All Projects

This is an attempt to achieve the goal once discussed in the forums and/or in irc. It should allow to not turn on the backlight on certain key presses. IIRC there is already a patch doing this. But a problem there was that the backlight is handled at the firmware level (as opposed to the app level) and hence it was not possible to take the context into account (whether we’re in the WPS or in a menu or…).

This tries to solve this problem. The approach is to provide a callback to the button code (firmware level) that would tell if the backlight should be turned on. The callback itself has access to the app level information (context).

The solution is not finished yet. TODOs:

1. Implement the backlight oracle (see the TODO mark in action.c)
2. Define the backlight properties for the key maps. It’s not done in this patch so it produces many warnings (unassigned struct member).

I nevetherless put up this patch to discuss the approach. Comments are welcome.

Here is my shot at what has been discussed on IRC a couple of days ago. Since I already provided a patch for reordering of items in the main menu (but it was rejected), I liked the idea because:

1. It’s more general (applies to all menus, not only the main one)
2. Menus are adjusted automatically to each user’s usage pattern
3. Even if not used for automatic reordering, it can be used for gathering statistics about the GUI usage and thus provide information for optimizing the menu structure

This is just the first quick try, but I hope the idea is clear: each menu item gets a unique ID. IDs are generated by macros, so no need to edit many source files; hopefully, there won’t be any name conflicts. If there will the line number (as in PP macro LINE) could somehow be incorporated into the ID. I just couldn’t figure how to do it. With lines incorporated, names become more unique but also more fragile.

Also, each item gets a counter for how many times the item was selected.

Then, before a menu is displayed, its items are reordered so that more frequently used items go to the top.

At shut down, the stats can be dumped to a file, and loaded again at startup.

There could/should also be settings telling:
- whether the stats should be gathered (maybe you’ve achieved your dream configuration want to freeze it?)
- whether the reordering should be made (useful e.g. for blind users)

TODOs:

1. Find a better place for the definition of GATHER_CALL_STATS so that this feature can be cut at compile time
2. Add something to the macros that would register each item’s stats struct in a global array. This array would then be used for dumping/loading
3. Implement dumping/loading the stats (now everything is initialized with zero)
4. Add the appropriate settings

Please don’t put this down immediately after you see “menu reordering.” Just live with the idea some time, and maybe you’ll like it. And I’d be glad if someone would help me with all this.

I’d definitely use it to reorder some items and would then freeze my stats once the desired and frequently used items are on the places where I’d like to find them.

This patch allows to change pitch not only by a certain amount of percents but also in halftone steps. Pitch changing mode is toggled by pressing MODE button while in the pitch screen.

Key mapping is only provided for H1xx since I don’t know which buttons are the best candidates for this on other players.

Also a small bug is corrected reg. nudging: if nudging would exceed the allowed range then it’s not executed. But on button release, the “denudging” was always executed – regardless whether the nudging was previously applied.

TODO:
- the active mode should somehow be displayed on the screen (an asterisk?)
- half tone pitching is not very accurate, the error is accumulated. But for simple pupose of transposing a piece it’s OK.

I’m very interested in your opinions about this patch.

First shot at nimbus-18 (my favourite shape). All basic glyphs have been designed, so the rest should be trivial. Markun, are you reading this?

This simplifies code, we don’t have to analyse whether it’s a .link file etc – each part has its own little task.

Also, two options are introduced:

1. In the link file, you can specify the display name for the entry. It should be separated from the path by the TAB. E.g.

/my/path/<TAB>My favourite songs

2. In the file, you can write

#Display last path segments=N

and then only the last N path segments will be displayed. This is handy if your music is in a deep folder.

Explicit name (with a TAB) has higher priority than N, i.e. if N=2 and you have

/aa/bb/cc/x.mp3<TAB>YYY

then the item will be displayed as YYY and not as bb/cc/x.mp3

I get some strange linker errors. Before I split the plugin into two everything worked OK.

Here is a patch that adds the ‘Startup Screen’ option to the playback menu (General Settings → Playback → Startup Screen). Possible values are:

- File Browser
- FM Radio (only on devices where it’s present)
- Playback (formerly known as ‘Resume on Startup’)
- Recording Screen (only on devices with recording capability, formerly known as ‘Start Rockbox in Recording Screen’)
- Last Active Screen

The options ‘Resume on Startup’ and ‘Start Rockbox in Recording Screen’ have been removed.

