'Old/Weird Laptops' Sought To Help Test Linux Kernel Backlight Drivers (arstechnica.com) 33
Do you have a laptop that's either "pretty old" or "weird in some other way"? Did it ship without Windows from the factory, or did you flash its firmware with coreboot? You could help the Linux kernel move its backlight code forward without abandoning quirky gear like yours. ArsTechnica: Hans de Goede, a longtime Linux developer and principal engineer at Red Hat, writes on his Livejournal about the need to test "a special group of laptops" to prevent their backlight controls from disappearing in Linux kernel 6.1. Old laptop tests are needed because de Goede is initiating some major changes to user-space backlight controls, something he has been working on since 2014. As detailed at Linux blog Phoronix, there are multiple issues with how Linux tries to address the wide variety of backlight schemes in displays, which de Goede laid out at the recent Linux Plumbers Conference. There can be multiple backlight devices operating a single display, leaving high-level controls to "guess which one will work." Brightness control requires root permissions at the moment. And "0" passed along as a backlight value remains a conundrum, as the engineer pointed out in 2014: Is that entirely off, or as low as the display can be lit?
Why bother? (Score:2)
Next week the CPU support will be dropped as "too old". Welcome to the new linux.
Re: (Score:2)
Yes, but don't worry... the kernel will be rusty by then. I'm sure that will fix everything.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
I do, but it's running DOS so I can program old two-way radios.
Re: Why bother? (Score:2)
A lot of POS and restaurant displays basically run modern 486/Pentium hardware, theyâ(TM)re great because theyâ(TM)re x86 compatible and only use a few watts, not even Intel Atom can boast sub-5W idle. Linux was slowly being used on those devices as well as DOS developers are harder to get by.
Re: (Score:3)
And not one of those vendors will be downloading the latest and greatest kernel and deploying out for YEARS. I have worked in that industry, and I can promise you even if they re-spin the board, they will continue to use the what I would guess is a very ancient (by your standards) version of the kernel. Heck, I would bet that if you ripped apart any terminal in any industry that is running Linux that the version on the kernel is in the 4.x.x world...if not older. I realize your argument might sound good to
Re: (Score:3)
Heck, I would bet that if you ripped apart any terminal in any industry that is running Linux that the version on the kernel is in the 4.x.x world
Naah, it's the Forever Kernel 2.6.x.
Re: (Score:1)
I don't disagree on that, but there are points that these vendors will jump to the latest because some third party driver happens to require it. Eg. during the pandemic, suddenly lots of restaurants wanted to integrate their 'web' orders directly with their POS. That's when they sell their latest devices with WiFi6 and Linux.
As I said, the problem is no longer the hardware, you can find someone in China to churn out any Intel processor forever, the problem is developers, and they'll get used to things like
Re: (Score:2)
Many industrial SoCs have 486 cores, for some reason, even though they run at semi-decent speeds (266+MHz).
If you see those order ticket screens at fast food places and such where they hit a small keypad to look at orders, that keypad contains typically a 486-core SoC with 1GB of RAM and an SD card for storage. They can run DOS, Windows, Linux, etc.
They often have USB for input (USB 2) and can be had cheap on the used market.
The core's
Re: (Score:2)
Re: Why bother? (Score:2)
> typically a 486-core SoC
I want 486 cores!
Forks! (Score:2)
But anyone can fork Linux!
Re: (Score:2)
Re: (Score:1)
Re: (Score:2)
Because hardware is not ideal, nor consistent.
Additionally, would that value be a FP16? FP32? Double-precision float? Fixed point?
To make things easy for the user, it requires a lot of thought and consideration into endless edge cases, which is exactly what they're requesting here, for users to share their most extreme edge cases so they can be brought into consideration.
Re: (Score:1)
Re: (Score:2)
I'm getting shades of "we'll never need more than 640KB of RAM" here.
Simple is easiest to get right. It doesn't mean the result offers what the user wants or needs.
Or do you think your cell phone, car, or computer are the simplest option available now? I'd disagree because not having them would be simpler.
Re: (Score:1)
Re: (Score:2)
Sticking with zero is off, one is as bright as it can be, and everything else is a decimal value in between.
Decimal? What if I don't want to build floating point math into my device driver? How about 0 to 255?
Re: (Score:1)
Re: (Score:1)
Re: (Score:3)
OK, philosophical blathering aside, I'm not going to support kernel drivers that need floating point. If you think it's a good idea, you do it.
Re: (Score:1)
Re: (Score:1)
the oldest laptop i have is (Score:2)
backlights fail (Score:2)
The backlight is probably the most likely thing to fail on an old laptop.
Re: (Score:2)
On old Toshibas the connector causes a lot of issues when it corrodes, a quick claen with deoxit d5, and all good. Deoxit is the best!
Re: (Score:3)
Deoxit is the best!
Will that clean Rust out of the kernel?
Re: (Score:2)
Maybe in the early days, but not any time recently if you're getting good laptops. I've had hard disks and even an interrupt controller fail, but no backlights. Cold cathode fluorescent backlights get noticeably dimmer over time, but LED backlights don't seem to have that issue.