|
82 | Base OS | Bug Report | Critical | Network Manager - Wifi Can't Reconnect | Closed | |
Task Description
When wifi drops out, or you have disconnected from it and wish to reconnect, you cant reconnect using network-manager.
The connecting icon spins for a while, then it pops up the 'Password' Dialog. If you enter your password, the cycle begins again. |
|
201 | Core | Bug Report | High | PND apps still use memory after they've quit | Closed | |
Task Description
After running and quitting a PND app, not all of the used memory is being freed.
This eats more and more memory over time. Just run 5 or 6 apps and you only have half of the memory left.
Sounds like either aufs or unionfs don't free the memory after unmounting.
This has been reported with HF4 and still is the case with HF5 Beta 2
-- downgraded to 'Researching' and 'Normal' priority until confirmed and details found |
|
2 | Core | Bug Report | High | pndnotifyd | Closed | |
Task Description
pndnotifyd needs to be restarted before it works (tested in image 2009-12-21)
This can manually be fixed by installing two packages:
http://www.angstrom-distribution.org/unstable/feed/armv7a/base/sessreg_1.0.4-r0.2_armv7a.ipk
http://djwillis.openpandora.org/pandora/packages/pandora-libpnd_1.0-r12.5_armv7a.ipk
And adding to lines to slim.conf:
sessionstart_cmd /usr/bin/sessreg -a -l $DISPLAY %user
sessionstop_cmd /usr/bin/sessreg -d -l $DISPLAY %user |
|
10 | Core | Bug Report | High | Permissions / sudoers isse in 12-21 image | Closed | |
Task Description
pndnotifyd will create .desktop files
Running those will fail; if you've just done a recent sudo, they'll run fine
Probably sudoers file needs fix |
|
14 | Core | To Do (Reminder) | Low | Add hup of pndnotifyd to .xinitrc | Closed | |
Task Description
Pretty minor risk, but if the guy creates multiple accounts on the machine, and logs out of one and into another .. we'll want to hup pndnotifyd
Alternative solution is to poll utmp (or inotify on utmp?) to watch for logins and handle it magicly. |
|
16 | Core | Bug Report | Medium | normal user login fails over serial/ttys | Closed | |
Task Description
Only root can login over serial or ttys, and only if it has no password. Ssh login works fine.
After some debugging it seems to be PAM issue, possibly recipes/shadow/ misconfiguration in OE. Maybe we sould use IMAGE_LOGIN_MANAGER = "busybox" instead of IMAGE_LOGIN_MANAGER = "shadow" as we don't need super security? |
|
48 | Core | To Do (Reminder) | Low | Create the TV Out Enable / Disable scripts | Closed | |
Task Description
The commands do exist, I just need to wrap them round some zenity scripts. |
|
87 | Core | Bug Report | Medium | default /etc/network/interfaces is causing problems | Closed | |
Task Description
The default /etc/network/interfaces has 192.168.1.0/24 subnet set up for usb0 which causes wifi connectivity to fail for users that have wifi on the same subnet.
All addresses should be edited out from /etc/network/interfaces, eth* stuff should probably be dropped too. |
|
90 | Base OS | Feature Request | Low | Adding more "general libs" to the OS image | Closed | |
Task Description
It would be nice if some more generic libs that could be used by software ported to the Pandora would be available in the base image. For example for Wesnoth we do require these additional libs (names taken from http://www.angstrom-distribution.org/repo/?pkgname=boost ):
* boost-iostreams
* boost-regex
* boost-serialization
Beside this Wesnoth does also require this package (from http://www.angstrom-distribution.org/repo/?pkgname=lua5.1 ):
* lua5.1
In general I would recommend pulling in all of boost. This means the dependencies as well as the recommends. Here is a short paste of what this means:
Depends:
libc6, libstdc++6, libgcc1, boost-filesystem, boost-thread, boost-date-time, boost-system
Recommends:
boost-date-time, boost-filesystem, boost-graph, boost-iostreams, boost-program-options, boost-python, boost-regex, boost-serialization, boost-signals, boost-system, boost-test, boost-thread
There are probably some other packages/lib that might make sense to directly have in the OS image, but they currently don't come to my mind, so I can't list them here. ;) |
|
101 | Core | Bug Report | High | Cannot unmount SD card or USB stick through the GUI | Closed | |
Task Description
When I rightclick and select "unmount" I always get an error, see below.
unmounting from a terminal window works fine.
