All Projects

When a X window which is displaying GLES2 render output is:
- Either dragged out of the visibile screen region (?)
- Is made larger than it was at the creation time of the GLES2 context
then:
- The performance will drop significantly as the EGL buffer swaps won't occur as fast as before,
- The render output will be corrupted and flicker
- In rare occasions crashes the application or locks up the entire system (which forces you to remove the battery)

The problem can not be solved without restarting the application (Possibly forcing players to complete the same goals again as they were unable to reach a savepoint or similar things).

This was confirmed in Mupen64Plus and the PowerVR GLES Samples and some of my own projects (which are, however, based on the PowerVR samples).
As I didn't see any other reports about this problems (but had it confirmed by other Pandora users too) I would expect that it can be fixed by using a newer X, kernel or combination of them as other OMAP users would be effected too.

When booting the pandora-OS from sd-card by using the official tar, you will end up with your sd-card being mounted twice: once as / and once as /media/<label>. I solved this by adding "/dev/mmcblk0" to "/etc/udev/mount.blacklist". Maybe this should be added to the default tarball? I'm not sure as it might be used in other ways maybe (beside SD-card-installation)?

Background: I ran "updatedb" as root and have been notified by a message that there was a mount-loop. After adding the above modification the system still booted fine and "updatedb" ran nicely without a warning.

I am running PandoraOS from sd-card installed with the official tarball "pandora-rootfs.tar.bz2".

I noticed that the control/keyboard-mapping in vi does not work correctly after switching to root (for example sudo su-). I solved this by copying over my users .vimrc to root-homedir. This should already be included in the tarball.

The patch implementing screen blanking uses the /usr/pandora/scripts/op_lid.sh script to control LCD backlit:
http://git.openpandora.org/cgi-bin/gitweb.cgi?p=openembedded.git;a=blob;f=recipes/xorg-driver/xf86-video-omapfb/pandora_backlight_dim.patch;h=4a0b8432d083c6e7297192a2920c7a777ae3d098;hb=16f3fff44323bd6b4349d909667b51a2fef3f278

The same result can be achieved by using the sys interface:
/sys/devices/platform/twl4030-pwm0-bl/backlight/twl4030-pwm0-bl/bl_power

This approach makes driver independent on particular distribution. The proposed patch is attached.

Try clearing shortcuts in Start->Settings->Window Manager->Keyboard. That doesn't work (for me).
This is a problem because these shortcuts are blocking key-combination in apps, like for example the keymapper in dosbox (CTRL+F1) or refreshing content (CTRL+F5). I tried manual editing or even deleting the configuration file, but changes are not recognized and the file recreated on reboot.

Please see this for GTK apps : http://library.gnome.org/devel/gdk/stable/gdk-Input-Devices.html
And see this thread : http://boards.openpandora.org/index.php?/topic/372-xournal-pdf-annotations/page__gopid__6836#entry6836

Ok, I just want to say that there is something missing about the font on the base image. I cant get what (at least for now)

Here is an issue : http://www.gp32x.com/board/index.php?/topic/54282-getting-vnc-viewer-to-work/

With my current Ogre build I get :
Warning: Unable to load any usable ISO8859 font
Error: Aborting: no font found

I've googled but so far no solution :(

That still need to be configurable, but chris_c does a good job I'ld like to see integraded in the main image :
http://www.gp32x.com/board/index.php?/topic/58794-a-script-to-shutdown-after-charging/
and :
http://chrisc.bedroomcoders.co.uk/?p=77

http://www.wader-project.org
http://trac.wader-project.org
http://trac.wader-project.org/wiki/Developers

this would be really usefull to monitor reception/network of mobile broadband dongle and it also counts ur mobile data usage (for those that have limits).

opengles and op_power.sh don't like each other very much.
whenever returning from low power to full power mode again, opengles rendering is just dead.

the reason for this issue are following 2 lines within op_power.sh

18 echo 0 > /sys/devices/platform/omapfb/graphics/fb0/blank
...
66 echo 1 > /sys/devices/platform/omapfb/graphics/fb0/blank

if these lines are removed/commented, the gles rendering just resumes fine.

using evdev xorg drivers and not the tslib drivers you can receive pressure axis data however there is
a problem with the values reported by the ads7846 module and the evdev driver has no calibration property for pressure

using
xinput test touchscreen
you can see the pressure axis ranging from 120-ish for hard press to about 150-170 for light press (low pressure values can
be quite erratic)

I could possibly hack the evdev module to possibly do something with this odd and narrow range of values, but I rather think looking
at the source that the ads7846 module needs some work.

is there a git tag that has a known working configuration? as I've had problems getting the kernel to compile the last few times I attempted...

If you hit the function button, then hit it again to lock the function keys, it doesn't lock the function key.

The delete key, hold shift and hit the back space button, it won't delete, meaning it won't remove the character to the right of the cursor. I saw this on the forum, and confirmed it on my pandora also.