Known issues:

1. The ‘Last Active Screen’ feature is not implemented yet, it’s equivalent to ‘File Browser’ for now (see file tree.c) – since I don’t know where is the best place to track the active screen.

2. Version of the settings structure is increased, so save your settings before installing this patch.

3. I’m not completely sure whether the ‘FM Radio’ option is implemented correctly, I used the code for ‘Recording Screen’ as a template.

So it would do no harm if a developer familiar with Rockbox structure would throw a critical look at it!

Model: iriver H120, Build: 2006-09-01 17:59, Default settings

Installed the build (settings were reset since the settings version was changed)
Changed file view to ‘All’ Went to ./rockbox/fmpresets, chose a preset file → radio is activated
Changed to preset mode
Pressed NAVI → station list displays
* Pressed DOWN → * the cursor doesn’t move *
Pressed DOWN once more → cursor moves
Left station list screen
Pressed NAVI → station list displays
* Pressed UP → * the cursor doesn’t move *
Pressed UP once more → cursor moves

Since I like the game but mostly as an exercise in programming RB, I’ve coded up the plugin for reversi game. It should work on all targets with a bitmap display.

As moves are made, they are stored in the game history so that it’s possible to save the track. It should be also possible to choose the playing strategy for each player.

TODOs:

- Really implement the automatic playing strategies (take them from hinversi?)
- Implement loading/saving a game from/to a file
- Implement robot players (provisions are made but the main loop should be adjusted)
- Make nicer GUI - …

Anyone interested?

The contents of the ZIP file should be placed to the ‘plugins’ directory, and the patch should also be applied.

This patch makes the pitch setting permanent. You can set the pitch value that should be used after the DAP is switched on. This can also be changed in the pitch screen but the value set there is NOT saved. So if you set the pitch to 150% in the pitch screen but have it set to 100% in the sound settings, 100% will be used after the next power on.

Also, the initial mode for changing pitch (small steps/semitones) can be set.

Manual is also updated.

I changed the symbol HAVE_PITCHSCREEN to HAVE_ADJUSTABLE_PITCH since it better tells what we’re at IMHO.

I have some files with cyrillic characters in the tags. But I don’t like the unifont since it’s too thin for my taste. So I added cyrillic chracters to the my favorite font nimbus-14 (it’s thicker and hence easier to read). And while I was at it I also added some east european characters since they all look very similar. I have not added any chinese/korean/japanese glyphs.

There is another patch for nimbus-14 with cyrillic and chinese glyphs, but it was created automatically from some TrueType font so the symbols don’t look very nice. Here every symbol was “designed carefully.”

Here is make take to make the order of the entries in the main menu customizable. Only the order may be changed, key shortcuts etc. is not changed.

The solution is a bit clumsy (if not ugly). To change the order of the entries, create a text file named “main_menu.mnu” in the folder “/.rockbox”. In this file, place the following lines in the order you’d like to have the entries in the main menu:

bookmarks
files
database
now playing
settings
recording
fm radio
playlists
plugins
system
shutdown

The line with “fm radio” should only be present on DAPs with FM radio. The line with “shutdown” should only be present on DAPs with char cells.

An example for H120:

bookmarks
files
now playing
fm radio
settings
playlists
recording
database
plugins
system

If something goes wrong (file does not exist, unknown entries, bad number of lines etc.) the root menu is not changed.

This feature can be propagated to other menues, but I think it would be an overkill. But the main menu is used often and hence should be optimized as much as possible to the taste of every single user.

If this will be propagated to other menues, we could create a field in the menu structure to hold the “id” of the menu entry.

Comments are welcome.

Here is my take on what has been proposed on irc on Apr 5 2007, about 13:00 (according to the irc log).

This did reduce the size of the binary, at least on my H120.

Here’s the extract from the irc log:

13:04:02 crop I’d reorganise the wps_data struct.
13:04:24 crop 1. make progress_xxx vars unsigned
13:05:15 crop 2. intriduce struct subline with the fields start_token_idx, possibly num_tokens, line_type, and time_mult
13:06:04 crop Introduce struct line with the fields start_subline_idx, possibly num_sublines, subl_expire_time, and curr_subline
13:07:12 crop The idea is that all sublines are stored in one array and the fields point to that array. Now it’s assumed that each line can have up to 12 sublines which is rarely the case I’d argue and this wastes space
13:07:43 crop Tokens are already stored this way.
13:08:07 crop Structuring the data into structures would also make the code more readable IMHO 13:10:09 crop And one more thing: why do we need tokens for NOP and EOL? Why not just store the tokens that belong to the subline? In my proposal, start_subline_idx would say where the subline starts (index in the tokens array). The num_tokens isn’t strictly needed since a subline ends where the next begins. The same idea applies to lines.
13:10:58 crop This would simplify the drawing code since you have clear bounds and don’t have to watch out if we’ve reached EOL. This should be done at parsing time IMHO

