Henri Verbeet a écrit :
On 16 February 2010 19:32, Christian Costa <[email protected]> wrote:
So we should never enter this function at all. Right ?
Yes.
And what do you mean by "something is broken" ? Are you talking about
wined3d code ?
It means there's code that marks a state dirty that doesn't exist.
Typically that's because the code doing that should check some limit
or if certain features are available.
That said I don't understand why we should not display the state. The fact
is we entered this code
and it seems, according to you, that knowing the state which produces the
error message makes the difference.
Yes, but you can't check that here, the state is always 0. You should
check "rep" for zero in IWineD3DDeviceImpl_MarkStateDirty() and
Context_MarkStateDirty(), and get a backtrace if it is. I.e.,
something like the attached patch. We can't add DebugBreak()'s like
that to wined3d, and often the error is mostly harmless, but I've been
thinking about adding extra validation code like that for e.g. the
state table and locking behind some debugging define in order to share
debugging/validation code.
That's clearer. Thanks.
I've just tested your patch. The incriminated state is
WINED3DRS_TWEENFACTOR.
I haven't investigated much though...
