Yes, think the issue was discussed and was put on my desk, but not administrated correctly.

Adding clip border to determine if we need to show the clipping

CLang

Moved code to DNA_view_types as a define

Agree. We should disable clipping in these cases.

Try to be more descriptive in why you think this fixes the issue.

Nothing wrong with the fix. I would like to see more context why this solution fix this issue in general. Can be added as a description to this diff. This way when can track source code back to the issue and the reason why it is added. Already useful from a few months from now.

@Brecht Van Lommel (brecht) I would say we can start with limited support for most cases. I would use unsupported only when system can actually crash. That said that means that the current AMD Windows test record needs to be set to limited as I haven't seen any crashes, only incorrect viewport renderings.

Has been addressed by other tasks

Changed redirect links

Added GPU Name to the message boxes.
Made the X11 Message box not resizable

• Fixed Memory leak
• Correct cleanup of X11 message box
• silent warning
• Correct data in support tests array
• fixed possible issues when opengl drivers return vendor, renderer or version using multiple lines

Fixed code review issues:

Removed the reverse add in draw manager

In D5955#137728, @Brecht Van Lommel (brecht) wrote:

I think we should name the buttons something else than Help and Ok, I expect many of users will just click Ok without reading. At least for me personally, I'm sort of conditioned to think that a Help button is not going to be all that helpful. Maybe something like this:

Find Latest Drivers

Continue Anyway

Updated text

Final fixes

@Brecht Van Lommel (brecht) In order to get a clean exit from this code I would check if we could change the callback to a function and show the message box in wm_init. this will also remove some repetitive code.

Removed unneeded \ in gpu_shader.c

• Added help button in windows
• Removed callback so we can do a more clean exit (basically it crashes and therefore it exited)
• Deduplicated with GPU package.
• Many small tweaks from code review

Some small fixes

Added Low level X11 Message Box

Added windows specific stuff to GHOST so I can switch back to linux

@LazyDodo (LazyDodo) added a diff on showing system warnings/errors in GHOST.

Ah, Seems like a display issue. The color values are correct, but seems like that when displaying the buffer is set or assumed to be pre-multiplied, but in fact workbench outputs straight alpha.
This makes all alpha values other than 1.0 and 0.0 displayed incorrectly.

@Paul Kotelevets (1D_Inc) Sorry, I don't get what you want to tell.

It is a limitation of the current system; B279 shows same behavior.

@Maxximo Ramirez (maxximo32) I have dozens of issues on my desk with all the same priority. I haven't looked at this issue at all.

Seems like we need to finetune this.
Workbench was never designed to be a renderer, so we were able to hide some of these artifacts. But I agree we should check what is causing this and find a solution.

When workbench was initially designed as a viewport only draw engine. We added background dithering https://developer.blender.org/rBf938ff1010851c511f7525f4c3378f86b4e25a55 . Later it became a scene render engine, but not all options were revisited.

Before commit please initialize to -1; 0 is an actual index.

Seems to be that rebasing introduced more changes...

small changes