Found I'd kept some notes about making the Pandora's wake on alarm function work...

drivers/rtc/rtc-twl4030.c

twl4030_rtc_remove (for when the driver is removed) and twl4030_rtc_
shutdown (for when the system is shutdown) functions need to change. Or at least the shut down does.

oh YUK who's been using goto....

changes in drivers/rtc/rtc-twl4030.c

twl4030_rtc_remove
commented out //mask_rtc_irq_bit(BIT_RTC_INTERRUPTS_REG_IT_ALARM_M);

twl4030_rtc_shutdown
replaced
// mask_rtc_irq_bit(BIT_RTC_INTERRUPTS_REG_IT_TIMER_M |
// BIT_RTC_INTERRUPTS_REG_IT_ALARM_M);
with
mask_rtc_irq_bit(BIT_RTC_INTERRUPTS_REG_IT_TIMER_M);

twl4030_rtc_suspend
replaced
// mask_rtc_irq_bit(BIT_RTC_INTERRUPTS_REG_IT_TIMER_M |
// BIT_RTC_INTERRUPTS_REG_IT_ALARM_M);
with
mask_rtc_irq_bit(BIT_RTC_INTERRUPTS_REG_IT_TIMER_M);

twl4030_rtc_init
added (before return!)
twl4030_rtc_alarm_irq_set_state(true);

Should probably check reg 0x2B with mask 0x08 to see if alarm should be left enabled or not ?
but only in twl4030_rtc_shutdown and ?remove? NOT in twl4030_rtc_init
should remove ever happen ???

I'm guessing it's best to not enable the alarm irq's all the time???

SD-Card in Pandora can only be unmounted by guest (uid=1000) but not by jgeiss (uid=1001).

Steps To Reproduce:

Insert SD-Card (or boot with SD-Card inserted),
Login as jgeiss (uid=1001),
Try to unmount SD-Card via XFCE-Popup-Menu (right mouse click on SD-Card-Icon and select unmount).

a) Host multiple volumes at once.
b) Select one for hosting first, later host additional ones. (I guess this scenario is more complicated)

This seems to be the case with Hot-fix 5 and 6 alpha 4.

I've been able to repeat this bug by re-flashing..

On a fresh re-flash.. installed community codec pack.. Videos will play fine on Panplayer, VLC and Gnome-Mplayer with default settings, which I believe is XV or SDL out in the case of VLC.

Run the TV-out script it will cause a blank black screen during playback.. Disabling TV-out, switching modes, rebooting.. battery out, nothing seems to allows it to work with XV/SDL out again.. I know If I switch to X11 it will make it "work" again, but I notice a bit of lag during playback using X11 compared to XV/SDL out.

This is the related documentation:
http://pandorawiki.org/Power_modes#Low_Power
(Please update accordingly as part of the issue resolution)

FROM MY USER EXPERIENCE:

If you put the Pandora into Low Power Mode, then press some keys, and then wake the Pandora back into Normal Mode, your input (both keys and nubs, haven't tried with USB input yet) seem to have triggered something while the device was in Low Power Mode!

What does really happen in Low Power Mode concerning execution and input?

a) Input is received AND triggers as application execution continues
or
b) Kernel/driver queues the input signals into a buffer, and executes them on wake?

If supposition a) or b) is true, then this would be one more reason to properly implement:  FS#259 

OBSERVANCE EXAMPLE 1:

1) Start MiniMenu. Mentally note down your active tab.
2) Put Pandora to Low Power Mode.
3) Press shoulder button R once.
4) Wake Pandora to Normal Mode. You are now one tab to the right of where you left. Input must have been caught in Low Power Mode, but wether execution of it happened while Low Power Mode or later after wake in Normal power Mode is unclear.

OBSERVANCE EXAMPLE 2:

1) Start gedit (a text editor). Insert the digit "1".
2) Put Pandora to Low Power Mode.
3) Press: CTRL-N 2 CTRL-N 3 CTRL-N 4. (Without the spaces)
4) Wake Pandora to Normal Mode.
5) For a fraction of a second you see the tabs building up. Either the input was really received AND executed while Low Power Mode and what you see is just a delayed window manager refreshment, or the input was queued in Low Power Mode, and only executed on wake.

Built-in keyboard input is a central thing on the Pandora, as it concerns almost all user interfaced apps!
I realized that I myself and also others submitted quite many reports concerning this issue.
Hence this meta issue is intended as an overview/accumulation/aid for those devs who are willing to overwork the whole issue.
If you realize new related issues, feel free to add them here.
If this my effort is contradicting the OpenPandora workflow, then pardon me, and instruct me, how else to handle issues of that kind. Thanks!

Keyboard low level: Driver, keyboard layout, post processing (hotkeys, input support, etc)
FS#138 FS#227 FS#242  FS#259 

Keyboard mid level: Application interfacing
 FS#102  FS#123 FS#238

Keyboard application internal level
 FS#157  (dupe:  FS#249 ) FS#243 FS#256

Thunar takes a long time going back to minimenu on exit. You don't notice this when in XFCE but then you are always in the GUI not returning to it.