Also see:
http://www.gp32x.com/board/index.php?/topic/54353-unmount-sd-card-hdd/
"Unable to unmount .... Cannot open /media/.hal-mtab" |
|
112 | Core | Bug Report | Medium | .pnd files with spaces in filename won't work | Closed | |
Task Description
pnd_run.sh barfs when given a file with space in it's name.
Also pnd_run bin doesnt enclose the .pnd name, it should.
/usr/pandora/scripts/pnd_run.sh -p "PicoDrive_180beta2 .pnd" -e PicoDrive.run
pnd set to `PicoDrive_180beta2 .pnd'
exec set to `PicoDrive.run'
/usr/pandora/scripts/pnd_run.sh: line 50: [: PicoDrive_180beta2: unary operator expected
when run from menu:
in fork!
df: /media/mmcblk1p1/pandora/desktop//PicoDrive_180beta2: can't find mount point
mountpoint:
/usr/pandora/scripts/pnd_run.sh: line 120: [: !: integer expression expected
not mounted on loop yet, doing so
Usage: grep [OPTION]... PATTERN [FILE]...
Try `grep --help' for more information.
Usage: grep [OPTION]... PATTERN [FILE]...
Try `grep --help' for more information.
error determining fs, output was cannot |
|
115 | Base OS | Bug Report | Medium | OS is missing libvorbisidec.so | Closed | |
Task Description
Please add to next hotfix, srb2 needed this file. My temporary hack was to modify the LD_LIBRARY_PATH (didnt like having to so this, but I think the way srb2 is loading the library it wouldnt work without it) and include the library in the pnd.
Once this lib is added ill do an update on srb2 and remove the LD hack. |
|
117 | Application | Bug Report | Low | Quake1 / Nanogl Lockup | Closed | |
Task Description
After the last TI gles driver update quake1 (vanilla glquake 1 sources) and nanogl started to lockup the pandora after a few minutes of gameplay. Oddly all the other apps using nanogl seem stable.
There is some bad code in nanogl (missing/null function pointers), which I need to clean up and afterwards reevaluate this issue. |
|
119 | Base OS | Feature Request | Low | ext4 support | Closed | |
Task Description
Don't know how much work this would be to get a later version of the linux firmware to add this, but it would be nice to have more filesystem support. |
|
128 | Core | Bug Report | Medium | pndnotifyd Segfault | Closed | |
Task Description
pndnotifyd Segfault when the PXML.xml attached with th pnd contain non-ascii characters (for translation) |
|
170 | Core | Bug Report | Medium | Executable directories broken | Closed | |
Task Description
Directory type packages fail to run. i.e. dropping a directory containing PXML.xml and a bunch of files onto the SD card rather than a .pnd package.
The application is detected and menu entries created correctly, however pnd_run.sh fails to bind mount the directory due to bogus double quoting. Looks like a failed attempt to handle spaces in the path. Patch below removes these incorrect quotes.
--- pnd_run.sh
+++ pnd_run.sh.old
@@ -161,2 +161,2 @@
elif [ $DFS = directory ]; then
- mntline="sudo mount --bind -o ro ${PND} "
+ mntline="sudo mount --bind -o ro \"${PND}\" " |
|
171 | Core | Feature Request | Medium | Update clock from hardware when restoring from low powe ... | Closed | |
Task Description
The clock occasionally does weird things during low power mode, sometimes gaining time, sometimes losing. This is most likely a result of the CPU being reduced to 14Mhz and the system clock being unable to keep up.
A convenient work around is to make a call to "/sbin/hwclock --hctosys" during wake-up to load the hardware clock back into the system clock. This may add an extra second or two to the wake-up time but will prevent confusion caused by clock being very incorrect. |
|
176 | Core | Bug Report | Medium | op_power.sh: problematic display brightness restoring a ... | Closed | |
Task Description
Since hotfix4 it is possible to reduce the value for display brightness below '3'. Thus the script op_power.sh (the one used for low power mode) has to be adjusted accordingly. Currently line 25 in this script looks like this:
if [ $oldbright -ge 3 ] && [ $oldbright -le $maxbright ]; then
Please change the "-ge 3" part to "-ge 1" so that lower modes are possible, too. Otherwise the screen will be set to max brightness if you have selected a value of '1' when going into powersave mode.
