Subject: Re: VGA driver changes for machines without built-in font
To: None <firstname.lastname@example.org>
From: Izumi Tsutsui <email@example.com>
Date: 01/25/2003 02:39:10
Hi, thank you for your comments.
In article <20030124155158.GA271@krishna>
> ../../../../dev/ic/vga_raster.c:140: `MC6845_NREGS' undeclared here
> (not in a function)
> cc1: warnings being treated as errors
Ah, sorry, I forgot to add a diff for mc6845reg.h in the latest patch..
> ../../../../dev/ic/vga_raster.c:268: warning: initialization discards
> qualifiers from pointer target type
> To fix that, I had to remove const qualifier you added to definition of
> struct videomode vga_console_modes. And then the kernel was successfully
Ok, I'll revert this.
> + 256, /* numbhars */
> - scr->cursortmp.enc = WSDISPLAY_FONTENC_IBM;
> + scr->cursortmp.enc = scr->encoding;
> Please leave encoding related code as it is. I'm going to resume working
> on this soon.
Hmm. My intention of these changes is to use encoding
of default wsfont in vga_no_builtinfont case.
Without these changes, if the kernel uses wsfonts for default font
other than WSDISPLAY_FONTENC_IBM, it shows some garbages.
(for example, it fills screen with '@' (0x40) instead ' ' (0x20)
on initialization with FONT_VT220ISO8x16, whose firstchar is 0x20.)
I guess these implecit FONTENC_IBMs are used for the encoding of
the builtin font (which is default), but is it wrong?
In my patch, scr->encoding is set in vga_raster_setup_font().
> vga_reset() and its friends are not necessary for i386. How about making
> it optionally compile via 'options VGA_RESET_ON_BOOT' (or something with
> a better name)?
Yeah, I also wonder if it should be optional or not..
At the first time I added 'options VGA_NO_BUILTIN_FONT'
to disable all related changes, but I removed it because
it was also useful to use optional fonts even on i386.
But actually vga_reset() is required by only some machines,
so I'll add options VGA_RESET to enable it.
(or needs a better name? :-)
I've put the latest patch here: