upd7220.cpp: make a bitmap partition of len==0 to be max size, fixes pc98 madoum1-2-3 [Angelo Salese, sampson] (#8009)

* upd7220.cpp: make a bitmap partition of len==0 to be max size, fixes pc98 madoum1-2-3 [Angelo Salese, sampson]

* pc98.xml: consistency DAC1BIT tag, update dbuster QA note

hash/pc98.xml

@@ -7286,7 +7286,7 @@ only have some part of Windows file and a Video driver(CLGD?).
 		</part>
 	</software>
 
-	<!-- boot OK on RS class, on intro uses PCM1BIT and some text doesn't seem to be drawn? -->
+	<!-- boot OK on RS class, on intro uses DAC1BIT and some text doesn't seem to be drawn? -->
 	<!-- black screen after Cocktail Soft logo on VM class (btanb?) -->
 	<software name="bishotsu" supported="partial">
 		<description>Bishoujo Tsuushin - Chat no Susume</description>
@@ -7603,7 +7603,7 @@ only have some part of Windows file and a Video driver(CLGD?).
 		</part>
 	</software>
 
-	<!-- This game is supposed to play sound effects through the beeper, but in MAME it just outputs a constant beep (PCM1BIT) -->
+	<!-- This game is supposed to play sound effects through the beeper, but in MAME it just outputs a constant beep (DAC1BIT) -->
 	<software name="blockouta" cloneof="blockout" supported="partial">
 		<description>Block Out (Patched?)</description>
 		<year>1991</year>
@@ -16327,7 +16327,7 @@ only have some part of Windows file and a Video driver(CLGD?).
 		</part>
 	</software>
 
-	<!-- Sprites leave "trails" on screen -->
+	<!-- boot OK, on VM class and RS class sprites leave "trails" on screen (GRCG or EGC regression) -->
 	<software name="dbuster" supported="partial">
 		<description>Dragon Buster</description>
 		<year>1989</year>
@@ -28336,8 +28336,11 @@ only have some part of Windows file and a Video driver(CLGD?).
 		</part>
 	</software>
 
-	<!-- no bitmap GFXs during gameplay -->
+	<!-- all three Madou Monogatari sports a "Madou Monogatari 1-2-3" title screen -->
-	<software name="madoum1" supported="no">
+	<!-- verify if these were really released as stand alone. -->
+
+	<!-- boot OK, uses DAC1BIT samples -->
+	<software name="madoum1" supported="partial">
 		<description>Madou Monogatari 1</description>
 		<year>1991</year>
 		<publisher>コンパイル (Compile)</publisher>
@@ -28357,8 +28360,8 @@ only have some part of Windows file and a Video driver(CLGD?).
 		</part>
 	</software>
 
-	<!-- no bitmap GFXs during gameplay -->
+	<!-- boot OK, uses DAC1BIT samples -->
-	<software name="madoum2" supported="no">
+	<software name="madoum2" supported="partial">
 		<description>Madou Monogatari 2</description>
 		<year>1991</year>
 		<publisher>コンパイル (Compile)</publisher>
@@ -28378,7 +28381,8 @@ only have some part of Windows file and a Video driver(CLGD?).
 		</part>
 	</software>
 
-	<software name="madoum3">
+	<!-- boot OK, uses DAC1BIT samples -->
+	<software name="madoum3" supported="partial">
 		<description>Madou Monogatari 3</description>
 		<year>1991</year>
 		<publisher>コンパイル (Compile)</publisher>

src/devices/video/upd7220.cpp

@@ -3,6 +3,7 @@
 /**********************************************************************
 
     Intel 82720 Graphics Display Controller emulation
+	also known as NEC uPD7220
 
 **********************************************************************/
 
@@ -22,7 +23,7 @@
     - A5105 has a FIFO bug with the RDAT, should be a lot larger when it scrolls up.
       Can be fixed with a DRDY mechanism for RDAT/WDAT;
     - Some later SWs on PC-98 throws "Invalid command byte 05" (zettmj on Epson logo),
-      actual undocumented command to reset something?
+      actual undocumented command to reset something or perhaps just check if upd7220 isn't really a upd72120 instead?
 
     - honor visible area
     - wide mode (32-bit access)
@@ -1730,18 +1731,29 @@ void upd7220_device::update_graphics(bitmap_rgb32 &bitmap, const rectangle &clip
 
 		if(im || force_bitmap)
 		{
-			if(area >= 3) // TODO: most likely to be correct, Quarth (PC-98xx) definitely draws with area 2. We might see an area 3 someday ...
+			// according to documentation only areas 0-1-2 can be drawn in bitmap mode
+     		// PC98 Quarth definitely needs area 2 for player section.
+			// TODO: what happens if combined area size is smaller than display height?
+			// documentation suggests that it should repeat from area 0, needs real HW verification (no known SW do it).
+			if(area >= 3)
 				break;
+			
+			// PC98 madoum1-3 sets up ALL areas to a length of 0 after initial intro screen.
+			// madoum1: area 0 sad==0 on gameplay (PC=0x955e7), sad==0xaa0 on second intro screen (tower) then intentionally scrolls up and back to initial position.
+			// Suggests that length should be treated as max size if this occurs, this is also proven to be correct via real HW verification.
+			// TODO: check if text partition do the same.
+			if (len == 0)
+				len = 0x400;
 
 			if(!interlace)
 				len <<= 1;
 
 			for(y = 0; y < len; y++)
 			{
-				/* TODO: again correct?
+				// TODO: not completely correct, all is drawn half size with real HW tests on pc98 msdos by just changing PRAM values.
-				         Quarth (PC-98xx) doesn't seem to use pitch here and it definitely wants bsy to be /2 to make scrolling to work.
+				// pc98 quarth doesn't seem to use pitch here and it definitely wants bsy to be /2 to make scrolling to work.
-				         Xevious (PC-98xx) wants the pitch to be fixed at 80, and wants bsy to be /1
+				// pc98 xevious wants the pitch to be fixed at 80, and wants bsy to be /1
-				         Dragon Buster (PC-98xx) contradicts with Xevious with regards of the pitch tho ... */
+				// pc98 dbuster contradicts with Xevious with regards of the pitch tho ...
 				uint32_t const addr = ((sad << 1) & 0x3ffff) + ((y / (mixed ? 1 : m_lr)) * (m_pitch << (im ? 0 : 1)));
 				for(int z = 0; z <= m_disp; ++z)
 				{