Somehow it looks like what is possible to select via the hotkeys and what is possible via the "application" differs. In the script op_lcdsettings.sh in line 11 a minimum brightness of 3 seems to be defined, with the hotkeys it is even possible to go down to 0 (screen off) or values like 1 and 2. This is the relevant line:
minbright=3
Brightness restoring appears to be problematic in op_lid.sh, too. Just have a look at line 25 in this script:
if [ $oldbright -ge 3 ] && [ $oldbright -le $maxbright ]; then
Another issue is that the clockrate of 14MHz is simply too low to be reasonable. Please do adjust it to 125MHz instead, since this seems to be what TI recommends and what makes most sense since starting with hotfix4 this is where the lowest powermode (OPP0) is entered. When clocking down to 14MHz issues like the clock not working correctly come up.
Relevant lines in op_power.sh:
line 14, restoring old clockrate:
if [ $oldspeed -gt 14 ] && [ $oldspeed -le 900 ]; then
line 72, setting the clockrate for low power mode:
echo 14 > /proc/pandora/cpu_mhz_max |
|
219 | Core | Bug Report | Medium | wrong owner of /home/user/Desktop | Closed | |
Task Description
I just reflashed with HF5, and it turns out that I cannot make shortcuts on the desktop. It is very simple to fix:
sudo chown janek.janek Desktop/
I remember that this bug was also present in HF4, I fixed it so quickly that I didn't even remember to report it. |
|
234 | Core | Feature Request | Medium | Changeable behaviour when closing the lid | Closed | |
Task Description
As it was requested on forum.gp2x.de and I think it's a good idea I've written some scripts to allow the user to change what happens when the lid is opened/closed.
There are 3 Settings:
-The old behaviour (turning off the screen)
-Go into low power mode
-Shutdown
Default behaviour is still the same: Turn the screen off. Maybe we could change this to low power mode? I don't know.
I made a lot of changes to op_lid.sh: http://pastebin.com/GT3qxWSx (diff: http://pastebin.com/Z8GwpDj8 )
I changed line 11 in op_power.sh to: if [ $powerstate -ne 0 ]; then
And wrote a gui for changing the setting using zenity: http://pastebin.com/9JvSTqfQ
(I still can't upload any files, don't know if it's the fault of my browser, that's why I use this paste service) |
|
237 | Core | Bug Report | Critical | /dev/mmcblkN - N sometimes points to left, sometimes to ... | Closed | |
Task Description
SPECIFICATION:
According to http://pandorawiki.org/SD_compatibility_list
/dev/mmcblk0 is the left SD card slot (the one being closer to the headphone check)
/dev/mmcblk1 is the right SD card slot
I trusted that device path specification.
And it was correct MOST of the time.
But sadly only most of the time, NOT ALWAYS!!!
This bug is severely dangerous, as a (possibly critical) command could unexpectedly affect a device other than the one you intended it for!
BUG INCIDENT:
Sadly, I realized this bug at the worst possible moment!
I issued the command: shred -fvz /dev/mmcblk0
but it erased the card in the RIGHT slot.
I realized that because the card inserted on the left has 16GB, and the one on the right 128MB.
I checked both the card positions and my entered command, to exclude my own mistake. Sadly it was not a clumsy mistake: This severe bug really exists!
BUG REPRODUCIBILITY:
I already slightly witnessed this bug during the last days, but thought it was just a coincidence.
I.e. in GParted the larger card showed up at the wrong device path.
I did a refresh and/or ejected/re-inserted the card, and the expected device path was back again, and I did not think further about this incident.
Only today with this fatal shred incident, I eventually realized this behavior is a bug.
Luckily I had my data backed up. |
|
279 | Core | Feature Request | Medium | pidgin requires (more) locales to work correctly | Closed | |
Task Description
Please add the following package to allow umlauts working in pidgin:
glibc-gconv-utf-16
After installing this package I no longer have a problem receiving umlauts in pidgin. Besides it might also make sense to add "glibc-gconv-cp1250" which might be required for western europe and koi8-r which seems to be missing for xchat and conversion of chars there.
