[PATCH 8/8] video: Convert UTF-8 input stream to the 437 code page
Andre Przywara
andre.przywara at arm.com
Mon Jan 10 12:08:34 CET 2022
On Mon, 10 Jan 2022 09:18:45 +0100
Heinrich Schuchardt <heinrich.schuchardt at canonical.com> wrote:
Hi Heinrich,
> /On 1/10/22 01:56, Andre Przywara wrote:
> > The bitmap fonts (VGA 8x16 and friends) we import from Linux use the
> > 437 code page to map their glyphs. For U-Boot's own purposes this is
> > probably fine, but UEFI applications output Unicode, which only matches
> > in the very basic first 127 characters.
> >
> > Add a function that converts UTF-8 character sequences into the
> > respective CP437 code point, as far as the characters defined in there
> > allow this. This includes quite some international and box drawing
> > characters, which are used by UEFI applications.
> >
> > Signed-off-by: Andre Przywara <andre.przywara at arm.com>
> > ---
> > drivers/video/Makefile | 1 +
> > drivers/video/utf8_cp437.c | 169 ++++++++++++++++++++++++++++++
> > drivers/video/vidconsole-uclass.c | 6 +-
> > include/video_console.h | 9 ++
> > 4 files changed, 184 insertions(+), 1 deletion(-)
> > create mode 100644 drivers/video/utf8_cp437.c
> >
> > diff --git a/drivers/video/Makefile b/drivers/video/Makefile
> > index 8956b5f9b00..5f9823dff9e 100644
> > --- a/drivers/video/Makefile
> > +++ b/drivers/video/Makefile
> > @@ -14,6 +14,7 @@ obj-$(CONFIG_DISPLAY) += display-uclass.o
> > obj-$(CONFIG_VIDEO_MIPI_DSI) += dsi-host-uclass.o
> > obj-$(CONFIG_DM_VIDEO) += video-uclass.o vidconsole-uclass.o
> > obj-$(CONFIG_DM_VIDEO) += video_bmp.o
> > +obj-$(CONFIG_DM_VIDEO) += utf8_cp437.o
> > obj-$(CONFIG_PANEL) += panel-uclass.o
> > obj-$(CONFIG_DM_PANEL_HX8238D) += hx8238d.o
> > obj-$(CONFIG_SIMPLE_PANEL) += simple_panel.o
> > diff --git a/drivers/video/utf8_cp437.c b/drivers/video/utf8_cp437.c
> > new file mode 100644
> > index 00000000000..cab68b92b6e
> > --- /dev/null
> > +++ b/drivers/video/utf8_cp437.c
>
> A translation from Unicode to CP437 is needed in the FAT driver (but
> missing), in Unicode Collation Protocol and here in video. So this
> functionality should live in lib/charset.c.
> Please, have a look at efi_fat_to_str() in
> lib/efi_loader/efi_unicode_collation.c. We should avoid code duplication.
Ah, didn't know, thanks for the heads up! Happy to move the code in there,
and will also look for prior art.
> Maybe we should drop CP1250 support to make our effort simpler? No
> defconfig uses it.
I need to have a closer look.
>
> > @@ -0,0 +1,169 @@
> > +/*
> > + * Convert UTF-8 bytes into a code page 437 character.
> > + * Based on the table in the Code_page_437 Wikipedia page.
