clb wrote on 2023-08-13, 16:26:

mkarcher wrote on 2023-08-13, 01:35:

You can easily program the Start Address Low and Start Address High registers in a way you can observe the "Keen issue" on the screen. [...] I don't see a way to find out whether this wrapping occurs or not without asking the user what's visible on the screen.

CRT Terminator is able to observe this, and route necessary information to the user. I wrote an unit test for this Tseng scrolling and DWORD address decoding behavior, something like below.

Oh, good point actually. The CRT Terminator gets to see all on-screen pixels, so you don't have to ask the user what pixels are sent from the graphics card to the monitor.

clb wrote on 2023-08-13, 16:28:

Pierre32 wrote on 2023-08-13, 10:29:

In the release version of the hardware, will your ribbon cable have a red core for pin 1? That might prevent some drama...

We are not planning on marking pin 1, but instead, the idea is to mark all the ground pins. This is because not all boards have pin 1 marked, so it is safer to locate via the ground pins. A multimeter in resistance mode can easily find the ground pins.

You're going to regret this. Most cards do have pin 1 marked and it is by far the simplist and most standard way to address cable orientation. Not marking pin 1 on the cable and forcing the user to dig out a multimeter is going to needlessly agitate most of your user base.

I understand the desire to mark the ground pins for those edge cases where that's the only reliable way to orient the cable, but you should absolutely do both. I'd suggest the usual red stripe on pin 1 and black for the grounds

Another vote for both. I'm cool getting the meter out and I'm sure most early adopters will be - but I think you should be prepared for this to find a wider market, and include every idiot-proofing measure you can.

Pierre32 wrote on 2023-08-13, 22:58:

Another vote for both. I'm cool getting the meter out and I'm sure most early adopters will be - but I think you should be prepared for this to find a wider market, and include every idiot-proofing measure you can.

I'm not cool with getting the meter out for no good reason. It's a matter of not wanting to be inconvenienced, not a skill issue.

If someone makes me perform needless chores, we're not going to be friends very long.

maxtherabbit wrote on 2023-08-13, 22:52:

I understand the desire to mark the ground pins for those edge cases where that's the only reliable way to orient the cable, but you should absolutely do both. I'd suggest the usual red stripe on pin 1 and black for the grounds

We can do both. There is certainly no need to raise ultimatums about being friends or not.

clb wrote on 2023-08-14, 06:08:

There is certainly no need to raise ultimatums about being friends or not.

🤣 that wasn't meant to be taken as some type of threat or personal attack, more of an explanation to the other poster about why I care about it and it not being "idiot proofing" per se

Yeh, no biggie at all. 👍

To anyone stumbling in, I posted a SNOOP log of a Headland card over on this thread while digging into something else: Can your VGA card provide a hardware source for random numbers?

Tests on a socket 7 system which has snoop in the BIOS, but it was left disabled for this.

Some odd colours in the captures, but all looked normal on the screen. Except the Ark, which did not like being snooped I guess.

Bonus ISA card test because it was sitting there

Superb, thanks for testing so many cards. It looks that all data that is printed is correct, the tool did not misdetect any card.

The S3 Trio colors going wrong might be an issue with enabling the Feature Connector there. Some S3 cards have a limitation that they require disabling half of the VRAM when Feature Connector is enabled (they cheaped out on outgoing bus lines, and shared Feature Connector with VRAM bus lines). That might cause the colors to go wrong, maybe.

The ARK 2000 situation is curious. I'll need to implement better debugging to help figure that out. Something you can try now is to run "snoop logfile" on that system, and then after the screen goes blank, reboot. Then there should exist a SNOOP.LOG file generated in the same directory of the SNOOP.EXE tool that shows some prints.Also you can try "SNOOP nos3test nomodetest novesatest disable" to skip a number of checks to see if one of those might be the cause of the blank screen. (I'll need to add more options to make that more fine grained)

Thanks for the pointers. I bring more ARK data.