At the moment the following locale related packages are installed (as of SuperZaxxon Beta3):
glibc-charmap-utf-8 - 2.9-r35.3.5
glibc-gconv - 2.9-r35.3.5
glibc-gconv-cp1252 - 2.9-r35.3.5
glibc-gconv-ibm850 - 2.9-r35.3.5
glibc-gconv-iso8859-1 - 2.9-r35.3.5
glibc-gconv-iso8859-15 - 2.9-r35.3.5
Please add:
glibc-gconv-cp1250
glibc-gconv-koi8-r
glibc-gconv-unicode
glibc-gconv-utf-16
Thanks! |
|
321 | Core | Feature Request | Low | Support overriding all category fields with .ovr | Closed | |
Task Description
Currently only the main category and the first sub category of the main category can be overridden using a .ovr file while libpnd uses additional PXML supplied category date to construct the desktop file. This makes it difficult to work with PNDs that supply unwanted data in these additional fields. |
|
322 | Core | Bug Report | Low | Allow removal of categories using the libpnd category c ... | Closed | |
Task Description
Since some PNDs provide nonstandard categories like GTK or GNOME, it is desirable to extend the libpnd category mapping with the ability to suppress individual categories completely. |
|
3 | Core | Bug Report | High | Permissions for brightness and CPU Speed | Closed | |
Task Description
/sys/devices/platform/twl4030-pwm0-bl/backlight/twl4030-pwm0-bl/brightness
/proc/pandora/cpu_mhz_max
Those values cannot be changed by the user, only by root.
Should be changed. |
|
4 | Core | Bug Report | Low | Missing Icons | Closed | |
Task Description
Some icons are missing:
XFCE4 Settings Manager: Accessibility, Appearance, Desktop, Display, File Manager, Keyboard, Mouse
The icons on the panel (Mail, Web, etc.) are also missing (they're invisibile) |
|
5 | Core | Bug Report | High | Touchscreen Calibration stops the First-Boot-Wizard | Closed | |
Task Description
When selecting the touchscreen calibration within the first-boot-wizard, the script stops after the calibration and the Pandora continues to boot. |
|
6 | Core | Bug Report | Medium | Calibrate Touchscreen within the Settings-Menu of XFCE ... | Closed | |
Task Description
It doesn't run. |
|
7 | Core | Bug Report | Low | No help or tips and tricks yet | Closed | |
Task Description
Help is missing as are the tips and tricks on startup. |
|
8 | Core | Bug Report | High | WiFi doesn't work | Closed | |
Task Description
Error inserting /lib/modules/2.6.27-omap1/kernel/drivers/net/wireless/wl12xx/wl1251_sdio.ko: -1 Unknown symbol in module |
|
9 | Core | Bug Report | High | pmenu doesn't work | Closed | |
Task Description
Running pmenu leads to:
ERROR: config_read_file() failed: skins/Platinum/skin.cfg (line 21 of config skin.c)
(config_skin.c:22)
ERROR: cfg_gui_read failed, exiting... ( main.c:1517) |
|
11 | Core | Bug Report | High | Console keymap: Two keys not mapped yet | Closed | |
Task Description
INS (Fn + Del) is not working in console yet.
ALT and CTRL are not working in console yet.
The rest of the console keys are fine (tested in nano).
Keys in X are mapped completely. |
|
13 | Core | To Do (Reminder) | Low | verify pndevmapperd is invoking scripts | Closed | |
Task Description
menu button is verified
verify power button
verify lid open/close |
|
15 | Core | Bug Report | Medium | Permissions / automounted SD Card | Closed | |
Task Description
Only root has write access to automounted SD Cards. The normal user can only read. |
|
17 | Core | Feature Request | Low | Backlight off when display shuts off | Closed | |
Task Description
Would it be easy to make the backlight shut off when the screen goes off after a period of inactivity? That would save a lot of energy both during the charge cycle and if left on with a battery running.
Thanks! |
|
18 | Core | Bug Report | Medium | Some errors in first-run-wizard.sh | Closed | |
Task Description
Line 154 and 157 need to be changed:
153 if [ $launcher == "xfce" ]; then
154 sed -i 's/.*sessions.*/sessions xfce4,pmenu/g' /etc/slim.conf
155 echo Xfce selected as default interface
156 else
157 sed -i 's/.*sessions.*/sessions pmenu,xfce4/g' /etc/slim.conf
158 echo PMenu selected as default interface
159 fi
They should be:
154 sed -i 's/.*sessions .*/sessions xfce4/g' /etc/slim.conf
157 sed -i 's/.*sessions .*/sessions pmenu/g' /etc/slim.conf
Note the empty space behind sessions. :) |
|
19 | Core | Feature Request | Low | XFCE-Menu Editor | Closed | |
Task Description
Add an XFCE-Menu Editor (using python).
