All Projects

The daily builds page - http://www.rockbox.org/daily.shtml - shows only () for the revision since the repo has been switched to git.

The rev field of /sites/download.rockbox.org/daily/build-info is used for supplying the revision of the daily builds and is currently empty.

Since the build-info file is also used by the Rockbox Utility, there may also be some issues there, although I have not confirmed that.

The win32 cross-compiled sim does not build since the 5b4a6c4 commit. The previous commit (cf333a6) builds correctly.

The following warning/errors occur during the build of an e200 sim:

CC firmware/target/hosted/debug-hosted.c
firmware/target/hosted/debug-hosted.c: In function ‘debug_exit’:
firmware/target/hosted/debug-hosted.c:32: warning: implicit declaration of function ‘SetConsoleOutputCP’
firmware/target/hosted/debug-hosted.c: In function ‘debug_init’:
firmware/target/hosted/debug-hosted.c:37: warning: implicit declaration of function ‘GetConsoleOutputCP’
firmware/target/hosted/debug-hosted.c:41: warning: implicit declaration of function ‘atexit’

LD rockboxui.exe
e200-sim/firmware/libfirmware.a(debug-hosted.o): In function `debug_exit':
firmware/target/hosted/debug-hosted.c:32: undefined reference to `_SetConsoleOutputCP'
e200-sim/firmware/libfirmware.a(debug-hosted.o): In function `debug_init':
firmware/target/hosted/debug-hosted.c:37: undefined reference to `_GetConsoleOutputCP'
firmware/target/hosted/debug-hosted.c:40: undefined reference to `_SetConsoleOutputCP'
collect2: ld returned 1 exit status
make: *** [e200-sim/rockboxui.exe] Error 1

In bootloader builds starting with r30548, the bootloader USB mode does not work. This was tested with an e200v2 on win7 and ubuntu.

The last working build is r30546 and r30547 does not build due to build errors.

Tested on r30933 vx747 sim

When changing the "Line Padding in Lists" setting from Automatic to Off, the setting saved in .rockbox/config.cfg is:

list padding: auto

Changing back to Automatic causes the line to be removed from .rockbox/config.cfg. Manually saving the .cfg file and the setting that is saved is:

list padding: off

When setting the backdrop from the context menu, the new backdrop is not displayed. If the device is powered cycled, or is cycled through a USB connection, the backdrop will then be displayed.

This behavior started with r30475. Fails with current svn (r30844).

Steps to reproduce:

1. Boot device and clear settings
2. Go to Settings→Theme Settings and select Clear Backdrop - the backdrop is removed
3. Optional - power cycle device to see that backdrop is still gone
4. Go to Settings→General Settings→File View→Show Files and select All
5. Go to Files→.rockbox→backdrops and highlight cabbiev2.bmp
6. Access the context menu and select Set As Backdrop

You will return to the backdrops directory, but the new backdrop will not be displayed. Revisions prior to r30475 would display the new backdrop at this point.

The virtual keyboard spacing is incorrect when using a custom loadable keyboard layout along with a theme which loads multiple fonts. The spacing of the virtual keyboard appears to be based on the largest font loaded, not the user interface font.

The first build that this behavior is noticeable is r30716, but that is not necessarily the failing build. With builds between r30589 and r30715, bringing up the virtual keyboard with a custom keyboard layout causes a data abort. The last build that works properly is r30588.

This is failure can be reproduced in the uisimulator.

Steps to reproduce:
1. Add a custom keyboard file to your .rockbox directory (http://www.rockbox.org/wiki/pub/Main/LoadableKeyboardLayouts/default.kbd is the default keyboard as a .kbd file)
2. Go to your .rockbox directory and "play" the file.
3. Bring up the virtual keyboard (i.e. rename a file) and notice the character spacing.
4. Load a theme which uses multiple fonts - you can modify the cabbiev2.wps file to load an additional larger font by adding a line such as:

%Fl(2,32-Terminus.fnt)

Make sure that font exists in your .rockbox/fonts directory. Also, change one of the viewport definitions to use the new font. Use a font which is large enough to be noticeable.
5. Bring up the virtual keyboard again. The spacing on the virtual keyboard has changed to match the size of the larger font which was loaded instead of the user interface font.

I have attached screendumps taken from the e200 sim. The correct.bmp screendump is using the default.kbd and an unmodified cabbiev2. The incorrect.bmp screendump has a cabbiev2 with the above %Fl line added and one viewport modified to use font number 2.

Since r30475, I am seeing screen corruption of the statusbar and/or backdrop on e200v1.

Reproduction procedure:

1. From power off, connect device to computer
2. Safely remove device and disconnect USB cable
3. If corruption does not happen immediately on USB disconnect, press a button or rotate the wheel
4. Attempting to enter the View buflib allocs debug screen at this point causes the device to hardlock

The corruption will also occur after connecting USB after device is already powered on, but it happens less often.

r30471 does not have this problem. r30475 up to current svn will fail using this procedure.

I am having USB connection problems on my e200v1 since r30701: Add missing usb_drv_init(), fixes FS #12303

When connecting USB to win7 computer, a message is displayed that says the device should be connected to a USB 2.0 port (all ports are USB 2.0). The device takes a long time to mount (if it ever does) and will sporadically dismount/remount.

Reverting r30701 on current svn corrects the problem.

This problem is not present in r30700 or prior builds.

I have noticed at least two situations where an incorrect language string is displayed since the genlang changes introduced with r30481.

The first is on System → Rockbox Info
With r30480, the third info line displays: MSD: Not present
With r30481, the third info line displays: HD1 Not present

The second is on the Rolo screen
With r30480, SELECT = Yes
With r30481, PLAY = Yes

Tested on e200 with clean installs

Album art is not being displayed with any theme since r30499.

Reverting back to r30498 allows album art to display properly.

Tested on e200 and e200 uisimulator.

The uisimulator will crash when loading a Base Skin (sbs).

Tested on clean e200 uisimulator.

First failing build - r30478
Last working build - r30477
Latest build tested - r30495 (still fails)

To recreate the failure:
1. start the uisimulator from a clean install
2. go to Settings → Theme Settings → Base Skin
3. select classic_statusbar

The uisimulator crashes after printing numerous SDL_WaitEvent() error messages and finally a Segmentation fault message.

Running with gdb produces:

Program received signal SIGSEGV, Segmentation fault.
0x0807ef05 in skin_data_reset (wps_data=0x81bda78)

  at /home/mc2739/rockbox/apps/gui/skin_engine/skin_parser.c:1400

1400 struct gui_img *img = (struct gui_img*)list→token→value.data;

The win32 cross-compiled sim also fails, but no error messages are output (the program just closes).

Tested with r30337 and fails on e200v1 - does not fail on e200v2

r30308 is the first failing build
r30307 is the last working build

When using a theme which includes an FMS,

1. boot player
2. start music playback
3. exit wps (either menu or power will cause the failure)
4. select fm radio
5. exit fm radio (either menu or power will cause the failure)
6. select resume playback

The e200v1 data aborts with the following message:

Data abort
at 000071F4 (0)

The address is the same with different themes.

Playing an mp3 file with cuesheet causes the simulator (linux and w32) to crash. The same setup on the actual device works properly.

The position of the mp3 file in the playlist can change the results (i.e. proper operation).

The last working revision is r29784 and the first failing rev (that builds) is r29789

Linux sim r30322 fails with the following gdb message:

Program received signal SIGSEGV, Segmentation fault.
cuesheet_subtrack_changed (id3=0x81a3790)

  at /home/mc2739/rockbox/apps/cuesheet.c:377

377 || (cue→curr_track_idx < cue→track_count - 1

Fails on device and uisimulator.

W32-sim provides the following message:
WARNING, bad file name lacks slash: cabbiev2
read_bmp_file: can't open 'cabbiev2', rc: -1

After loading cabbiev2 theme from settings, the backdrop loads properly and is retained after reboot.

Each time a file is modified, a new entry for the file appears in the file list. The files are not duplicated if dircache is disabled. The file is not duplicated in the file system.

The display of radio art in FMS is not working properly since r29259 (works fine in r29258).

The following errors occur when entering the FM screen in preset mode while tuned to a preset with a radio art jpg available.

On e200v1, the radio art is sometimes displayed without being resized. Other times the radio art is not displayed, as if it does not exist. Sometimes a data abort crash occurs.

On e200v2, a data abort crash always occurs.

On the sim (e200v2), the sim crashes with a Segmentation fault.

On the win32 sim (e200v2), the sim crashes with no error message.

This bug may be related to  FS#11997 

The Fuzev2 will freeze after displaying the bootloader logo while the firmware logo is displayed.

This has been discussed on the forums here: http://forums.rockbox.org/index.php?topic=26055.0

Forum user Super_Dork_42 has sent me his failing Fuzev2 for testing, and I have found the following:

1. Builds later than r27502 (including v3.7 and v3.7.1) freeze on the firmware logo screen.
2. Build r27501 boots properly, but the FM Radio is not detected.
3. The OF diagnostic info shows HERMON/Si4702
4. si4700_read_reg(DEVICEID) returns 0xC834 not the 0x1242 expected for si4702

r28278

Fails on device and sim (e200v1 and e200v2)

1. start the resistor plugin
2. select “LED resistor calculator” 3. select “2v (Common red, orange)” 4. select “20mA - Most common…” 5. use the virtual keyboard to input a supply voltage - I used “5”

After leaving the virtual keyboard, a resistor image is drawn and plugin immediately returns to menu.

At this point, the scrollwheel does not work until button is pressed. The button is ignored and the scrollwheel starts to work. This scrollwheel/button behavior has been previously documented in several other Flyspray tasks: FS#8816,  FS#10097 , FS#10883 (and others)

Building w32 cross-compiled simulators in linux and cygwin do not build since r27873.

Pondlife opened a thread about the cygwin issue on the forum: http://forums.rockbox.org/index.php?topic=25575.0

The error produced with the linux w32 cross-compiled simulator is similar:

CC apps/language.c
LD chessclock.rock
LD credits.rock
/home/mc2739/rockbox/test-trunk/build/apps/plugins/plugin_crt0.o: In function `atexit’:
/home/mc2739/rockbox/test-trunk/apps/plugins/plugin_crt0.c:41: multiple definition of `_atexit’ /usr/lib/gcc/i586-mingw32msvc/4.2.1-sjlj/../../../../i586-mingw32msvc/lib/dllcrt2.o:/build/buildd/mingw32-runtime-3.15.2/build_dir/src/mingwrt-3.15.2-mingw32/dllcrt1.c:161: first defined here
collect2: ld returned 1 exit status
make: * [/home/mc2739/rockbox/test-trunk/build/apps/plugins/chessclock.rock] Error 1
make: * Waiting for unfinished jobs….

tested with r27834

Load a theme that includes a radio screen (fms)

Go to settings→theme settings→radio screen to list available fms files. The currently loaded fms file is not hilighted when entering the menu.

If you do the same with the base skin or while playing screen, the currently loaded sbs or wps file is hilighted.

This patch makes the sleep timer operational in the simulator. Current svn allows you to set the sleep timer, but the timer does not count down.

Instead of shutting down as on target, playback is stopped and you are returned to the menu.

The simulator will crash on start when it loads a theme which contains an SBS that has a comment line prior to the %Vi line and additional viewport definitions after the %Vi line. The same theme loaded after the simulator has successfully started does not cause a crash and operates properly.

This failure started with r27665. A simulator built with r27664 does not crash.

Tested on e200 v1/v2 linux/win32 simulators. The linux version spits out a bunch of "SDL_WaitEvent() error" messages and then gets a Segmentation fault. The windows version just shuts down the window. If run with –debugwps, in each case the "Loading '/.rockbox/wps/foo.sbs'" message is displayed before the crash. The similar messages for
the WPS and FMS are not displayed.

This simple SBS will cause the crash:

#
%Vi(-,0,0,176,194,-)
%V(0,194,-,-,-)
TEST

Substituting the %V with %Vd/%Vl as below will also cause the same crash:

#
%Vi(-,0,0,176,194,-)
%Vd(a)
%Vl(a,0,194,-,-,-)
TEST

Removing the comment line in each case will allow the simulator to start properly.

With a theme that includes an FMS with playbar, if you switch to the FM Radio right after playing a song with a cuesheet, the cuesheet tick marks from the song will still be displayed on the playbar in the FM screen.

r27563 tested on e200v2 target and sim (also fails on Version 3.6)

Steps to reproduce:

1. Boot Rockbox and clear settings.
2. Turn on cuesheet support.
3. Reboot to enable cuesheet support.
4. Load a theme that includes an FMS with a playbar.
5. Start playing a song that has a cuesheet.
6. Return to the main menu.
7. Start the FM Radio.
8. Notice the tick marks on the playbar.

If you leave the radio, play a song without a cuesheet and then return to the radio, the tick marks will be gone.

The Clip+ manual is currently using the images of the clip. Since the clip+ has now hold switch, this causes confusion for users who see an image with a hold switch on the side.

Additionally, there is now mention of the key sequence for the soft-hold function (home+select).

r26513

Tested and fails on e200v2 and e200v2 sim

There seems to be a problem with the %cs tag when certain screens are switched.

To reproduce the problem, I used the kleanew3 theme. This theme was disceussed in the forum: http://forums.rockbox.org/index.php?topic=24894.0 and is available here: http://www.mediafire.com/file/mzt0mqrwdhj/kleanew3.zip

I am attaching screen dumps to show correct and incorrect operation.

The steps to reproduce the problem:
1. Clear settings.
2. Set FM Region.
3. Add an FM Preset file and load it.
4. Install the kleanew3 theme. The theme requires 15-Adobe-Helvetica and 24-Terminus-Bold which are available in the font pack.
5. Switch to the kleanew3 theme.
6. Verify correct theme operation by going to the FM Radio, switching to the main menu and switching back to the radio.
7. Stop the radio (power button on e200).
8. While FM Radio is highlighted in the main menu, enter the context menu an reload the FM Preset file.
9. When the screen switches to the FM Radio, the %cs tag seems to be confused as items normally in the FM Screen are missing, and items from the menu are shown.

Stopping and restarting the radio will correct the problems.

Select a theme with custom FMS and then select a theme without custom FMS.

The custom FMS is still shown when going to the radio screen.

r26429 and later

Tested on e200v2 with inbuilt FM screen and custom FMS

A marker only appears for the first preset.

After playing music, the marker is no longer visible

When connecting e200v1 to USB with dircache enabled the device will freeze up before going to the USB screen. This does not happen on every connection but on the majority of connections. The user interface is frozen, but the backlight will still go off after the timeout and turn on with a button press. If dircache is turned off, the freeze does not occur.

Works fine with r26221 and fails with r26222

Only tested on e200v1 - may also fail on other players

Steps to reproduce:
1. Install r26222 build
2. Reset settings on boot
3. Turn on dircache
4. Reboot
5. Insert USB
6. If USB screen appears, safely remove and repeat step 5

In the PDF Manuals, Appendix B, WPS Tags the first WPS tag in each of the tables is shifted to the right by one character.

The 3.5 release manuals look correct.

Release 3.5 - current

The %Sx tag added in r22837 is not documented in the manual or on the CustomWPS wiki page.

I am seeing high dircache stack usage after USB connection on my e280v1.

I tested as far back as r24419 and up to current (r24794).

Steps to reproduce:

1. Clear all settings on boot
2. Check dircache usage on debug screen
3. Attach USB cable
4. Eject or safely remove
5. Remove USB cable
6. Check dircache usage on debug screen
7. Enable dircache
8. Reboot
9. Check dircache usage on debug screen
10. Attach USB cable
11. Eject or safely remove
12. Remove USB cable
13. Check dircache usage on debug screen

With r24419, the stack usage for steps 2, 6, 9, and 10 were 2%, 3%, 2%, and 82%. This 82% level was also present after rebooting.

With r24657, there was an increase in stack usage. The levels recorded were: 2%, 3%, 2%, and 93%.

With r24708, there was another increase in stack usage. The levels recorded were: 4%, 7%, 4%, and 97%.

Current svn gives the same values as those recorded for r24708.

As I have enabled other features in Rockbox, I have started to see “Stkov dircache (0)” panics. These panics continue after reboot and require editing the config.cfg file in USB mode to disable dircache. After disabling dircache, it will reboot without problems.

wav playback is broken since r24689 - it works with r24688

tested on e200v1 and e200v2 - target and sim

when the wav file in a playlist, a codec failure splash appears and then the next song is played

when the is played as the only file in a directory, no splash is displayed, you are just returned to the file browser

With dircache disabled, if you are in the file browser and insert or remove the microSD card it will consistently update the list by adding <microSD1> on insert and removing it when the card is removed.

With dircache enabled, the list is not consistently updated.

Tested with r24224 and as old as r22000 on e280v1
Tested with r24224 on e260v2

Process to reproduce:
1. Boot and clear all settings
2. Enable dircache setting
3. Power off and reboot
4. Enter the file browser by selecting Files on the Main Menu
5. Insert and remove microSD card and observe the Files list

If the microSD card was not inserted on boot, the list is not updated on the first insert. After the card is removed, subsequent inserts/removals seem to update properly.

If the microSD card was inserted on boot, the list is not updated on the first removal. Upon insert, the list may or may not update. The updates are very inconsistent on subsequent inserts/removals. If you exit the file browser and then reenter, the correct state is shown.

The website has several dead links since the target rename.

Rockbox Manual Install - http://www.rockbox.org/download/byhand.cgi 1. iAudio M3 PDF manual - http://download.rockbox.org/release/3.4/rockbox-iaudiom3-3.4.pdf 2. iPod Mini 1st gen PDF manual - http://download.rockbox.org/release/3.4/rockbox-ipodmini1g-3.4.pdf 3. iPod Mini 2nd gen PDF manual - http://download.rockbox.org/release/3.4/rockbox-ipodmini1g-3.4.pdf 4. SanDisk Sansa e200R - http://download.rockbox.org/release/3.4/rockbox-sansae200r-3.4.zip (e200r uses e200 release)
5. SanDisk Sansa e200R English voice - http://download.rockbox.org/release/3.4/sansae200r-3.4-english.zip 6. Toshiba Gigabeat F/X - http://download.rockbox.org/release/3.4/rockbox-gigabeatfx-3.4.zip 7. Toshiba Gigabeat F/X PDF manual - http://download.rockbox.org/release/3.4/rockbox-gigabeatfx-3.4.pdf 8. Toshiba Gigabeat F/X English voice - http://download.rockbox.org/release/3.4/gigabeatfx-3.4-english.zip

Rockbox - Daily Builds - http://www.rockbox.org/daily.shtml 1. SanDisk Sansa e200R (no latest link)
2. SanDisk Sansa e200R old - http://www.rockbox.org/dl.cgi?bin=sansae200r (not really a dead link)

It seems like the build server is getting the builds mixed up from time to time. There is a report in the forum about the r24112 e200v2 build actually being an ipod build - http://forums.rockbox.org/index.php?topic=14064.msg159849#msg159849 - the downloaded file (rockbox-sansae200v2.zip) is actually an ipod video build. This appears to be the only download that is not correct for this build.

This has happened in the past at least one other time. The time I remember was talked about in IRC and had a number of files that were incorrect.

Skipping to the previous track without cuesheet support enabled works as follows:
If track elapsed time is less than 3 seconds, skip to previous track
If track elapsed time is 3 seconds or more, skip to the start of current track

Skipping to the previous track with cuesheet support enabled currently works as follows:
If first segment elapsed time is less than 3 seconds, skip to previous track
If first segment elapsed time is 3 seconds or more, skip to the start of first segment
In any other segment after the first, skip to the start of the previous segment

This patch makes skip previous consistent on a track with cuesheet support. If skip previous is pressed in any segment before 3 seconds have elapsed, skip to the previous segment (or previous track if in the first segment). If skip previous is pressed in any segment with 3 or more seconds elapsed, skip to the start of the current segment.

Since the target rename commit (r23883), the nightly manual build process will attempt to build manuals for all “usable” ports (stable or unstable status). This causes problems since some of the unstable ports do not have manuals which complete the build process. The manual build for these targets produces the message “! LaTeX Error: File `platform/xxx.tex’ not found.” and waits for a user response.

I propose adding “manual ⇒ ‘none’,” to the hash array in tools/builds.pm and appropriate logic to www/manual.pl to bypass these ports.

I assume that the server that does the manual builds will also need some changes.

The ports in question are:
cowond2
mrobe500
samsungyh820
samsungyh920
samsungyh925
sansae200r

Since the sansae200r uses the sansae200 manual, I have already committed a change to tools/builds.pm for that (r23933). I have still listed it here in case further changes are necessary on the manual build system.

When recording to the microSD card on an e200v2 the message “The disk is full. Press PREV to Continue.” is splashed when the recording is stopped. The recording may or may not exist when you look on the microSD card. If the recording does exist, it has been truncated. The microSD card had approximately 100mb available.

Recording on the internal storage works fine. When recording in mp3 mode, the file was 1k larger than shown on the recording screen.

Rockbox build = r23746

Steps to reproduce:
1. Create a directory for the recordings on the microSD card.
2. In the file browser, highlight this directory and bring up the context menu.
3. Select “Set As Recording Directory” 4. Next, I went to Settings → Recording Settings → Format and chose “MPEG Layer 3” but the default PCM Wave also caused the same failure.
5. I started FM Radio and through the context menu started recording (MIC recording also fails)
6. On the recording screen, press the play button to actually start recording.
7. After some time, press the play button to pause recording. Make note of the size.
8. Press the power button to save the recording.
9. The disk full splash appears.
10. Press PREV to clear the splash.
11. Exit the context menu and the FM Radio.
12. Browse to your recording directory. If the file exists, check the properties. It will be smaller than the size noted on the recording screen.

Here are some of the recording size to file size comparisons I noted:
80.4 → 79k
106k → 95k
508k → 493k
243k → 238k

This patch prevents power switch activation when coming out of hold by not allowing the power GPIO to be read until 1 second has passed since the hold switch has been released.

This bug was introduced with r23258 (works with r23257)

Steps to reproduce
1. hold record button while powering on
2. e280 locks up while displaying the verbose bootloader screen.

I commented the splashf() on line 538 in apps/main.c and it did not lock and did reset the settings.

This patch cleans up the pictureflow manual button map. It also adds iAudio M3 remote control mappings and allows the manual to build again.

Rbutil v1.2.3 will not autodetect e200v2 in win7.

It will autodetect e200v2 in winxp.

It will autodetect e200v1 in win7.

This patch adds an item to the calendar menu to allow switching the display style between the default European style (first day of week is Monday) and the US style (first day of week is Sunday).

Manual patch is also included.

r22365 to current (r22588)

does not occur with r22364

tested on e200, e200v2, e200v2 sim

tested with clean build and settings reset

Steps to reproduce:
1. add gameboy rom to dap
2. select gameboy rom to start rockboy
3. press power button to access rockboy menu
4. select Load Game or Save Game

It appears that the status bar is disabled in all plugins with r22365, except the status bar appears over top of the first entry in the load or save game list of the rockboy plugin. The status bar may not appear immediately, but will appear after a refresh/update. With r22364, the status bar is not disabled in all plugins, and does not overwrite the first entry of the load and save game lists of the rockboy plugin.

I do not have a concern with the status bar being disabled in the plugins, but it should not be overwriting the load/save game lists.

Beginning with r22418 the volume changing tag %mv does not work properly in the Wide Cabbie theme.

This theme displays a large volume graphic and volume text when the volume changes. Prior to r22418, this worked properly. Since r22418, the graphic is not displayed, but the text is displayed.

The graphic is preloaded prior to being displayed by the following line:

%xl|G|bigvolume-176x220x16.bmp|52|36|10|

The conditional to display the viewports is:

%?mv<%Vdh%Vdi|%?Fn<%?C<%Vdz%Vdd%Vde%Vdf|%Vda%Vdc>|%?C<%Vdz%Vdd%Vdg|%Vdb»>

And the viewport definitions:

%Vl|h|0|20|-|110|-|-|-|
%?pv<%xdGa|%xdGb|%xdGc|%xdGd|%xdGe|%xdGf|%xdGg|%xdGh|%xdGi|%xdGj>
%Vl|i|0|120|-|57|-|-|-|
%acVolume: %pvdb

I have attached a screendump of the missing graphic.

r22225
tested on e200 and e200v2

Power on device
select theme with backdrop such as default cabbiev2
power off device
"Shutting down…" splash is displayed over backdrop

Power on device
select theme with WPS backdrop such as default cabbiev2
resume playback
power off device
"Shutting down…" splash is displayed over WPS backdrop

This is also noticed if a panic occurs. This becomes a problem with the default cabbiev2 WPS. The second line of the panic message overlays the top yellow "NOW PLAYING" bar and is unreadable.

Closed tasks  FS#5222  and  FS#9098  seem to describe the same problem during shutdown, but not the problem during a panic.