Re: [PATCH 0/25] MR7226: ntdll: Implement in-process synchronization via the Linux "ntsync" driver.

[Aida Jonikienė (＠DodoGTA)]

Aida Jonikienė (@DodoGTA) commented about dlls/ntdll/unix/sync.c:
> {
>      union select_op select_op;
>      UINT i, flags = SELECT_INTERRUPTIBLE;
> -
>
> unsigned int ret;
>
> if (!count || count > MAXIMUM_WAIT_OBJECTS) return STATUS_INVALID_PARAMETER_1;
> -
>
> if (TRACE_ON(sync))
> -
>
> {
> -
>    TRACE( "wait_any %u, alertable %u, handles {%p", wait_any, alertable, handles[0] );
> -
>    for (i = 1; i < count; i++) TRACE( ", %p", handles[i] );
> -
>    TRACE( "}, timeout %s\n", debugstr_timeout(timeout) );
> -
>
> }
> -
> -
>
> if ((ret = inproc_wait( count, handles, wait_any, alertable, timeout )) != STATUS_NOT_IMPLEMENTED)

This is going to fail with no server fallback with completion waits
(because `completion_wait_ops` has `no_get_inproc_sync`)

That failure can have realistic consequences (like GTA 3 crashing because
of an `quartz` assert caused by this function failing)

So can there be an extra `STATUS_OBJECT_TYPE_MISMATCH` check here
(or actually implementing completion wait support if that's even possible)?