This way, people can edit the Startmenu :)
http://my.opera.com/Texel/blog/2008/02/20/xfce-applications-menu-editor-0-4 |
|
20 | Core | Bug Report | Low | pndevmapperd: op_lid.sh / op_lidtoggle.sh | Closed | |
Task Description
pndevmapperd tries to run /usr/pandora/scripts/op_lidtoggle.sh, however, the filename is /usr/pandora/scripts/op_lid.sh
Either rename the file or change the call ;) |
|
21 | Core | Bug Report | Low | op_bright.sh doesn't work correctly | Closed | |
Task Description
The file has to be changed to make it work.
The working script is:
#!/bin/bash
#get value range
minbright=3
maxbright=$(cat /sys/devices/platform/twl4030-pwm0-bl/backlight/twl4030-pwm0-bl/max_brightness)
curbright=$(cat /sys/devices/platform/twl4030-pwm0-bl/backlight/twl4030-pwm0-bl/brightness)
device=/sys/devices/platform/twl4030-pwm0-bl/backlight/twl4030-pwm0-bl/brightness
if [ ! $1 ]; then
newbright=$(DISPLAY=0:0 zenity --scale --text "set brightness" --min-value=$minbright --max-value=$maxbright --value=$curbright --step 1)
else
newbright=$1
fi
if [ $newbright -le $minbright ]; then newbright=$newbright; fi
echo $newbright > $device |
|
22 | Core | Bug Report | Low | pndevmapperd: Powerbutton / hold doesn't work. | Closed | |
Task Description
Moving the powerbutton doesn't get recognized by pndevmapperd.
When moving the powerbutton to hold, it reports:
keyboard.c: can't emulate rawmode for keycode 152 |
|
23 | Core | Bug Report | Very Low | slim-init doesn't pass session to .xinitrc | Closed | |
Task Description
It seems like slim doesn't send the session to .xinitrc
I included an "echo $1 >> ~/debug.txt" in the script and the file was completely empty.
That's why it always starts xfce4, not pmenu: The fallback is startxfce4. |
|
24 | Core | Bug Report | Low | op_menu.sh not working | Closed | |
Task Description
Reports:
xfce4-appfinder: Cannot open display: .
Fix might be (according to vimacs):
Change those two similar looking lines to:
sudo -u $xfceuser DISPLAY=:0.0 xfce4-appfinder
and:
pidlist=$(pstree -lpA | grep pnd_run.sh | sed -ne 's/.*-\(.*\)(\([0-9]\+\))/\2\n \1/p' | DISPLAY=:0.0 zenity --list --multiple --column "pid" --column "name" --title "kill" --text "which apps should be killed" | sed 's/|/\n/')
Needs to be checked. |
|
25 | Core | Bug Report | Medium | QWERTY Slowdowns | Closed | |
Task Description
Cleaned up a bit according to everything we know.
EDIT: notaz did it again. Fixed it. If you find any weird side effects, let us know :)
http://notaz.gp2x.de/tmp/Xorg
Replace the file /usr/bin/Xorg
Be sure to change permissions to 755
Switching between DPAD and keyboard does increase CPU usage on X A LOT. It even tends to crash ALSA if done too extensive.
According to notaz, this is due to some very old crap that's still included in X.
Internally it has one mouse and one keyboard, everything else is copied every time you press a key and bunch of callbacks are called, messages are sent to all screens and windows.
New tests from EvilDragon:
Disabling xfwm4 makes things better (ALSA won't crash anymore), but it still lags. Running it in minimal X (directly from .xinitrc) strangely fixes this - X won't eat up more than 17% CPU,
regardless how many buttons you press.
OLD Entry:
The qwerty controls when pressed are causing the system to slowdown making the application running to pause/delay. Ive seen in a least 2 application glquake and supertux. For example pressing space would cause the hiccups and shortly after the pressing stopped the application would regain normal running speed. The other joystick/nub controls work fine. glquake is getting controls direct from the event files. Supertux is using the SDL layer.
This was done on angstrom image from dec 19th.
EDIT: More tests done by EvilDragon with the latest image (pandora-xfce-image-omap3-pandora-20100102-new-kernel-fn-n-m-support.tar.bz2)
When you play with the DPAD and press a keyboard button, the app is making the small pause. You can press as many keyboard buttons as you want after that, there won't be another pause.
