News for hackers. Stuff that matters.
Re:ReactOS 0.4.11 Released by tombert on Monday March 04, 2019 @10:58AM (#19302181)
Despite being somewhat of a Unix/Linux junky, I really want ReactOS to be a thing. I would love to get older family members who refuse to upgrade from Windows Vista to something that is still being updated and will work on their older hardware.

That said, the last three times I've tried dogfooding ReactOS, I had trouble keeping it stable for more than 20 minutes, which is a complete non-starter for getting non-technical people using it.

It seems like the people working on the project are very smart and motivated, so I have to wonder why this is still a huge issue? I know almost nothing of OS/Kernel dev, so is it basically impossible to salvage all the stability work from the Linux kernel?

Re:ReactOS 0.4.11 Released by yjftsjthsd-h on Monday March 04, 2019 @11:18AM (#19302345)
> is it basically impossible to salvage all the stability work from the Linux kernel?

Short answer: Yes.

Medium answer: That would be a near-complete rewrite of the project, since the ReactOS kernel and Linux have nothing in common. Different architectures, code styles, conventions, APIs, ABIs (which is super important for drivers). Probably the best option would be to port all of the ReactOS userland to run on WINE (which would actually be cool, and might be easy), and possibly write a Linux kernel module that provided a NT-compatible driver ABI (which would probably be hell to write, but if it's even possible then it would be a huge boon to Linux hardware compatibility... probably).

Re:ReactOS 0.4.11 Released by slacka on Monday March 04, 2019 @11:57AM (#19302707)
> APIs, ABIs (which is super important for drivers)

Yes, only userspace code would be reusable with a linux based kernel. Might as well start from scratch.

One more option that's being tried is to split the difference by integrating more of Wine's code. This approach is being testing in the Arwinss branch[1]

[1] https://reactos.org/wiki/Arwinss

Re:ReactOS 0.4.11 Released by poizan42 on Monday March 04, 2019 @12:54PM (#19303410)
> and possibly write a Linux kernel module that provided a NT-compatible driver ABI

NDISwrapper, Captive NTFS, Linuxant? Sure none of them implement(s|ed) the whole interface, but it shows that it is far from an impossible task.

Re:ReactOS 0.4.11 Released by DC-3 on Monday March 04, 2019 @12:10PM (#19302837)
> dogfooding

Or as I like to call it, 'using'.

Re:ReactOS 0.4.11 Released by tombert on Monday March 04, 2019 @01:27PM (#19303778)
Well, I meant this in the "practice what you preach" sense. I actually nuked my laptop and went all-in with React for a couple days each time. "Using" is more vague and could just mean I tried running it in a VM for ten minutes.
Re:ReactOS 0.4.11 Released by devrod on Monday March 04, 2019 @04:26PM (#19305705)
> practice what you preach

But that's not dogfooding. If you were a ReactOS developer using ReactOS to compile future version of ReactOS, then you would be "eating your own dogfood", but simply using it as a daily driver doesn't count.

Re:ReactOS 0.4.11 Released by tombert on Tuesday March 05, 2019 @12:21PM (#19312160)
I apologize for misusing the term.
Re:ReactOS 0.4.11 Released by coldacid on Monday March 04, 2019 @01:53PM (#19304083)
Yeah, it's only dogfooding if you're the developer of a project.
Re:ReactOS 0.4.11 Released by bgorman on Monday March 04, 2019 @12:13PM (#19302872)
The main issue is developing device drivers is very complicated and difficult to get right. A sensible approach IMO would be to have ReactOS run virtualized on top of the Linux kernel. That way Users get all of the stability benefits of Linux, plus hardware support. The ReactOS team would then be able to concentrate on Windows compatibility, while only developing a few drivers for virtualization. The downside with this approach is a huge performance hit on computers without virtualization extensions. However I think the main use case of ReactOS is to provide a way to run old windows programs on modern hardware, so I don't think this would be a concern for most potential commercial users.
Re:ReactOS 0.4.11 Released by restalis on Monday March 04, 2019 @12:52PM (#19303380)
Running it in a VM under a Linux host is exactly how a lot of involved people have it, the reason being that that setup was more accessible, more stress-tested, and thus got more stable compared to a real-hardware setup. The goal however is to develop it to be compatible with current Windows NT drivers, which in some cases still happens to have received more love from their vendors compared to their Linux versions.
Re:ReactOS 0.4.11 Released by jolmg on Monday March 04, 2019 @12:15PM (#19302889)
That's Wine, isn't it?
Re:ReactOS 0.4.11 Released by iwalton3 on Monday March 04, 2019 @12:39PM (#19303207)
Wine translates Windows syscalls to POSIX syscalls without using virtualization.
Re:ReactOS 0.4.11 Released by restalis on Monday March 04, 2019 @11:25AM (#19302416)
"is it basically impossible to salvage all the stability work from the Linux kernel?"

If you'll search through ReactOS codebase, you'll definitely encounter Linux pieces of code, thus it's safe to assume ReactOS doesn't suffer from NIH syndrome and that it salvages whatever it can from any given legit available source. Other than that, ReactOS has its own NT reference (which covers stability too) that it has to follow and that isn't any of the easy-available *nix architectures, so there's that.

Re:ReactOS 0.4.11 Released by Paianni on Monday March 04, 2019 @07:59PM (#19307222)
Server 2008 is based on Vista and still supported...