
However, it seems cleaner this way, and I wasn't sure whether these
replacement functions should be exported or something, and if so, how.

> > +    HANDLE new_handles[count + 1];
> > +    if (count > 0)
> 
> This check is not needed.

My memcpy manpage doesn't document calling it with a 0 size and a NULL
source pointer :/ But if you're sure, right, I'll remove it.

> > +    /* We do not support the bWaitAll case */
> > +    if (bAlertable && ! bWaitAll)
> 
> Should probably assert(!(bAlertable && bWaitAll)).

Okay. Thanks!

-- 
Pierre Ynard
"Une âme dans un corps, c'est comme un dessin sur une feuille de papier."