Adding "discard" after "break" in WebGL (ANGLE) shader causes wrong image to be rendered

[Alastair_D_1]

The files associated with the bug are here:

https://github.com/mc-imperial/shader-compiler-bugs/tree/master/Intel-HD-Graphics-520-WebGL-ANGLE/wrong_images/large-v100-webgl-e404f04495815667_inv_variant_71

System Setup Information:
-----------------------------------------

System Used: Dell XPS
CPU SKU: i7
GPU SKU: HD 520
Processor Line: U-Series
System BIOS Version: Dell Inc. 1.0.0, 10/09/2015
CMOS settings: I can check on request
Graphics Driver Version: 20.19.15.4501
GOP/VBIOS Version: Intel Video BIOS
Operating System: Windows 10
OS Version: Pro
API: WebGL 1.0 (OpenGL ES 2.0 Chromium), WebGL vendor: Google Inc., WebGL renderer: ANGLE (Intel(R) HD Graphics 520 Direct3D11 vs_5_0 ps_5_0), broswer version: Google Chrome Version 54.0.2840.99 m (64-bit)
Occurs on non-Intel GPUs?: Haven't tried

Steps to Reproduce:
-------------------------------
1. Open the Chrome browser
2. Navigate to http://htmlpreview.github.io/?https://github.com/mc-imperial/shader-compiler-bugs/blob/master/utils/webgl/webgl_viewer.html
3. Paste the contents of original.frag into the text box
4. Click "Execute"; the rendered image is the expected image for this shader, and should look like original.png
5. Paste the contents of variant.frag into the text box
6. Click "Execute"; you should find that the rendered image is significantly different from the expected image rendered previously

Expected Results:
-------------------------------
Both shaders should render an image that looks like original.png.  This is because the only difference between the .frag files is:

- There is a "discard" statement after a "break" statement in
  variant.frag; because the "discard" statement is unreachable code,
  this should have no effect on computation.

Actual Results:
-------------------------------
The shaders render significantly different images, which they should not.

Additional Information:
-------------------------------
N/A

[Alastair_D_1]

I had meant to post this under here:

https://software.intel.com/en-us/forums/graphics-driver-bug-reporting

Is it possible for it to be moved there?

[Michael_C_Intel2]

Hi Alastair,

Can you see if the issue reproduces on this new driver? 

https://downloadcenter.intel.com/download/26404/Intel-Graphics-Driver-15-45-

I will move this to the sub forum for you.

-Michael

[Alastair_D_1]

Hi Michael

Yes, this does reproduce under the new driver.  The symptoms are a little different, but having the "discard" after the break or not completely changes what is rendered.

I'd be interested to know what the issue is.

Thanks

Ally

[Michael_C_Intel2]

Thanks! I have the ticket filed and will the driver team to investigate.