Release PNDstore
- Thread starter Tempel
- Start date
Tempel
Active Member
If that were the case, there would be a warning printed, and it would keep running. Segfaults should only appear in external, non-Python code. So it could be a bug in, say, PyGTK, but it's more likely an error in how PNDstore interfaces with libpnd, which it only does when it needs to update the list of installed packages. That's why I suspected it might be one bad PND causing it. But since it hasn't happened again, I'll conclude it was gremlins. peelie, you need to keep those gremlins away from your Pandora.
Caine
Hardcore Member
Unless you are generating your own python bytecode at runtime (which is fun, but I know you are not doing that), the most likely sources are c-libraries and python bindings of those such as e.g. PyGTK or especially ctypes interfacing with libpnd. Personally, I would start looking at the latter one.
I've been working a bit on turning XBMC into a full featured Pandora front-end application and as the first step in that direction I have plugged PNDStore directly into XBMC through its addons. Whenever I perform a local database update (i.e. database_update.update_local()) it seems to (consistently) crash XBMC. This is probably unrelated and most likely caused by an improper setup on my end. Last time I checked out your source was about a week ago or so, so if you recently fixed something there then let me know. I'll try to keep an eye open for that segfault but it is not my first priority now.
[edit]My error seems to occur when finding the PNDs on the searchpath in the line:
search = libpnd.disco_search(searchpath, None)
[/edit]
[edit2]pndrun_xbmc.out reports:
Code:
[--------------------------------]{ App start }[--------------------------------]
sh: lsb_release: not found
sh: lsb_release: not found
ALSA lib pcm.c:7234:(snd_pcm_recover) underrun occured
/usr/pandora/scripts/pnd_run.sh: line 24: 2370 Segmentation fault "./$EXENAME" $ARGUMENTS
[--------------------------------]{ App End }[--------------------------------]This occurs exactly on the invocation of libpnd.disco_search(searchpath, None). So it seems that is causing the segfault (at least in my case).
The value of searchpath was: "/media/*/pandora/apps:/media/*/pandora/desktop:/media/*/pandora/menu:/usr/pandora/apps"
[/edit2]
[edit3]
Hmmm. I moved all pnds of the searchpath except for a freshly downloaded xbmc.pnd to see if there was a corrupt pnd present.
This changes the error to the following (putting the pnds back generates the segfault again):
Code:
[--------------------------------]{ App start }[--------------------------------]
sh: lsb_release: not found
sh: lsb_release: not found
ALSA lib pcm.c:7234:(snd_pcm_recover) underrun occured
ALSA lib pcm.c:7234:(snd_pcm_recover) underrun occured
*** glibc detected *** ./lib/xbmc/xbmc.bin:: double free or corruption (out): 0x018c9500 ***
/usr/pandora/scripts/pnd_run.sh: line 24: 3806 aborted "./$EXENAME" $ARGUMENTS
[--------------------------------]{ App End }[--------------------------------]The error still occurs on the same line, i.e. invocation of libpnd.disco_search(searchpath, None).
[/edit3]
-Caine.
Last edited by a moderator:
Tempel
Active Member
Caine: Thanks for the investigation! Unfortunately, I don't know what to do about your results. In disco_search, I've only ever been able to induce a segfault by setting searchpath to None. Even an empty string works fine. Since you say that searchpath is set to the libpnd default, I'm not sure where the segfault could be appearing. I could perhaps read through the libpnd code to see if there could be a problem in pnd_disco_search, but I hardly know C and probably wouldn't be very effective at bughunting. So sorry Caine, I have to leave you on your own for now!
But out of curiosity, how are you interfacing with PNDstore? Your .outs seem to show calls to pnd_run.sh, yet you mention calling specific functions rather than the command-line interface. What magic are you performing in there? It probably has nothing to do with the segfaults, but since I can't figure out the segfaults, who knows.
But out of curiosity, how are you interfacing with PNDstore? Your .outs seem to show calls to pnd_run.sh, yet you mention calling specific functions rather than the command-line interface. What magic are you performing in there? It probably has nothing to do with the segfaults, but since I can't figure out the segfaults, who knows.
Caine
Hardcore Member
No problem. I'll track it down sooner or later
No magic at all. I simply replaced your GUI with a python file of my own. I.e. I have a single python unit to implements a plugin for xbmc (which itself is launched by pnd_run). This python unit imports database_update.py and packages.py to do the work quite similar to __init__.py in pndstore_gui.
So, in a sentence, source level integration where I only replace the gui.
[edit]Did some more testing. I enabled the logging in libpnd by calling pnd_log_set_buried_logging as well as pnd_log_to_stdout (By the way, Tempel, you didn't expose any of those functions yet in your libpnd.py). Now I can see which pnd file was last encountered by pnd_disco_callback.
