r328: OS/2 Warp 3 format error; CD access crashes PCem

[neozeed]

The 2.11 install was straightforward. I was very conservative in the setup (486 66Mhz, ET4000AX, SB16, 8MB ram, 500MB HD). Ram was bumped up to 32MB after install and running ok at 1024x768. Multimedia services works partially. It messes up the SB16 IRQ and DMA and those need to be changed. For some reason only digital audio outputs anything, MIDI doesn't work. I haven't really checked into why that is yet.
For the life of me I can't find the TCP/IP install disks anywhere. I thought I had the complete IBM software sets but I was wrong.

The os2 sound blaster drivers are crazy.. creative did their best to block as many clone cards as possible. It was a real pain back then.

I have tcp base 2.0 which works fine on 2.1 .. load the hotfixes though!

[ppgrainbow]

Running the Windows Clock locks up the Windows subsystem in OS/2 2.1 also.

That's awfully bizarre to see the Win-OS/2 Clock to lock up under Win-OS/2.

Does the Windows 3.1 version of Clock work under Win-OS/2 or do you get the same issue?

[amadama]

Midi doesn't seem to work on Warp either. Maybe it's me, can someone try to play a midi file on their Warp 3 setup?

[neozeed]

Midi doesn't seem to work on Warp either. Maybe it's me, can someone try to play a midi file on their Warp 3 setup?

While looking at sound_sb.c I noticed this in the sb_pro_v2_init that I could just add in the following line:

Code: Select all

mpu401_uart_init(&sb->mpu, 0x330);

And now I can play DooM using general MIDI on port 0x330. The AWE32 sets this up as well, but I wanted the sound IRQ on 5, which the older soundblasters let you configure, but not the AWE32 for some reason. I'm wondering if there is a reason why the MPU401 couldn't be enabled or disabled at will for all the sound cards? Or even maybe on it's own?

And while we are talking about old hardware and limitations, I went ahead and tried to add a 1GB disk (63 sectors, 16 heads, 2080 cylinders) and noticed that with the Award SiS and above, if I put it in LBA mode, I can exchange data between that 486 and the Pentiums OK.

Also some volume levels are so insanely quiet. I have to crank my sound up to 100% and hope nothing else makes any noise.

[Battler]

- neozeed: The IRQ 7 as default for sound cards makes me surprised they even work, since PCem also assigns the exact same IRQ to the emulated LPT 1 port.

[neozeed]

- neozeed: The IRQ 7 as default for sound cards makes me surprised they even work, since PCem also assigns the exact same IRQ to the emulated LPT 1 port.

Yeah, I know so many people who did this IRL and had nothing but grief with it. I know virtual stuff is more forgiving, and PCI can happily share, but I've always liked to keep my peripherals separate.

[ecksemmess]

Don't install quemm. or emm386 or anything else like it.

try to keep the configuration as PLAIN and SIMPLE as possible.

More like, the bug in PCem needs to be fixed. Again, I did I=B000-B7FF on real hardware all the time and never had these problems. Only in PCem does that cause problems.

PCem's memory map isn't like your 'real pc' and this has nothing to do with WARP.

viewtopic.php?f=2&t=303

A bit off topic, but my curiosity compels me to ask: neozeed, can you clarify what you meant here? How is PCem's memory map unlike that of a "real PC"? I've read that other thread you linked, but it's not really clear which part of it is supposed to be relevant to that comment.

[neozeed]

A bit off topic, but my curiosity compels me to ask: neozeed, can you clarify what you meant here? How is PCem's memory map unlike that of a "real PC"? I've read that other thread you linked, but it's not really clear which part of it is supposed to be relevant to that comment.

Well for one thing you can actually look at it with absolute certainty. The debug versions dump out the physical memory, so there is nothing hiding or obscuring things so you can actually see what is where. Plus who knows how many cards, ROM's or memory mapped buffers other people have on other machines. Nothing is guaranteed which is why the whole UMB/EMS time was a really really dark time. You had programs like Return to Zork that needed 563,424 base ram, along with 19,519kb for sound drivers. Sure 582,583 bytes is a joke today, but back then it was a nightmare to deal with. And you couldn't give directions that would work 100% of the time on everyone's PC as to what memory was available in the UMB because you have no idea what cards they have or how they are configured. The same goes with IO ports, and IRQ's. This is the hell that was MS-DOS+ISA.

DPMI was one way around this, as you could just use memory above 1MB, and access it as one nice big segment. If you want a 4MB zone, just malloc(4*1024*1024); and you have 4 megabytes. DPMI is also why Windows 3.0 was so popular, as people had all this RAM in their PC but couldn't use it. And EMS frames take away from UMB's so there is that tradeoff as well.

PCem is really cool as I can give you my config files (PCem/config.sys/autoexec.bat) and you can reproduce what I have.

Also when it comes to drivers and whatnot the first thing to do is don't try to load them high. Things can get unpredictable as some drivers just won't work in the UMB. So it's best to load low, and without any memory manager first. If everything is fine, then try to load high. Otherwise if it fails who knows why? Like everything else, it's about changing one variable at a time, and recording results. Otherwise you are just going to make a mess, and nobody will be able to reproduce what happened.

[ecksemmess]

Return to Zork required 19,519 *KB* of RAM for its sound drivers?! No wonder I had so much trouble getting that game running! In all seriousness, yes, 580-some KB was an incredibly daunting requirement back in the day. I remember. Oh god, do I remember. I was there for "the hell that was MS-DOS+ISA", and I can vouch for everything you say. I don't know about you, but as I recall it, the ritual was generally "Get new game -> Attempt to install, configure, and run said new game -> Spend sixteen straight hours trying to divine meaning from inscrutable error messages, diving into the murky depths of the system, modifying one setting in one config file at a time until all 3,280,476 permutations have been exhausted, praying to various deities and promising them a religious conversion in exchange for their intervention, etc. -> Fail miserably anyway -> Pick up the phone (which, if lucky, MAY have been cordless, lol) and start calling everyone in the literal, physical address book on the desk who might have more expertise and beg them for their advice -> ??? -> Profit!", or something along those lines. Anyway, I was just confused by your comment because I thought you meant to say that PCem's memory map is categorically unlike that of any real PC. Now I see that you just meant to point out that the memory maps of any two "real PCs" from that era wouldn't even match each other, so of course no one should expect PCem's memory map to perfectly match that of any one given real PC. If that is in fact what you meant, then of course I agree.

I'll spare you the litany of RtZ quotes that I sorely want to use to close out this post...

[neozeed]

I'll spare you the litany of RtZ quotes that I sorely want to use to close out this post...

Who're you?

Don't matter.

Want some rye? Of course you do!

Oh and I found this save game, if you want to beat it on the quick! I know it works with my floppy version.

Which is one thing, after installing OS/2, Windows NT, Office, and something like this from floppies it was not a fun experience... Kind of funny how PCem can force us to re-live slow POST times, small disks, and memory/irq/dma conflicts!

[ppgrainbow]

Has anyone tried r373 of PCem yet?

This revision fixes the ATAPI interrupt status following Mode Select command when attempting to install OS/2 Warp 3 using the virtual CD-ROM ISO or real CD-ROM disk.