Sansa e280 with FM tuner, r19871

My settings are:
- status bar: no
- the WPS used is very simple, it does contain the %we tag to switch the status bar on

But the status bar is not shown in the WPS

Here is the WPS I use:
%we

%ac%s%?ia<%ia|%d2>
%ac%s%?it<%it|%fn>

%al%pc%ac%pp/%pe%ar-%pr
%pb

%ac%s%?It<%It|%Fn>

Experienced with r17749.

I started playback by selecting the first file in a directory (so that the playlist was created ‘on the fly’). All files are mp3’s encoded at 128 …(don’t remember the unit).

There are 18 files in the directory (it’s ‘Chess’ by Andersson & Ulvaeus).

The last track was playing when playback stopped. There were still 1:28 in the track to play.

Player still reacted to some key presses (backlight went on). I tried to shut the player down. The splash ‘Shutting down…’ appeared but the player didn’t shut down.

It seems that this bug only shows up if the playlist is long enough since I didn’t experience it when I started a directory with the third from the end file.

Correctly handle files with long names containing non-latin characters (more precisely: files whose names, when utf8 encoded, require more than 255 bytes). This avoids player freeze when such files are present in the file system.

The patch does the following. When reading a entry for a file with long name, the name is first utf8-decoded into a temporary buffer. Then the check is made whether the content of the buffer would fit into the dir entry. If not, the long name is ignored and the short DOS name (which is OEM encoded) is used instead.

Beware: If this short name contains characters from the upper half of the code page, that name may appear with strange glyphs or empty character in the file browser. This is because Unicode characters 0×80 to 0x9F are control characters. They can’t be displayed.

This adds the printf format for pointers. Use %p or %P.

Example:

int *p;
printf(”Int value at %p is %d”, p, *p);

Also corrected the code indentation and factored out duplicated code.

This is a fix for the bug reported in  FS#7027 

As of now, file name extensions are shown (or not) based on the option “Show files” (All/Supported/etc.) This patch introduces an explicit setting for this.

I read about this in irc logs and liked the idea. The patch is really simple. It can be included without any side effects. But if the RB developers are not excited about this feature I wouldn’t insist that it MUST be included.

r19933: the lamp plugin can only be left via shutdown (long press of POWER), at least on my Sansa e200. All other buttons are ignored.

BTW: there is the patch FS#8934 that implements switching the lamp on and off. Why isn’t it committed?

The implementation of the force_backlight_on was not correct IMO. If the current setting value is OFF then calling the function wouldn’t have any effect.

This is a fix for the bug reported in  FS#9894 

This is a font I created for using in the FM radio screen. I contains only digits, the period and the default character that will be used to display all other letters. Ascent and descent are set to 0 so that there will be no vertical spacing.

Update: I added latin letters so that the font can be included into the font pack.

This patch implements the idea proposed and discussed in http://forums.rockbox.org/index.php?topic=24952.0

It adds a new option of how to create and maintain the list of the recent bookmarks (MRB): one per track. With this option, if a bookmark is added to the MRB, other bookmarks from the same playlist and the same track are not retained.

Comments are welcome.

I think it would be good if we could introduce such option for “normal” bookmark files too, i.e. to store only one bookmark per track in a bookmark list. But this would be matter of a separate patch.

This happens on my Sansa e280v1 with r31634-120108

I wanted to record something using the inbuilt microphone. The menu "RB Info" tells me there is space on the internal card:

132MB/7.45GB

When I start the recording, and then stop it while the WRS screen shows about 50KB, I get the message "Disk is full. Press PREV to continue."

After that, there is no recorded file. (The file should be created in the root directory of the internal card; WRS shows "Filename: /./R_MIC…")

But there IS space on the card! I could copy a file (via USB) of size 100 KB without problems.

So I think something is fishy in the recording code.

Is this somehow related to  FS#7721 ?

How to reproduce:

1. Go to the FM radio screen
2. Bring up the main menu
3. In the main menu, select ‘Now playing’ or ‘Resume playback’

Expected: FM screen is shown
Observed: DAP tries to start MP3 playback

This is a version of nimbus19 with some capitals made one pixel wider so that the gap between the legs of, say, H, has the same width as the legs’ thickness. Please try this and decide whether the font looks better that way (=more readable). If yes, we could provide a nimbus19w version of the font with wider capitals. It could be used on DAPs with large screen width.

