07685: sonic, cabalus, possibly others using upd4701a: mapping joystick to trackball causes input to go haywire

ID	Category [?]	Severity [?]	Reproducibility	Date Submitted	Last Update
07685	DIP/Input	Minor	Always	Jun 12, 2020, 09:42	Nov 3, 2021, 08:47

Tester	oboewan42	View Status	Public	Platform	MAME (Official Binary)
Assigned To	cuavas	Resolution	Fixed	OS	Windows 10 (64-bit)
Status [?]	Resolved			Driver
Version	0.221	Fixed in Version	0.238	Build	64-bit
		Fixed in Git Commit	d29287e	Github Pull Request #

Summary	07685: sonic, cabalus, possibly others using upd4701a: mapping joystick to trackball causes input to go haywire
Description	Moving an analog joystick axis that's mapped to a trackball axis - specifically one connected to a uPD4701A - causes the input to go haywire, even when sensitivity is set to 1. Mouse input appears to work fine.
Steps To Reproduce	Map an analog stick to a trackball (MAME is set up this way by default if you have a controller with an analog stick), then move the stick and watch the analog controls go haywire. Visible in-game, in the test menu in sonic, and in the "Analog Controls" menu. Compare with version 0.185.
Additional Information	I vaguely remembered being able to play SegaSonic perfectly fine with an analog stick in an older version of MAME, so I checked for a regression. Last working version was 0.185, it broke in 0.186. Looking in the whatsnew it looks like there was a rewrite of the code for the uPD4701A around that time. Tested in sonic and cabalus, both exhibit this issue, haven't had a chance to look at anything else (what other games use it?) Centipede works fine, so it seems to be limited to that chip.
Github Commit

Flags
Regression Version	0.186
Affected Sets / Systems	sonic, cabalus, possibly others using upd4701a

Attached Files

No.17745 Tafoid Administrator Jun 14, 2020, 19:00	I don't have a decent non-mouse analog device to use for testing at this time, but I've acknowledged for now.
No.17752 AJR Developer Jun 18, 2020, 17:47	This may be a problem with not accounting for PORT_RESET, which the uPD4701 emulation now depends on. Other sets to test include arknoid2, ridleofp and simpbowl.
No.17753 hap Developer Jun 18, 2020, 19:10	arknoid2 and ridleofp have the same issue (did not test simpbowl)
No.19409 AJR Developer Nov 1, 2021, 12:36	I wish there were some way to reproduce this without additional hardware.
No.19415 cuavas Administrator Nov 3, 2021, 08:47	The symptom is “fixed” but the real problem is that PORT_RESET is cancer and should be killed. It simulates an incremental encoder where the counter is automatically reset by the vertical blanking signal of the first screen in the system (or the 60Hz virtual first screen update for screenless systems). This is almost never correct. The systems/devices need to be fixed so they reset their up/down counters at the correct time.