[0.6.0+58+g2881cff75] Linux: Strange interpolation issues

[mjr4077au]

System:
Arch Linux
SDL 2.0.12
GCC 9.3.0

Games:
Duke 3D
Shadow Warrior
(presumably others)

Basically in the latest git head version, when moving forward, backwards or sideways there is noticable stuttering occurring but not when looking around with the mouse. It's as if the games aren't interpolating bwtween opos and pos values or something.

This issue does not occur when I build the 6.0.0 release version on the same system.

[Graf Zahl]

I'm afraid that you'll have to track down the commit when this started yourself. I cannot see this on Windows.

[mjr4077au]

Will have a look before it gets lost with more backend changes. FWIW, this isn't my native platform either, just noticed it when I was trying to help someone get a compile going on Discord

[mjr4077au]

It's so hard making all this work commit-by-commit since it's all quite broken until the end, but I believe the math addtions in f671eb6 are highly likely to be the cause.

There's also other issues with this commit as it brings in files like https://github.com/coelckers/Raze/blob/ ... /m_fixed.h which has significant overlap with the definitions in source/build/include/pragmas.h.

I can confirm that none of the extensive commits at the end to do with all the palette work are causing this, it's something older than this: 29f95e2.

[Graf Zahl]

I know that m_fixed.h has some overlap but I'm going to need this in the backend for the renderer to work.

If you think it's the math, you can edit cmath.h and comment out the USE_CUSTOM_MATH and USE_FAST_MATH #defines and see if that makes any difference.

[mjr4077au]

This took such a long time to determine... PR to fix my reported issues is here.

I apologise about the noise regarding the math libs. What was happening was that I was creating branches at certain commits and then cherry-picking this commit as part of fixing build issues. Only when I stopped those cherry picks and just fixed the build errors as they occurred was I able to isolate the true issue.

I suspect there will need to be a similar fix done for the Cocoa backend and have tagged _mental_ in the pull request since he does those builds.