This started in Hotfix 6 Alpha4 I believe

Since the default inputrc doesn't contain any functionality (simple cli movements like ctrl-left or crtrl-right to skip words don't work) this diff may be useful to add to the cli-experience. I've also been trying to get the delete-key to work, but this is apparently already a known issue (see FS#227). I can remove this from the diff if you want. (I couldn't attach a file to the task somehow)

31,32c31,33
< # "\e[3~": delete-char
< # "\e[2~": quoted-insert
---
> #"\e[3~": delete-char
> "\e[^?": delete-char
> "\e[2~": quoted-insert
44,47c45,50
< # "\e[5C": forward-word
< # "\e[5D": backward-word
< # "\e\e[C": forward-word
< # "\e\e[D": backward-word
---
> "\e[1;5C": forward-word
> "\e[1;5D": backward-word
> "\e[5C": forward-word
> "\e[5D": backward-word
> "\e\e[C": forward-word
> "\e\e[D": backward-word

In GTK applications, such as Truecrypt, NoteCase Pro or Mousepad, the file open / file save dialogs have shortcuts to mounted file systems on the left side.
If clicking such a shortcut entry, the corresponding file system's contents are shown in the right dialog pane.
This works for file systems, which were available at boot time.
For file systems mounted after booting, e.g. by inserting an SD card or connecting a USB drive, when clicking such an entry, an error message appears saying
"Could not mount [volumen name]
Failed to execute child process 'gnome-mount' (No such file or directory)"

This is healed by
sudo opkg install gnome-mount
(which instals gnome-mount, nautilus, gvfsd-ftp, gvfs (upgrade), and a lot of libs (upgrades)..

This has been seen in SuperZaxxon release 1.5, not in Beta3 (but Beta3 is the only choice in Flyspray's dropdown currently).

I reported that error for HF6 I think, and it seemed to be fixed in SuperZaxxon Beta5. A regression?

Using pygame I try to play a music file starting from some position in the middle::

import pygame
pygame.init()
pygame.mixer.init()

pygame.mixer.music.load('some_file.ogg')
pygame.mixer.music.play(0,60)

This should start playing the file from the firs minute, instead on the pandora the file plays from the start.

I've tried the same code in a debian (sid) chroot (from extend utils) and it works correctly, as it does on my pc (debian wheezy).

I suspect that the issue may be present also in .next (I've done a very quick test on a friend's pandora).

Some relevant version numbers:

* libogg.so.0.6.0 (on debian libogg.so.0.8.0)
* libSDL_mixer-1.2.so.0.10.1 (on debian libSDL_mixer-1.2.so.0.12.0)

When the taskbar is set to autohide, the Pandora button does not bring up the menu anymore.

With SuperZaxxon Final, I've noticed that stuff running on the Pandora like Thunar and bash are seeing a different interpretation of filenames on disk than when I remove the SD card and insert it into my desktop PC.

Specifically, the following two mismatches appear to be present when using a FAT32-formatted SD card:

First, the Pandora's VFAT support seems to be be operating with different case-handling settings than desktop Linux distros.

Filenames set in all uppercase are forced to all-lowercase on the Pandora (whether they're set on the desktop or the Pandora) but inserting the card into a PC reveals that the. All-lowercase filenames set by software on the Pandora risk appearing in all uppercase when the card is inserted into a PC. (I think the grsync PND is what created the files in question)

Among other things, this forces me to use EITHER the PND-based copy of grsync over the network OR a desktop-based copy of rsync with an SD card reader but not both because the kernel will preserve case differences like ALBION.BAT vs. albion.bat but rsync think they are separate files. (resulting in Flash-killing, time-wasting deletion and re-creation)

It also makes for irritating entries like "zzt" in DOSBox or "ddr" in PyDance where I can't capitalize the filename properly without employing Department of Redundancy Department with "DDR Songs" or "ZZT.The.Game" as an excuse for mixed case.

Second, SuperZaxxon Final seems to be using a different (though still Unicode-capable) filename encoding than all the desktop distros I've tried.

Filenames are preserved properly and all-ASCII filenames are displayed properly in all cases, but any non-ASCII characters appear as gibberish on whichever system was not used to set them.

I've confirmed this problem with these two filenames:
- 03 - The Foggy Dew with Sinéad O'Connor.flac
- Lucky Star - Native Misao (Touhou - Native Faith) ???????????????????×??????.mp4

As a Canadian user, my desktops use the "en_CA.utf8" locale and the only other distros I've found which have this problem are Slax and the 1.0 release of its successor, Porteus, both of which use ISO-8859-1 (latin1) for filenames.

As verified with xev, an attached bluetooth keyboard can not generate colon (shift semicolon). xmodmap shows that (unlike a standard Linux system) shift-semicolon is mapped to NoSymbol. While it's find that there's a special symbol to get semicolon on the built-in keyboard, the other binding really should be there in support of attached keyboards.