[PATCH] winmm: Don't mask out SND_ALIAS_ID or SND_FILENAME in sndPlaySound

[Andrew Eikum]

On Wed, Dec 05, 2012 at 11:06:48PM +0800, Dmitry Timoshkov wrote:
> Andrew Eikum <aeikum at codeweavers.com> wrote:
> 
> > > > +    br = sndPlaySoundA((LPCSTR)SND_ALIAS_SYSTEMASTERISK, SND_ALIAS_ID|SND_SYNC);
> > > > +    ok(br == TRUE || br == FALSE, "sndPlaySound gave strange return: %u\n", br);
> > > 
> > > This kind of test is broken.
> > > 
> > 
> > Care to elaborate? Being this terse helps no one.
> 
> What is this code testing the return value for? Why are there all these casts?
> 

The call is only supposed to return TRUE or FALSE, so I guess it's
checking that the return value isn't 3 or something. Not very useful,
sure, but it looked odd to just have a series of sndPlaySound calls in
a row without ok() calls. The real test is to make sure sndPlaySound
doesn't crash, as it does without my patch.

The casts are to avoid compilation warnings. sndPlaySound takes a
string as its first parameter, but it also accepts symbols like
SND_ALIAS_SYSTEMASTERISK, which has the value 0x2A53. A cast is needed
to avoid the warning about converting an integer to a pointer.

Andrew