dev builds
themes manual
device status forums
mailing lists
IRC bugs
dev guide

Rockbox mail archive

Subject: Re: fredwbauer: r30826 - in trunk: apps apps/gui firmware firmware/export

Re: fredwbauer: r30826 - in trunk: apps apps/gui firmware firmware/export

From: Jonathan Gordon <>
Date: Tue, 25 Oct 2011 10:19:48 +1100

On 25 October 2011 05:16, Fred Bauer <> wrote:
> If this is reverted, can it be delayed until after the release or
> until there is an alternative? (Jonathan said last night on IRC that
> he would work on it.) Several things were broken before r30826 and
> there are a considerable number of files that still expect
> font_get(FONT_UI) to provide the user selected font instead of the
> largest font index of the loaded fonts. Suppose a theme uses a small
> font with a very limited number of glyphs for a WPS numerical display
> and it happens to be the last font loaded by the skin parser. If all
> these files are not rewritten to use global_status.font_id[] the next
> display of  FONT_UI is going to cause the player to saw through the
> hard disk because of all the cache misses. (There can be three disk
> seeks for every cache miss.)

I didnt have time to work on this last night, hopefully I'll post a
patch tonight

> If r30826 is reverted, all the ~40 files that depend on
> font_get(FONT_UI) should be updated to use global_status.font_id[].
> FONT_UI should then be removed from font.h and considering the time
> that FONT_UI has been around, it would also be helpful to put a
> comment in the SVN commit and in font.h that FONT_UI is now deprecated
> so that contributors will not be publicly upbraided for trying to fix
> it when broken.

Relax, noone is being publicly unbraided. And yes, a note abou FONT_UI
being deprecated should have been done with the buflib commit

> Finally, bug hunting is not especially fun. What incentive does
> someone have to contribute to the project if all their ideas are
> promptly rejected without consideration? I'm aggravated. At this
> point, I really feel like I could write a plugin that cured cancer,
> AIDS, and hangovers and JD would reject it. Improve the v-keyboard? JD
> opposed. Allocate a sensible number of glyphs worth of font memory
> instead of over 1250 for an iPod Video (8818 for a Fuze!)? JD opposed.
> The same for skin fonts? JD accepted but then silently reverted. Save
> a glyphcache file per font? JD agreed with that...after he rejected
> the idea and I wrote an alternative. Now he's giving me grief because
> I didn't know about the unpublished deprecation of FONT_UI. Good
> times!

I have no idea what you're talking about. Yes I initially had
reservations with the glyph allocation patch, but you posted numbers
and I shut up (actually I thought i replied to the tracker but
apparently didn't). Bug fixes should never be done secretly, had you
asked on IRC I would have said that your bugfix wasnt ideal. Actually
thats probably a lie because I think the whole lifetime of this patch
was when I was offline, hence me bringing it up on the mailing list.

>> This message: [ Message body ] [ More options ]
>> Related messages: [ Next message ] [ Previous message ]
>> From: Jonathan Gordon <>
>> Date: Sun, 23 Oct 2011 16:31:50 +1100
>> Can this please be reverted? firmware/fonts.c should now know or care
>> about the ui font at all, and post buflib fonts it doesnt.
>> screen_access.c added a helper to set the font
>> (screens[screen].set_font() ) which should be being used by the
>> keyboard and lrcviewer. If those are still having issues then that
>> helper might need more work. (actually it appears keyboard.c needs to
>> use font_get() which has no helper to get
>> global_status.font_id[screen].
>> Ideally a font_get_height() function be added which (to the screen
>> api) to get the correct font height seen as that is what most calls to
>> font_get() actually care about.
>> On 23 October 2011 04:13, <> wrote:
>> > Date: 2011-10-22 19:13:33 +0200 (Sat, 22 Oct 2011)
>> > New Revision: 30826
>> >
>> > Log Message:
>> > Add functions font_set_ui() and font_get_ui(). The font returned by FONT_UI used to be fixed at zero but since buflib-fonts (r30589) can be different, depending on the order of loads and unloads. Fixes broken behavoir in virtual keyboard (FS#12336), lyrics player (FS#12306), and hopefully, FS#12337
>> >
> ...
>> > _______________________________________________
>> > rockbox-cvs mailing list
>> >
>> >
>> >
>> Received on 2011-10-23
> --
> Fred W. Bauer
Received on 2011-10-25

Page template was last modified "Tue Sep 7 00:00:02 2021" The Rockbox Crew -- Privacy Policy