> > + */
> > +
> > +#include <linux/types.h>
> > +
> > +uint8_t code_points_00a0[] = {
> > + 255, 173, 155, 156, 7, 157, 7, 21,
> > + 7, 7, 166, 174, 170, 7, 7, 7,
> > + 248, 241, 253, 7, 7, 230, 20, 250,
> > + 7, 7, 167, 175, 172, 171, 7, 168,
> > + 7, 7, 7, 7, 142, 143, 146, 128,
> > + 7, 144, 7, 7, 7, 7, 7, 7,
> > + 7, 165, 7, 7, 7, 7, 153, 7,
> > + 7, 7, 7, 7, 154, 7, 7, 225,
> > + 133, 160, 131, 7, 132, 134, 145, 135,
> > + 138, 130, 136, 137, 141, 161, 140, 139,
> > + 7, 164, 149, 162, 147, 7, 148, 246,
> > + 7, 151, 163, 150, 129, 7, 7, 152,
> > +};
> > +
> > +uint8_t code_points_2550[] = {
> > + 205, 186, 213, 214, 201, 184, 183, 187,
> > + 212, 211, 200, 190, 189, 188, 198, 199,
> > + 204, 181, 182, 185, 209, 210, 203, 207,
> > + 208, 202, 216, 215, 206
> > +};
> > +
> > +static uint8_t utf8_convert_11bit(uint16_t code)
> > +{
> > + switch (code) {
> > + case 0x0192: return 159;
> > + case 0x0393: return 226;
> > + case 0x0398: return 233;
> > + case 0x03A3: return 228;
> > + case 0x03A6: return 232;
> > + case 0x03A9: return 234;
> > + case 0x03B1: return 224;
> > + case 0x03B4: return 235;
> > + case 0x03B5: return 238;
> > + case 0x03C0: return 227;
> > + case 0x03C3: return 229;
> > + case 0x03C4: return 231;
> > + case 0x03C6: return 237;
> > + }
> > +
> > + return 0;
> > +};
> > +
> > +static uint8_t utf8_convert_2xxx(uint16_t code)
>
>
> This is duplicate to include/cp437.h
But this is the other way around, isn't it? The above is from CP437 to
UCS-2, this one here is from UCS-2 to CP437 (and it's misnamed).
But I can certainly move it in there.
> > +{
> > + switch (code) {
> > + case 0x2022: return 7;
> > + case 0x203C: return 19;
> > + case 0x207F: return 252;
> > + case 0x20A7: return 158;
> > + case 0x2190: return 27;
> > + case 0x2191: return 24;
> > + case 0x2192: return 26;
> > + case 0x2193: return 25;
> > + case 0x2194: return 29;
> > + case 0x2195: return 18;
> > + case 0x21A8: return 23;
> > + case 0x2219: return 249;
> > + case 0x221A: return 251;
> > + case 0x221E: return 236;
> > + case 0x221F: return 28;
> > + case 0x2229: return 239;
> > + case 0x2248: return 247;
> > + case 0x2261: return 240;
> > + case 0x2264: return 243;
> > + case 0x2265: return 242;
> > + case 0x2310: return 169;
> > + case 0x2320: return 244;
> > + case 0x2321: return 245;
> > + case 0x2500: return 196;
> > + case 0x2502: return 179;
> > + case 0x250C: return 218;
> > + case 0x2510: return 191;
> > + case 0x2514: return 192;
> > + case 0x2518: return 217;
> > + case 0x251C: return 195;
> > + case 0x2524: return 180;
> > + case 0x252C: return 194;
> > + case 0x2534: return 193;
> > + case 0x253C: return 197;
> > + case 0x2580: return 223;
> > + case 0x2584: return 220;
> > + case 0x2588: return 219;
> > + case 0x258C: return 221;
> > + case 0x2590: return 222;
> > + case 0x2591: return 176;
> > + case 0x2592: return 177;
> > + case 0x2593: return 178;
> > + case 0x25A0: return 254;
> > + case 0x25AC: return 22;
> > + case 0x25B2: return 30;
> > + case 0x25BA: return 16;
> > + case 0x25BC: return 31;
> > + case 0x25C4: return 17;
> > + case 0x25CB: return 9;
> > + case 0x25D8: return 8;
> > + case 0x25D9: return 10;
> > + case 0x263A: return 1;
> > + case 0x263B: return 2;
> > + case 0x263C: return 15;
> > + case 0x2640: return 12;
> > + case 0x2642: return 11;
> > + case 0x2660: return 6;
> > + case 0x2663: return 5;
> > + case 0x2665: return 3;
> > + case 0x2666: return 4;
> > + case 0x266A: return 13;
> > + case 0x266B: return 14;
> > + }
> > +
> > + return 0;
> > +}
> > +
> > +uint8_t convert_uc16_to_cp437(uint16_t code)
>
> We should not duplicate efi_fat_to_str() but use a common function.
Definitely, I wasn't aware of this code, thanks for the heads up.
> > +{
> > + if (code < 0x7f) // ASCII
> > + return code;
> > + if (code < 0xa0) // high control characters
> > + return code;
> > + if (code < 0x100) // international characters
> > + return code_points_00a0[code - 0xa0];
> > + if (code < 0x800)
> > + return utf8_convert_11bit(code);
> > + if (code >= 0x2550 && code < 0x256d) // block graphics
> > + return code_points_2550[code - 0x2550];
>
> How about ÄÖÜß and other European letters?
