-- This patch is a bug fix.

-- Description on the bug: using 8 bits/pixel framebuffer, all (splashscreen) bitmaps with palette colors whose RGB components differ in the 5 high bits in the blue byte and/or in the 6 high bits in the green byte do not display correctly.

-- How my patch fixes this bug: computing correctly the palette entries.

The C code fills a 16bit palette entry with 3 RGB bytes in the form

RRRR RGGG GGGB BBBB [....|....|....|....] masks: (R,G,B) (0xf800, 0x07e0, 0x001f) fedc ba98 7654 3210

The used bits are the highest 5 bits for the red and blue bytes and the 6 high bits for the green byte. This did not change neither in PowerPC nor in PXA in all cvs.sf.net revisions I've found.

The computation may be either: colreg = ((red & 0xf8) << 8) /* [RRRR|Rrrr|....|....] */ | ((green & 0xfc) << 3) /* [....|.GGG|GGGg|g...] */ | ((blue & 0xf8) >> 3); /* [....|....|...B|BBBB]bbb. */ or: colreg = ((red << 8) & 0xf800) | ((green << 3) & 0x07e0) /* versus " << 4" */ | ((blue >> 3) & 0x001f); /* versus " >> 0" */ I choose the latter form because it's the way you adopted.

-- A way of demonstrating that the patch actually fixes something is simply displaying such a bitmap (either splashscreen or simply as argument to "bmp display"). For instance obtaining a 640x480 8 bpp bitmap from http://www.dreamvideo.it/video/immagini/monoscopio.jpg: the original code displays greys as yellowish and blues as rubbish, while the patched code displays everything correctly.

-- (no new features)

-- CHANGELOG entry

* Corrected the computation of framebuffer palette for 8bpp lcd bitmaps Patch by Francesco Mandracci, 16 Sep 2005

-- CREDITS entry

N: Francesco Mandracci E: francesco.mandracci@primaelectronics.com P: ID=0xF6B25635 P: Fingerprint=4059 B2F9 E987 105D 00C8 3DEA 848C F56A F6B2 5635 D: Hacking PXA270

-- (no new board)

-- (no new configuration options)

-- The patch itself: ------------------------------------------------------------------- Index: common/lcd.c =================================================================== RCS file: /cvsroot/u-boot/u-boot/common/lcd.c,v retrieving revision 1.5 diff -p -u -r1.5 lcd.c --- common/lcd.c 4 Jul 2005 00:03:16 -0000 1.5 +++ common/lcd.c 16 Sep 2005 13:07:00 -0000 @@ -638,8 +638,8 @@ int lcd_display_bitmap(ulong bmp_image, bmp_color_table_entry_t cte = bmp->color_table[i]; ushort colreg = ( ((cte.red) << 8) & 0xf800) | - ( ((cte.green) << 4) & 0x07e0) | - ( (cte.blue) & 0x001f) ; + ( ((cte.green) << 3) & 0x07e0) | + ( ((cte.blue) >> 3) & 0x001f) ;

#ifdef CFG_INVERT_COLORS *cmap = 0xffff - colreg; -------------------------------------------------------------------

-- (single file)

-- (single modification)

-- ok for all boards

-- modifications are at the minimum

-- thinking that the memory footprint is affected in a reasonably way

-- in order to respect the 40Kb size limit I'm not sending the very bitmap file

Ciao Francesco Mandracci