On Thu, 8 Apr 2021 10:39:38 GMT, Ajit Ghaisas <aghai...@openjdk.org> wrote:
>> src/java.desktop/macosx/classes/sun/java2d/metal/MTLGraphicsConfig.java line
>> 149:
>>
>>> 147: try {
>>> 148: // getMTLConfigInfo() creates new MTLContext, so we should
>>> first
>>> 149: // invalidate the current Java-level context and flush the
>>> queue...
>>
>> The old discussion was related not only to the comment but to the
>> invalidateCurrentContext, do we need to do it?
>
> This is the only place where we use MTLContext.invalidateCurrentContext() -
> which when processed in MTLRenderQueue - clears some native stuff and nulls
> out both mtlc and dstOps pointers maintained in MTLRenderQueue.m. I think,
> this will be important once we get rid of SET_SCRATCH_SURFACE under
> JDK-8263309.
But why you need to invalidate context here? Why do you need "clears some
native stuff and nulls out both mtlc and dstOps pointers maintained in
MTLRenderQueue.m"?
In OGL the getCGLConfigInfo() change the state of the OGL state due to
"makeCurrentContext", this is why we need to update the lava level state to
"invalid", otherwise we will get a mismatch between the state in the native and
java state.
-------------
PR: https://git.openjdk.java.net/jdk/pull/3357
