|
Post by goemon on Jan 13, 2021 20:14:39 GMT
Has somebody made some serious testing of TEOS Beta4 so far? Space Harrier is very buggy. The screen is shown 4 times on the screen.
|
|
sarge
Punkic Cyborg
Posts: 104
Fave PCE Shooter: Soldier Blade
Fave PCE Game Overall: Dracula X: Rondo of Blood
Fave PCE RPG: Dungeon Explorer II
|
Post by sarge on Jan 13, 2021 20:51:26 GMT
I'm seeing the same behavior on my end, too. Both Japanese and USA ROMs.
|
|
|
Post by goemon on Jan 13, 2021 21:27:19 GMT
I tried also other Sega games and they worked. Space Harrier is the only game so far showing such an error.
|
|
sarge
Punkic Cyborg
Posts: 104
Fave PCE Shooter: Soldier Blade
Fave PCE Game Overall: Dracula X: Rondo of Blood
Fave PCE RPG: Dungeon Explorer II
|
Post by sarge on Jan 14, 2021 0:09:24 GMT
I know it's not recommended to run it as a main OS, but I wonder what would happen if it were booted that way, instead of through KRIKzz's OS first? (I'm not brave enough to give it a shot, although I suspect it wouldn't ruin anything.)
|
|
|
Post by goemon on Jan 14, 2021 0:44:33 GMT
I haven't tried that. But I think the game should run without trouble in any case. Must be a strange bug.
|
|
|
Post by goemon on Jan 15, 2021 20:09:02 GMT
Okay, I tested most of the japanese Hu Card collection and found a few more games with the same 4 screen issue. There are also two games with different problems. Here it goes: Shiryou Sensen < 4 screen problem Tower of Druga < hangs at a green screen and doesnt start Wonder Momo < looks like a corrupted titel screen Out Live < 4 screen problem Benkei Gaiden < 4 screen problem Gekisha Boy < 4 screen problem
I didnt tested most of the sports games and Mahjong games.
|
|
sarge
Punkic Cyborg
Posts: 104
Fave PCE Shooter: Soldier Blade
Fave PCE Game Overall: Dracula X: Rondo of Blood
Fave PCE RPG: Dungeon Explorer II
|
Post by sarge on Jan 17, 2021 4:54:05 GMT
I can confirm all of these but Wonder Momo (looks fine on my end), and I also have two versions of Benkei Gaiden, one of which is labeled "alt 1" that works while the other does the four-screen thing.
|
|
|
Post by dshadoff on Jan 17, 2021 5:25:45 GMT
This is very interesting. I expect that it comes from these games failing to re-initialize all of the video registers which are set by TEOS, assuming that the "start-up" state settings are appropriate.
One such setting that isn't re-initialized is the dot-clock of the VCE; these games likely use 256 pixels per line (5MHz), whereas TEOS sets it to 512 pixels per line (10MHz).
Another set of possible settings would be the horizontal and vertical apertures (start/size/end); these are almost certainly set by TEOS, and although the cards are likely to have initialized some of these values, it is possible that they didn't initialize all of them.
Unfortunately, we don't know the startup values for the video registers, and I don't believe that their values can be queried.
|
|
|
Post by dshadoff on Jan 20, 2021 4:57:13 GMT
Oh... good news/bad news...
Good news: I found the teos repository on GitHub, to look at the code and see if there was something missing...
Bad news: half of the code seesm to be missing. teos_bank1.s and teos_bank2.s (maybe more)
|
|
|
Post by goemon on Jan 20, 2021 7:47:18 GMT
cool for looking into it is elmer no more around to fill in the missing code?
|
|
|
Post by spenoza on Jan 20, 2021 13:54:13 GMT
PM him and he's likely to respond.
|
|
|
Post by dshadoff on Jan 20, 2021 15:58:20 GMT
Already done.
|
|
|
Post by dshadoff on Jan 24, 2021 7:13:17 GMT
I have confirmed that the main issue is that TEOS leaves the VCE in 10MHz mode, and simply assumes that the game will reset this (which is not the case for these games).
I am hoping that elmer will return soon, so that the fix can be properly integrated back into the codebase.
In the meantime, here is a patch for the TEOS binary, if you have a hex editor:
At offset $2D9F, change: From: 4C 00 22 To: 4C F0 BF
At offset $3FF0, change: From: FF FF FF FF FF FF To: 9C 00 04 4C 00 22
This will reset the VCE back to the 256-dot mode, which fixes all of the 4-screen type issues. I also did not see an issue with Wonder Momo, (though not sure whether that's my ROM or whether the above patch fixes things).
The issue with Tower of Druaga remains, however, and that's likely related to interrupt enable/disable. If true, that would be harder to patch in the ugly way described above (but easy with the original source code and build environment).
|
|
|
Post by theshadowrunner on Jan 24, 2021 7:27:37 GMT
dshadoff, thanks for the quick fix. Isn't there a quick and dirty way, possibly as well, to double the font size in the UI via hexediting ? On GT (original lcd) the game list is very hard to read, pretty much a guessing game in itself. I asked elmer in the past if he could implement a way for users to provide their own font before he disappeared. But a quick hack would make my day as well, if you have a little time.. Thanks !
|
|
|
Post by dshadoff on Jan 24, 2021 15:26:31 GMT
Sorry, that would not be easy by any stretch - even if I had the whole source code. The pogram was designed to be 512 pixels acreen, so that it can display a lot of information on the screen. I expect It would be faster and easier to get a screen upgrade.
I also looked into Tower of Druaga's remaining issue, which seems to be waiting for an interrupt that never occurrs (or never gets past the interrupt mask). Adding CLI to re-enable interrupts isn't sufficient to fix this, so it must be related to interrupt generation. I suspect that the VDC's control register needs to have at least the VSYNC bit set (value 8, VDC register 5), and that probably isn't set as part of the transfer-of-control to the HuCard being loaded.
|
|