The letters changed are: A, B, D, G, H, J, O, P, Q, R, U, V, X. That’s it for now.

If the font doesn’t look nice then this task should be abandoned.

This patch adds a new replaygain type “off” and eliminates the menu setting “Enable replaygain”. I hope I got it right

Following would be nice: you press a special key and get the list of all possible keys (in the current context) with their meaning. This would help if you don’t know exactly what you can do in a screen.

The comment field in the attached file should be “ÄÖÜäöüß” (German umlauts; the problem also occurs when e.g. cyrillic letters are used). But in “Show ID3 Info” screen it is shown as in the attached image. The font used is nimbus-14, it contains the appropriate glyphs (so it’s because of the bad font).

I don’t know whether this is a problem with ID3 reading or just displaying.

Observed in r14585.

Introduce a new WPS tag (%ms) which indicates whether te study mode is activated.

This patch lets the compiler to issue a compile error when the lamp plugin is built for a player without the backlight feature.

Without this patch, the whole file is #ifdef’ed out so that there will be no plugin entry point. The patch makes the compile process fail with a describing error message.

The patch does the following:

1. Reading of the file is now done char by char. This makes it not dangerous if the file is big
2. Code readability is improved via introducing of get_cell
3. Slight speed improvement in calculating the next generation (n[4] is taken from the previous generation)
4. A first step is made in decoupling of the grid size from the LCD size

When I start the simulator for sansa e200, the screen goes black after some seconds of inactivity. When I then press buttons, the screen does not turn on, i.e. it remains completely black.

Seen with r24075. I suspect that the bug was introduced in r24042 and its companions.

I've seen some comments about the reason and a possible fix in IRC but have not tried them out.

Seen with r25455 on Sansa e200 v1

Display the hotkey assignments (menu Settings→General Settings→Hotkey→View Hotkey Settings). The items in the displayed list do not scroll so that if the function name is long it’s not possible to read all of it.

Observed with the SVN revision 12728

How to reproduce:

1. Call up the main menu and go to ‘System’. At the top, the menu title ‘System’ is displayed with the user selected font
2. Position the cursor on ‘Version’ 3. Press RIGHT (RB logo with the version number at the bottom is displayed) and after one second press LEFT.

Now I’m back in the ‘System’ menu (as expected) but the title ‘System’ is displayed with the system (and not user selected) font

Implement the "anti karaoke" mode. In karaoke mode, only the differences between left and right channels are retained, the "common" part of the left and right channels is cut off.

What I'd like to have is a mode where only the voice "in the center" is left and voices that are present only in the left or only in the right channel are cut off (or significantly lowered).

As of now, %cP produces “am/pm” and %cp produces “AM/PM” (lowercase produces uppercase and vice versa). I think this should be changed.

Another proposal: let’s make the RTC tags “atomic”, i.e. let’s write “%ca %cb” instead of “%ca bc”. The parsing will become simpler as well as handling as well as they will be more like other tags (each %-tag produces a piece of information and stands on its own). There are no other tags that require a terminating char.

The patch adds the possibilty to move the cursor with shift in the other dimension. For example, if you move the cursor up and it wraps (jumps to the last row), the column is also changed (one to the left). Thus you can reach every calculator button by just keeping pressing up or down or left or right.

As of now, the button combination is only defined for Sansa e200/c200 (scroll wheel back and fwd) and only for moving up/down since that’s what the wheel does in menus and lists.

Also introduced constants for the number od button rows and columns. And renamed ‘global’ variables for the cursor positions. m and n are too short to be a global name IMHO.

The title says it all, I have nothing to add

…before the final ‘and’

The description of the tags is not very clear. One could e.g. understand that they can work as a switch for conditionals. This patch uses the “active” verbs instead of passive.

In the e200 v1 simulator, if I position the cursor on a long name in the file browser, it does not scroll. This happens in r20829. With r20828, everything works as it should. Hence r20829 must have introduced a bug.

I have not tried whether the same happens on target.

The patch does just what the title says (and nothing else). I think the new name more clearly states what the field means. I hope I catched all occurences.

In a separate patch I'll probably rename the field "mp3entry.embed_cuesheet" to "mp3entry.embedded_cuesheet".

The manual say how to change the menu but not how to restore it.