First up: SNOOP nos3test nomodetest novesatest disable

I tried the switches one by one and narrowed it down to nomodetest - which got snoop enabled and some more lovely colouring:

SNOOP logfile attached.

Pierre32 wrote on 2023-08-21, 09:08:

I tried the switches one by one and narrowed it down to nomodetest - which got snoop enabled and some more lovely colouring:

Huh, out of all the disable flags, that one was the one that was most unexpected.

What that check does is it just simply switches video mode from Mode 03h 80x25 text to Mode 13h 320x200 and back: https://github.com/juj/crt_terminator/blob/16 … P.CPP#L121-L126

I can only think that either getting the current video mode ID does not work on that card quite well, or Borland Turbo C++ has a bug with reusing the REGS structure for multiple int86() calls. (I have seen that behave oddly in the past).

I might create some specific debug builds later to figure out what is going wrong on in the code with that card specifically. Oh, and looks like I've snuck in a bug at some point that the PCI Vendor and PCI Device are getting printed twice.

And the detection of a complex form of Hidden RAMDAC register is not working for that card. It looks like it should have printed just "B1h 00h" there instead of the long sequence it did.

Updated SNOOP version v.Aug 23 2023 now detects whether the Vertical Retrace interrupt is supported on the board, and whether the Light Pen Address registers can be used as a source of Hardware RNG (Can your VGA card provide a hardware source for random numbers?). Additionally, running

1SNOOP perf

will measure VGA 320x200 mode individual pixel read/write performance.

I don't yet know of safe it is to unconditionally test for IRQ2. So if there are any issues, I'll have to move that behind an option.

A BIG test (snoop 2023-08-20) - 40! AGP and PCI cards. Have some issues there. 🙁 Can you reformat snoop output so it will definitely fit in 80x25 or make it work in 80x30, 80x50, 80x60 text mode(s)?

Superb work! Thanks for testing so many cards.

Of particular surprise to me is the Voodoo 3: it diagnoses Palette Snoop setting to be cleared on video mode change. Before this I knew of only one specific Cirrus Logic model (Diamond Speedstar 64 1 MB PCI CL-GD5434) that did that. That quirk is a little bit awkward, but well fixable with a TSR, so not a huge deal.

Also of interest are the lines where it says "VGA Adapter: ... unknown ...", since those signal that there is room to improve the detection code in SNOOP. I'll route back to improving that later.

The output height is a constant struggle. I've worked a ton to try to squeeze it fit on a single 80x25 screen, but these new cards that support many modes that definitely won't do. I did have a version that outputted 80x50 earlier, but cleared it due to my code not working on some old ISA cards well. I'll need to look into seeing how to improve that.

Something that would be interesting to test with the new SNOOP version I just updated, is if the AGP Voodoo 3 card will report the "lp-rng" capability (Lightpen RNG).

ViTi95 reported in Re: Can your VGA card provide a hardware source for random numbers? that his PCI Voodoo 3 had it, so that makes me wonder if the AGP variant also does. If so, then the RNG.ZIP from Can your VGA card provide a hardware source for random numbers? should also work on the AGP Voodoo 3 card.

Made some new SNOOP and RNG (3dfx) logs including problematic cards.

Thanks! Indeed looks like all Voodoo 3s then have Light Pen entropy, independent of whether they are PCI or AGP form. Nice 😀

New version SNOOP v.Aug 24 2023:
- Fixed duplicate PCI adapter printing
- Made lp-rng detection more accurate
- Fixed Hidden RAMDAC detection
- Switch video mode to 80x50 if output would scroll beyond 80x25
- Toggled "perf" option to a "noperftest" negative option like the other settings.

I tried to run SNOOP on my Siemens Nixdorf Scenic Pro M5 with integrated Matrox MGA graphics (like a Mystique 220 http://www.vgamuseum.info/index.php/cpu/item/ … sg-mystique-220 )
(BTW doesnt support RNG 😉 )
It crashes after the first line and i have to reboot the computer manually.
Just to let you know.