Mesa 21.x Seems To Muck Up Gamers' Trust Factor For Counter-Strike: Global Offensive

[bachchain]

So far we have not seen official feedback from Valve since January. This should be quite simple to confirm/deny...

You've never dealt with Valve before, have you? Their capacity for silently ignoring problems is second-to-none.

[stormcrow]

You've never dealt with Valve before, have you? Their capacity for silently ignoring problems is second-to-none.

Still haven't figured out Unix permissions systems, that's for sure. All files in a game are marked executable whether they're library, config file, or data assets. They're only quick to make changes if their cash cows are in jeopardy.

[JonnyRobbie]

If you code the game properly it's impossible to wallhack. Sure you can make the textures transparent but it will show nothing because the server will plain not send what's on the other side of the wall to the clients.

You do realize that they have a system addressing exactly this and have been there for several years. They absolutely do have an occlusion-based anti-wh. It's just at some point, they have to start rendering the player if not only just because you cannot see an eneme, you can hear an enemy, so the positional info has to be transmitted to you anyways. But most importantly, they have to prevent popping up, which was a problem when they first introduced this method.

You obviously just want to rage at Valve without knowing much.

Wait, it's 2021 and CS:GO is still using OpenGL not Vulkan?
(DotA moved to Vulkan a while ago so I'm confused)

The general gist of it that the current codebase of csgo is an atrocious spaghetti monster of ludicrous code and changing something so fundamental would be so hard, compounded to impossibility thanks to the lack of incentive in the infamously flat Valve structure.

[HenryM]

the trust factor problem started a couple weeks before clamp_div_by_zero and glthread were made default, but some time after clamp_div_by_zero was added as an option in mesa-git, maybe a month.

neither option appears to cause problems before mesa 21. I have one computer on mesa 20.3 running both and the account that is used daily on that machine still has a good trust score.

[Davidovitch]

I've been playing with MESA_GLTHREAD on for months, but flipping that back to FALSE today, after reading this article, suddenly made me end up with shouting and insulting dudes again in competitive matchmaking. Could be a coincidence of course, or maybe the change is what triggered a lower trust factor? So either from False to True or the other way around?

[smitty3268]

Random guess, Valve knows how many threads they expect the game to use, this was based on older mesa versions. When the thread count doesn't match, this likely could be from cheats running their own thread, but since it's not fool proof, it just affects trust factor.

Some anti-cheat software uses timing as well, I believe. Run some GL commands, and if it returns instantly it's because the commands are just being ignored rather than executed - except in this case, it's being returned immediately because that stuff is thrown onto another thread.

[piorunz]

Don't play stupid DRM games, problem solved.

[skeevy420]

Don't play stupid DRM games, problem solved.

I think some form of DRM for online games is acceptable. Cheaters suck.

[stormcrow]

Don't play stupid DRM games, problem solved.

Good luck with that.

[thedukesd]

You do realize that they have a system addressing exactly this and have been there for several years. They absolutely do have an occlusion-based anti-wh. It's just at some point, they have to start rendering the player if not only just because you cannot see an eneme, you can hear an enemy, so the positional info has to be transmitted to you anyways. But most importantly, they have to prevent popping up, which was a problem when they first introduced this method.

Believe what you want. As long as the server sends information that is not actually needed by the client you can't prevent wall hacking.
Now the poping in/out can't even be prevented starting from a particular ping value.

you can hear an enemy, so the positional info has to be transmitted to you anyways.

You are 100% wrong here.

There is a difference between trashing Valve and wanting them to improve. The fact that they don't look to want to improve is another discussion.
Also the fact that I don't give much information about how to fix it has to do with the fact that Valve is greed (just look at their 30% cut).
I also don't really give much information to people that should think a bit more, maybe just maybe I don't really want to write it all (the ones working at cheat tools sell them so I don't really want to help them get money... i'm willing to help for free an open-source project as long as they are willing to accept my help, but if they refuse my help in a way or another I will just move on, rest is entirely done based on the money offered by the other side, it's not me who made this world rotate around money).

An yes if you ask me the entire math has to be done server side to prevent most of the cheating. Now the fact that I can't name a single one that didn't cheap out here is another discussion. Then to cover the fact that they cheap out add stupid anti-cheat stuff that ends up hurting the game performance. So basicaly the cost is moved to the player that sometimes even payed for the game... I don't talk here about Valve, this happens with every single one out there...

As long as you keep too much stuff client side someone motivated enough will just have the game running in a virtual machine 1/pc 1, video output sent to virtual machine 2/pc 2 and here the stuff from the cheat tool is being added to the game (with stuff that is being processed from the router in front of virtual machine 1/pc 1). No anti-cheat tool will detect what is done on pc 2 because it has no access to what's done on pc 2 or router level...