Any "Übergrößengeschäft" should be covered by the international
characters section above, between 160 and 255. Do you miss anything in
particular?
> > +
> > + return utf8_convert_2xxx(code);
> > +}
> > +
> > +uint8_t convert_utf8_to_cp437(uint8_t c, uint32_t *esc)
> > +{
> > + int shift;
> > + uint16_t ucs;
> > +
> > + if (c < 127) // ASCII
> > + return c;
> > + if (c == 127)
> > + return 8; // DEL (?)
> > +
> > + switch (c & 0xf0) {
> > + case 0xc0: case 0xd0: // two bytes sequence
> > + *esc = (1U << 24) | ((c & 0x1f) << 6);
> > + return 0;
> > + case 0xe0: // three bytes sequence
> > + *esc = (2U << 24) | ((c & 0x0f) << 12);
> > + return 0;
> > + case 0xf0: // four bytes sequence
> > + *esc = (3U << 24) | ((c & 0x07) << 18);
> > + return 0;
> > + case 0x80: case 0x90: case 0xa0: case 0xb0: // continuation
> > + shift = (*esc >> 24) - 1;
> > + ucs = *esc & 0xffffff;
> > + if (shift) {
> > + *esc = (shift << 24) | ucs | (c & 0x3f) << (shift * 6);
> > + return 0;
> > + }
> > + *esc = 0;
> > + return convert_uc16_to_cp437(ucs | (c & 0x3f));
> > + }
> > +
> > + return 0;
> > +}
> > diff --git a/drivers/video/vidconsole-uclass.c b/drivers/video/vidconsole-uclass.c
> > index 420fd86f9ac..ca6e1a2620c 100644
> > --- a/drivers/video/vidconsole-uclass.c
> > +++ b/drivers/video/vidconsole-uclass.c
> > @@ -546,6 +546,7 @@ static int vidconsole_output_glyph(struct udevice *dev, char ch)
> > int vidconsole_put_char(struct udevice *dev, char ch)
> > {
> > struct vidconsole_priv *priv = dev_get_uclass_priv(dev);
> > + uint8_t cp437;
> > int ret;
> >
> > /*
> > @@ -587,7 +588,10 @@ int vidconsole_put_char(struct udevice *dev, char ch)
> > priv->last_ch = 0;
> > break;
> > default:
> > - ret = vidconsole_output_glyph(dev, ch);
> > + cp437 = convert_utf8_to_cp437(ch, &priv->ucs);
> > + if (cp437 == 0)
> > + return 0;
> > + ret = vidconsole_output_glyph(dev, cp437);
> > if (ret < 0)
> > return ret;
> > break;
> > diff --git a/include/video_console.h b/include/video_console.h
>
> This should go to include/charset.h.
>
> Best regards
>
> Heinrich
>
> > index a908f1412e8..f2d05e7f4e7 100644
> > --- a/include/video_console.h
> > +++ b/include/video_console.h
> > @@ -83,6 +83,7 @@ struct vidconsole_priv {
> > int escape_len;
> > int row_saved;
> > int col_saved;
> > + u32 ucs;
> > bool cursor_visible;
> > char escape_buf[32];
> > };
> > @@ -304,6 +305,14 @@ static inline int vidconsole_memmove(struct udevice *dev, void *dst,
> > return 0;
> > }
> >
> > +/*
> > + * Convert an UTF-8 byte into the corresponding character in the CP437
> > + * code page. Returns 0 if that character is part of a multi-byte sequence.
> > + * for which *esc holds the state of. Repeatedly feed in more bytes until
> > + * the return value returns a non-0 character.
>
> Please, follow the style described in
> https://www.kernel.org/doc/html/latest/doc-guide/kernel-doc.html#function-documentation.
> This will allow inclusion in the HTML documentation (see
> https://u-boot.readthedocs.io/en/latest/api/index.html).
>
Sure, I love and use kerneldoc, just missed that in that three-year old
patch ;-)
Cheers,
Andre
>
> > + */
> > +uint8_t convert_utf8_to_cp437(uint8_t c, uint32_t *esc);
> > +
> > #endif
> >
> > #endif
>
More information about the U-Boot
mailing list