Open maz-1 opened 3 years ago
Hi!
Is there a small repo I can clone and test/repro this bug?
I ran into a bug (independently) that I think is the one causing this.
The texture is destroyed, but then when recreated the exact same address ptr is provided by malloc.
Thus later on when VulkanRenderPassDescriptor is recreating VulkanRenderPassDescriptor::calculateSharedKey
looks up a cache and sees there is already a cached entry (which is for the old ptr). Then VulkanRenderPassDescriptor::setupFbo
sees fboDesc.mRenderPass
is already initialized.
This bug can cause crashes, not just wrong render areas.
I'm already on it.
OK I just pushed https://github.com/OGRECave/ogre-next/commit/15e41fddd517b19d59ad850a91e3335a52c8af78 to fix this bug.
Could you confirm if your bug is fixed or still present? Thanks
just upgraded ogre to master, problem still exists on win & linux
Ouch. Ok, I'll check out your sample when I get some time.
I tried last Monday but I had to setup several dependencies before it could build
System Information
Detailled description
I'm try to embed ogre render window into qt5 widget as described here: https://forums.ogre3d.org/viewtopic.php?f=2&t=96408&p=551295#p551295
All three backends ( OpenGL, DirectX11, Vulkan) can be embedded successfully. OpenGL & DirectX11 can adjust render area when widget is resized, However Vulkan does not
Tried OpenGL & Vulkan under Archlinux and all three backends under windows 10.
Ogre.log
Callstack