FS#66700 - [vulkan-validation-layers] Please consider offering unstripped binaries
Attached to Project:
Arch Linux
Opened by mark r (io7m) - Monday, 18 May 2020, 19:34 GMT
Last edited by Laurent Carlier (lordheavy) - Thursday, 18 June 2020, 13:04 GMT
Opened by mark r (io7m) - Monday, 18 May 2020, 19:34 GMT
Last edited by Laurent Carlier (lordheavy) - Thursday, 18 June 2020, 13:04 GMT
|
Details
Description:
I recently encountered an issue whilst writing some Vulkan code that resulted in a crash inside the validation layers. This was a bug in _my_ code, not in the layers, but the only way that could be determined was because I went and compiled a debug version of the layers that had not had symbols stripped. See: https://github.com/KhronosGroup/Vulkan-ValidationLayers/issues/1847 Given that the Vulkan validation layers are ostensibly a development tool, and given that stripping symbols demonstrably reduces their utility... Is there any chance you'd consider making an exception to the usual Arch policy of stripping libraries? Additional info: No extra info. Steps to reproduce: Introduce a bug in your code, and then try to use the validation layers to find it. :) |
This task depends upon
Closed by Laurent Carlier (lordheavy)
Thursday, 18 June 2020, 13:04 GMT
Reason for closing: Fixed
Additional comments about closing: vulkan-validation-layers-1.2.140-2
Thursday, 18 June 2020, 13:04 GMT
Reason for closing: Fixed
Additional comments about closing: vulkan-validation-layers-1.2.140-2
> Given that the Vulkan validation layers are ostensibly a development tool, and given that stripping symbols demonstrably reduces their utility...
> Is there any chance you'd consider making an exception to the usual Arch policy of stripping libraries?
I would go further that without debug symbols the validation layers are of no real use to anyone developing a vulkan application, and indeed I can't think of any other usecase for them.
FYI: When VK_LAYER_KHRONOS_validation is active in combination with nvidia-vulkan from AUR (had the issue with older drivers as well) loading vkGetImageMemoryRequirements2 via vkGetDeviceProcAddr returns a NULL - which the validation presumably tried to forward me to. Still, without the validation layer i get the correct function pointer. Loading the extension variant vkGetImageMemoryRequirements2KHR works either way...
I can work with that, thanks!