Well, if nothing becomes available I am more inclined to buy a retrotink 4k. This costs up to 1000€ but will probably be much less painful in the long run (also has automatic phase adaption). Should be available in a couple of weeks/months. Especially with switching between 720x400 VGA (which is 320x200), 720x400 EGA (which is also 320x200), 720x400 text mode, and 720x400 VGA, which is 640x400 interlaced) automatic phase adaption will possibly solve all the issues at once.

Saying that doesn't exactly put your feature requests at the top of the pile. Then again, it's a useful feature for those on Windows, so I don't mind seeing someone get paid to do it.

Yeah, I dont see anyone doing these adjustments anyways unless they need the features themselves, like streamers and co. But like me, its unlikely that any of the few streamers (who cover vga games in the first place plus have a used 1000$ datapath), got the skillset to do that unfortunately.
And I wouldnt even be able to see how much work that is. But I assume anything will take between 1-2 hours (just to have a look) up till dozens of hours to kill bugs or add features. With the rates you mentioned thats an insane amount of money, with no guarantee of getting a solution out of it. So...I am just being reasonable. I definitely do appreciate the work you did so far. I mean at this point in time datapath+vcs is still the best option to capture vga. And whether the Retrotink 4k (first one to have VGA support) will be the "can-do-it-all" tool, is yet to be seen (but it looks promising).

If you think the rates are insane then what you had was insane value, a fluke. You're trying to replicate it and it's not happening. Hardware always needs software so it always comes full circle.

Version 3.1.0 is out. UI improvements and some others. Full changelog on GitHub. https://github.com/leikareipa/vcs/releases/tag/v3.1.0

Version 3.2.0 is out. Along with version 3.1.1, it adds a few things and removes a few things. Full changelog for both versions on GitHub. PCem 17 capture is now supported via the MMAP interface, though tested only briefly.

Version 3.3.1 is out, which together with version 3.3.0 adds two previously Windows-only features: alias resolutions and capture rate throttling. Also adds other things, like the ability to lock a video preset so it's always active, and a more economical eco mode.

Version 3.4.0 is out, and there was a 3.3.2 in between that fixed some issues. The new version adds some things, among them the requested feature of a color domain setting.

What distro is known to work for VCS? I get a segmentation fault with VCS but suspect its my slightly weird ubuntu setup, so going to nuke it and install a more standard setup.
I think I want kernel 5.15 (I did patch the datapath driver to compile on newer kernels but maybe thats whats problematic, 5.18+ changed the names of the DMA api), QT5 and OpenCV 4.

Currently have a R580 video card in the machine, but have quite a few different ones to swap if NV is known to work.

Just want to get a known good setup before I hit the debugger 😀

I use Ubuntu 20.04. You can compile with the virtual capture backend (CAPTURE_BACKEND_VIRTUAL in vcs.pro) to rule out a problem with the Datapath capture chain. Otherwise compile with VCS_DEBUG_BUILD and see if the debugger tells you anything.

Have it working now, all compiles clean with Ubuntu 20.04, but it still crashed at startup. Little bit of debugging and figured it out, patch incoming this evening or maybe tomorrow (its just one line, do you prefer here or github PR?). It looks like a clean install with no aliases crashes due to trying to a split string that doesn't have the right characters. My fix just adds a defensive line to ignore the permanent setting alias line that's incorrect. There may be a better fix (where the incorrect alias line comes from?), but don't know the code base so just fixed it in the simplest way.

VCS preview quality is awesome! Though have some problems with OBS window capture quality and dropped frames if not in ECO mode. Will investigate over next few evenings.

Thanks. Looking at the code my guess is since the alias resolution loader defaults to an empty string it generates a non-empty QStringList with an empty string and .at() asserts. I've fixed it by defaulting to an empty QStringList, which doesn't enter the split loop. Ideally the assertion would show a more meaningful message though, since it's there to catch malformed strings.

No idea about OBS. VCS will readily drop frames if you navigate the UI, but generally shouldn't happen otherwise. You can look at the output latency and see if you're spiking above what's needed for the capture refresh rate.

Are there any links to the Windows version of this? I can't find it anywhere.

How come more Windows folk aren't contributing to this? Some unemployed Linux user can't be the only one who can maintain the Windows version.

Because windows people are outside touching grass?

luckybob wrote on 2024-03-17, 15:27:

Because windows people are outside touching grass?

Nah, grass ain't growing here just yet.

I hate to stereotype people, but a (presumably) linux person asking why the windows community isn't out programming for "fun", is so on brand it physically hurts.

I'm not trying to disparage you, I swear. I just bought Factario, and it is going to kill hundreds if not thousands of hours of off time.

luckybob wrote on 2024-03-17, 19:33:

a (presumably) linux person asking why the windows community isn't out programming for "fun", is so on brand it physically hurts.

So is a Windows person wanting to be spoon fed a binary instead of going through a few (and rather well documented, in this case) build steps. It's an open source project, the author owes us nothing.

I bet that even if there was a zipped, prebuilt binary available for download, someone would complain that they expected a setup.exe instead.

Yeah, it's a niche in a niche, in a niche basically.
Like the group of people using a datapath the same way we are, and has a interest to even use custom software, must be extremely small (maybe 100 people or so?), but then having an overlap with programmers who are actually familiar with that stuff (I do have some programming knowledge in other fields but wouldn't be able to build an executable out of these source files, never done it, never had to look into it), and then having an overlap with people who are also fine with dedicating their time on that...and then that overlap would then also need to be a Windows user. And I guess it's the last overlap where we see, probably statistic probability fails us, as vbee happens to be a Linux user. ^^
And with new tech available now, I doubt the situation will shift to the better, as the demand will also shrink.

elszgensa wrote on 2024-03-17, 19:46:

luckybob wrote on 2024-03-17, 19:33:

a (presumably) linux person asking why the windows community isn't out programming for "fun", is so on brand it physically hurts.

So is a Windows person wanting to be spoon fed a binary instead of going through a few (and rather well documented, in this case) build steps. It's an open source project, the author owes us nothing.

I bet that even if there was a zipped, prebuilt binary available for download, someone would complain that they expected a setup.exe instead.

Sounds about right to me. I personally can faff about with hardware for hours on end. But you put any sort of software in front of me and I will shut down like a monkey doing a math problem. Hell, I've used the same windows theme since win7, because I just cannot bring myself to fiddle with software.