Add tests for these protocol sequences and for ncacn_ip_tcp.

services: Always initialise the out buffer in svcctl_QueryServiceStatusEx and svcctl_QueryServiceConfig2W.

services: svcctl_GetServiceDisplayNameW and svcctl_GetServiceKeyNameW should have string attribute on output buffer.

They also should not have two parameters for specifying the size of the buffer.

The buffer size should also not include in the nul-terminating character.

The ref count will be released by SC_RPC_HANDLE_destroy when the
returned handle is destroyed.

Change svcctl_OpenServiceW to use InterlockedIncrement to increment
the ref count since the service is in the database and therefore could
be shared and SC_RPC_HANDLE_destroy doesn't take the scmdatabase lock
when decrementing the ref count.

Based on a patch by Louis Lenders. Although MSDN claims this should be in
shell32, it really is in shdocvw.

This should be the last unnecessary getFormatDescEntry() call.

Rather than looking it up all the time.

We can use the information from format_desc instead.

As far as I can tell nothing depends on this being 1, and there's some code
that checks for WINED3DFMT_UNKNOWN and uses 0 for the byte count in that case.

This mostly just reduces the nesting to something more sane.

Comparing gl_shaders against NULL after potentially dereferencing it a few
lines earlier probably confuses Coverity. Checking num_gl_shaders instead
should avoid this.

This was partially handled by ungetting the end group token (i.e. '}')
so that it was read again at the proper place.  Unfortunately there is a
read hook that maintains a stackTop variable, which is decremented when
receiving this end group token.  Therefore stackTop would get
decremented twice, and the rich text file would end prematurely.

This is fixed by incrementing the stackTop variable to compensate for it
being decremented twice.