nanogui: Thread: Freetype on 4bpp display

I'm attempting to use Freetype2 (version 2.1.7) on an ARM720T
device with a 4bpp display. I'm not having luck getting any
output using a truetype font. I can use the built-in fonts
sucessfully, and I have verified that Freetype is functional
by dumping out glyphs. I guess the issue must be somewhere in
the blitter. Any clues would be appreciated.

-- sg

: I'm attempting to use Freetype2 (version 2.1.7) on an ARM720T
: device with a 4bpp display. I'm not having luck getting any
: output using a truetype font. I can use the built-in fonts
: sucessfully, and I have verified that Freetype is functional
: by dumping out glyphs. I guess the issue must be somewhere in
: the blitter. Any clues would be appreciated.

Check the config file and build and determine exactly which
4bpp driver you're using (drivers/fblin4*.c); your version
may not have the blit implemented.  Craig Anderson just
finished a new fully complete 4bpp driver implementation,
its available in CVS as drivers/fblin4.c.

Regards,

Greg

I'm using fblin4rev.c, which doesn't seem to have the new code.
What routine would I port in order to get freetype to work?

Thanks,
sg

> Check the config file and build and determine exactly which
> 4bpp driver you're using (drivers/fblin4*.c); your version
> may not have the blit implemented.  Craig Anderson just
> finished a new fully complete 4bpp driver implementation,
> its available in CVS as drivers/fblin4.c.
>
> Regards,
>
> Greg
>

> I'm using fblin4rev.c, which doesn't seem to have the new code.
What routine would I port in order to get freetype to work?

The linear4_blit routine.

The applyOpR code doesn't have to be inserted for font
drawing, but is used for and/xor/or with images.

Regards,

Greg

I did a quick sanity check by just using the fblin4.c code to
make sure that I could see some freetype output when using
the new code. The effect is the same as before: no output from
freetype.

I suspect it's something other than the blitter. I put in some
more debugging probes, and see that no matter what size I set
for the font, I always get back the same font metrics. For example,
I have set sizes of 10, 12, 16, and 20 and always get back the
same info:

maxwidth = 11
height = 13
baseline = 11
descent = 101058054
maxascent = 101058054
maxdescent = 101058054
linespacing = 101058054
firstchar =
lastchar = ÿ
fixed = FALSE
"Test1234567890ABCDEFGHIJKLMNOPQRSTUVWXYZ": width=269, height=13, base=11

The descent, etc info looks fishy. Anyway, I guess there is some
problem in realizing the font.

Regards,
sg

>> I'm using fblin4rev.c, which doesn't seem to have the new code.
> What routine would I port in order to get freetype to work?
>
> The linear4_blit routine.
>
> The applyOpR code doesn't have to be inserted for font
> drawing, but is used for and/xor/or with images.
>
> Regards,
>
> Greg
>

> I did a quick sanity check by just using the fblin4.c code to
make sure that I could see some freetype output when using
the new code. The effect is the same as before: no output from
freetype.

Good idea; its probably not the blitter then.  You should
still check that the blit routine is present in fblin4rev.c, 
though.



> I suspect it's something other than the blitter. I put in some
more debugging probes, and see that no matter what size I set
for the font, I always get back the same font metrics. For example,
I have set sizes of 10, 12, 16, and 20 and always get back the
same info:

I haven't tested the freetype 2 code, perhaps you should
try using the older freetype 1 libs, and see whether this
still happens.  Also, test with one of the .ttf files on the
distribution, I've checked them and they work with FT1.

Regards,

Greg

> I haven't tested the freetype 2 code, perhaps you should
> try using the older freetype 1 libs, and see whether this
> still happens.  Also, test with one of the .ttf files on the
> distribution, I've checked them and they work with FT1.

I'm pretty sure that Freetype2 is functioning correctly. I can
load the fonts from my app in both monochrome and 8-bit grey
mode. I've also dumped glyphs from the font realized inside
font_freetype2.c, and they are OK as well. Unfortunately, I
don't have enough RAM on my device or I would trace through with
gdb. As it is, I'm stuck with printf.

BTW, what is the proper procedure for submitting patches?

Regards,
sg

> I'm pretty sure that Freetype2 is functioning correctly. I can
load the fonts from my app in both monochrome and 8-bit grey
mode. 

The design of Microwindows is such that the code paths
should be identical except for the driver code.  All blit
bitmaps are stored in hw format.  If this works on the
desktop with X11, but not on the target, we likely have
either an endian or driver problem.


> BTW, what is the proper procedure for submitting patches?

Send them to me in "diff -urN" format, if possible.  I'd like
to see your glyph display debug mod, it looks nice.

Regards,

Greg

Here's the userspace program I've been using to test freetype.
I put similar code in font_freetype2.c to make sure that freetype
was working correctly in the nano-X shared lib. There, I changed
all printf to DPRINTF and set DPRINTF to use syslog.

I'll take another crack at freetype in a couple of weeks, but for
now I have embedded some bitmapped fonts suitable for my purpose.
I need to get a text editor, ebook reader and an HTML viewer
going on my device. Any recommendations for lightweight GUI apps
would be welcome (I have about 150K in which these should fit, so
C++ is out).

-- sg

> Send them to me in "diff -urN" format, if possible.  I'd like
> to see your glyph display debug mod, it looks nice.
>
> Regards,
>
> Greg
>