08799: revx, revxp5: Steven Tyler cutscene static colors corrupted

ID	Category [?]	Severity [?]	Reproducibility	Date Submitted	Last Update
08799	Color/Palette	Minor	Always	Dec 1, 2023, 17:00	5 days ago

Tester	yoshi200213	View Status	Public	Platform	MAME (Official Binary)
Assigned To	AJR	Resolution	Fixed	OS
Status [?]	Resolved			Driver	midway/midxunit.cpp
Version	0.261	Fixed in Version	0.280GIT	Build	64-bit
		Fixed in Git Commit	f123770	Github Pull Request #

Summary	08799: revx, revxp5: Steven Tyler cutscene static colors corrupted
Description	hello in the cutscene of Steven Tyler talking to you in the first level he static effect when the transmission is interrupted has corrupted colors compared to old mame versions. It's supposed to be a regression.
Steps To Reproduce	1. start the level. 2. play the entire level until the Steven Tyler cutscene. 3. see the static.
Additional Information
Github Commit

Flags	64-bit specific
Regression Version	0.224
Affected Sets / Systems	revx, revxp5

Attached Files	graphic bug.png (393,186 bytes) Dec 1, 2023, 17:00 Uploaded by yoshi200213
	i.zip (875,245 bytes) Mar 2, 2024, 16:26 Uploaded by yoshi200213
	1.zip (1,108,248 bytes) 5 days ago Uploaded by Osso save state for MAME 0.279

No.21985 yoshi200213 Tester Feb 25, 2024, 15:08	but the problem really happens, the cutscene at the end of the first level is glitched and this happen in recent versions of mame
No.21990 Osso Moderator Feb 28, 2024, 06:19	Could you provide a save state right before the issue appears?
No.21992 yoshi200213 Tester Mar 2, 2024, 16:15	here I have the Save state
No.21993 yoshi200213 Tester Mar 2, 2024, 16:26	here I have a save file
No.22824 yoshi200213 Tester Jan 10, 2025, 18:41	I tested the game on mame 0.273 and the bug still occurred Please fix this issues one for more Here's a video of the game on insert coin channel ?si=JKWfuHphbUG9VM7z
No.22825 yoshi200213 Tester Jan 10, 2025, 18:41 edited on: Jan 10, 2025, 18:42	It's about the Steven tiler glitched static
No.22827 PL1 Tester Jan 10, 2025, 21:19 edited on: Jan 10, 2025, 23:01	To be clear: There should be black and white static at the beginning and end of the video. This part _looks_ good. (It may be inverted.) The correct video glitch looks like a VCR tracking error/horizontal wavyness. (happens 3x during this 20 second video) The incorrect video glitch looks like it has (inverted?) color static added on top of the VCR tracking error/horizontal wavyness and the broken glass at the bottom/front of the screen is the wrong color. https://raw.githubusercontent.com/PL1-Arcade/Misc/refs/heads/main/Revx_Video_Glitch.png This recording shows it working correctly at 4:05 - 4:26, right? (Glitch in video above is at 5:00 - 5:20.) Scott
No.23551 yoshi200213 Tester 6 days ago	i tested the game on version 0.279 and the bug is still not fix. I notice that the regression start on version 0.224 5 years ago
No.23552 hap Developer 6 days ago	Don't do a topic-bump like that.
No.23553 yoshi200213 Tester 6 days ago	ops sorry
No.23554 Osso Moderator 5 days ago edited on: 5 days ago	FWIW this patch seems to avoid the issue, but causes other problems. Maybe it can point to the bug: diff --git a/src/mame/midway/midxunit_m.cpp b/src/mame/midway/midxunit_m.cpp index 6b2507592d2..fe7664bdddb 100644 --- a/src/mame/midway/midxunit_m.cpp +++ b/src/mame/midway/midxunit_m.cpp @@ -309,9 +309,9 @@ uint32_t midxunit_state::dma_r(offs_t offset, uint32_t mem_mask) uint32_t result = 0; if (ACCESSING_BITS_16_31) - result \|= m_video->dma_r(offset * 2); - if (ACCESSING_BITS_0_15) result \|= uint32_t(m_video->dma_r(offset * 2 + 1)) << 16; + if (ACCESSING_BITS_0_15) + result \|= m_video->dma_r(offset * 2); return result; } @@ -320,7 +320,7 @@ uint32_t midxunit_state::dma_r(offs_t offset, uint32_t mem_mask) void midxunit_state::dma_w(offs_t offset, uint32_t data, uint32_t mem_mask) { if (ACCESSING_BITS_16_31) - m_video->dma_w(offset * 2, data & 0xffff); - if (ACCESSING_BITS_0_15) m_video->dma_w(offset * 2 + 1, data >> 16); + if (ACCESSING_BITS_0_15) + m_video->dma_w(offset * 2, data & 0xffff); }
No.23555 hap Developer 5 days ago	Try this?: void midxunit_state::dma_w(offs_t offset, uint32_t data, uint32_t mem_mask) { // word smearing if (mem_mask != 0xffffffff) data \|= (data & mem_mask) << 16 \| (data & mem_mask) >> 16; m_video->dma_w(offset * 2, data & 0xffff); m_video->dma_w(offset * 2 + 1, data >> 16); }
No.23557 Osso Moderator 5 days ago edited on: 5 days ago	Yes, it seems good to me with hap's patch.
No.23558 hap Developer 5 days ago	FWIW regression is probably from: https://github.com/mamedev/mame/commit/efed2ea2c21c9ac4dec8b3682efce8d731623abb I'll give AJR a poke, see if he thinks it's caused by a bug in his commit, or that 'word smearing' makes sense here.
No.23559 AJR Developer 5 days ago edited on: 5 days ago	I took a look at the TMS34020 manual, and I don't see any reference to word smearing. f1237705805a051f140fbd31330cad621f955e54 should fix this, at least for now.
No.23560 hap Developer 5 days ago edited on: 5 days ago	I was thinking a quirk in the DMA controller. Doesn't that commit reintroduce the 'U76 bad' error? edit: yes, it does And Osso mentions "other problems", or is it only this error?