BUT as soon as you press the DPAD again, it pauses again. Then the next pause does happen when you press a keyboard button, etc.
So it feels like it pauses when you switch between DPAD and keyboard.
The most important thing: This only happens when you run SDL Games from X. It doesn't occur when running SDL games from the console. |
|
26 | Core | Bug Report | Medium | pndevmapperd only works when manually started | Closed | |
Task Description
Though pndevmapperd runs after booting (ps -A lists it), it doesn't react to anything.
EDIT: It DOES do something. It creates defunct op_menu.sh with each press.
When you run it manually from the terminal, the menu button works find and shows the appfinder. |
|
27 | Core | Bug Report | Low | xfce4-brightness-plugin: 0 isn't a good minimum value | Closed | |
Task Description
The xfce4-brightness-plugin can be set to a minimum of 0.
This is kinda confusing to the normal user, as it makes the screen totally black.
Maybe this could be changed to a minimum of 3? |
|
28 | Core | Bug Report | High | pmenu only shows preview pictures when selecting the PN ... | Closed | |
Task Description
When selecting a PND for the first time, it only shows a red X.
When selecting the PND for the second time, it shows the preview picture. |
|
29 | Core | To Do (Reminder) | Low | ToDo: Map Fn-U and Fn-I with brightness controls | Closed | |
Task Description
Include brightness change hotkeys. |
|
30 | Core | Bug Report | Medium | New issues in current image (pandora-20100102-new-kerne ... | Closed | |
Task Description
* It takes forever to startup.
It hangs while trying to initialize the new SDIO-WiFi:
[ 2363.922271] wl1251: WARNING WLAN not ready
[ 2363.927581] wl1251_sdio mmc2:0001:1: firmware: requesting wl1251-fw.bin
[ 2423.930328] wl1251: ERROR could not get firmware: -2
[ 2424.033447] wl1251: ERROR sdio write failed (-110)
[ 2425.786071] wl1251: ERROR sdio read failed (-110)
[ 2425.790893] wl1251: WARNING WLAN not ready
[ 2425.795776] wl1251_sdio mmc2:0001:1: firmware: requesting wl1251-fw.bin
* OpenPandora Start-Button is too small (has the same width as a normal icon) |
|
31 | Core | Bug Report | High | X11 Crashing with omapfb | Closed | |
Task Description
Cleaned up a bit.
* Mouse is inverted if you run SDL Games from the console.
* Mouse is not properly calibrated when you run SDL Games under X fullscreen. Fixed with notaz patch that also fixes the X11-Crash with omapfb (see below)
Regarding the X11-Crash when using SDL-Fullscreen-Apps together with omapfb:
notaz nailed down the problem: It's actually xf86-input-tslib's fault-
omapfb only acted as catalizator here, causing SDL to use different code path and call different X functions.
His patch also almost increases the touchscreen problem (see above): It runs perfect for games in 800x480.
However, on small issue is left on apps with lower resolution which use their own mouse cursor:
As soon as you move the mouse outside of the game resolution (i.e. SuperTux runs 640x480), the mouse cursor stops at the end of the virtual screen and you have an offset.
This only seems to happen when the game uses it's own mouse cursor. If you go to the level editor of SuperTux and leave, the normal X cursor appears and the problem doesn't occur anymore.
So this is probably something that needs to be fixed in the game, not in the driver.
A compiled version (for testing):
http://notaz.gp2x.de/tmp/tslib_drv.so --- needs to be copied here: /usr/lib/xorg/modules/input/tslib_drv.so |
|
32 | Core | Bug Report | Low | pndnotifyd: Crawling through directories it shouldn't i ... | Closed | |
Task Description
My current configuration is:
Rootfs on first SD Card, 2nd partition.
If I put my WIZ SD Card with tons of files into the second SD Slot while booting, pndnotifyd eats up about 95% of CPU time.
After about 8 - 9 minutes, it stopped eating the CPU and the system is working normally.
This doesn't happen when I put in the card when XFCE4 is already running.
Using LSOF, I can confirm it's crawling where it shouldn't:
pndnotify 1825 root 6r DIR 179,9 1441792 62488 /media/mmcblk1p1/roms/c64/Tapes
Could it be that on first initialization it hasn't loaded the config and crawls through the whole SD Card?
On a side note: I got 44700 files on that SD Card... so if it needs about 8 minutes, that does mean it can do over 5000 files a minute! |