Some PND files are overrepresented in the crashes which suggest malformed PNDs or some other corner case. E.g. i'm seeing audacity very often before the segfault. When scanning /usr/pandora/apps/, I see consistent crashes on op_tvout.pnd. The strange thing is, PNDstore itself doesn't seem to experience those crashes even though it is calling exactly the same code.
I recall that XBMC actually uses a builtin python interpreter, maybe that one is causing some differences...
[/edit]
Last edited by a moderator:
Tempel
Active Member
New beta here! I'm calling this one version 1.0 because you need to empty the pndstore appdata directory before using (no need anymore)!
Get the 1.0 beta here! Despite the version number jump, there are very few user-facing changes here. But there are some significant backend changes, so I need it abused as best as you can.
The repo v3 change required a lot of changes to the backend code (but very little of the API, so you should be clear Caine). But though I've tested it as best I can, I can't be fully confident in it until it's used for real for a little while. So please keep an eye out for any inconsistencies in the package lists, like if new packages don't appear or installed packages aren't aware of upgrades, since those would be critical failures here.
And to remind you, you need to empty the pndstore appdata directory before using! Expect me to say that a lot in the near future.
Get the 1.0 beta here! Despite the version number jump, there are very few user-facing changes here. But there are some significant backend changes, so I need it abused as best as you can.
- Updated to fully use repo spec version 3.0. This should make list updating faster and more bandwidth efficient.
- Made remote list updating much more robust. Not that's it's needed for milkshake's repo, but now I say bring on the hastily-coded alternate repos!
- Now searches all the directories it should for PNDs, including /media/*/pandora/mmenu.
- Ctrl+Q can be used to close the GUI.
The repo v3 change required a lot of changes to the backend code (but very little of the API, so you should be clear Caine). But though I've tested it as best I can, I can't be fully confident in it until it's used for real for a little while. So please keep an eye out for any inconsistencies in the package lists, like if new packages don't appear or installed packages aren't aware of upgrades, since those would be critical failures here.
And to remind you, you need to empty the pndstore appdata directory before using! Expect me to say that a lot in the near future.
Last edited by a moderator:
hi anychance next release install location can be chosen? or have defaults to /sdcard/pandora/menu/apps
; /sdcard/pandora/menu/emulators; /sdcard/pandora/menu/games?
cheers
milkshake
Advanced Member
It already lets you pick from the default locations that you would expect... i.e.
/sdcard/pandora/apps/
/sdcard/pandora/desktop/
/sdcard/pandora/menu/
meandu229
Zubeman
Feature request, the location to save is set as a default for new time (if available)
E.g I always install things to /media/PANDY/pandora/menu
but the 2 partitions on my boot card are always before that in the drop down box,, (yes I am that lazy)
Love the app makes my life easier.
E.g I always install things to /media/PANDY/pandora/menu
but the 2 partitions on my boot card are always before that in the drop down box,, (yes I am that lazy)
Love the app makes my life easier.
meandu229
Zubeman
I do that aswell, Apps, Emus, Games, System, would it be possible to have a config to make it go into the folder from the 1st category from pxml?
stratego
Still Fresh
The package list is not updated on my OP. When I click on Update lists, the message Updating remote package list... appears in the bottom but nothing happens.
I deleted the folder pndstore in appdata, and I can see that a new one has been created, containing 2 files (app_database.sqlite and pndstore.cfg), but the time stamp hasn't changed since the time when I executed PNDstore.
I have noticed that there is a pndstore at /mnt/utmp. Should I delete that folder first?
Thanks a lot.
Last edited by a moderator:
Caine
Hardcore Member
Nice, I'll check this out later.
Tempel, don't hesitate to make API changes for my sake. It's not a lot of code and easily adjusted.
Tempel, don't hesitate to make API changes for my sake. It's not a lot of code and easily adjusted.
you do have to wait a bit on initial update
stratego
Still Fresh
I thought so, and I waited around 10 minutes. Is it normal?
stratego
Still Fresh
Now waited for almost 30 minutes, nothing changed (still says Updating remote package list...). Files in both /mnt/utmp/pndstore and /media/PANDORASD1/pandora/appdata/pndstore have timestamp from 14:33 (that's when I launched PNDstore).
Internet is working, because I can surf with Midori.
Any ideas? Thanks a lot.
Internet is working, because I can surf with Midori.
Any ideas? Thanks a lot.
stratego
Still Fresh
OK, somehow worked. I don't know, I just closed and tried again a couple of times.
Anyway, thanks for the help. And it may help somebody else who may have the same issue.
Anyway, thanks for the help. And it may help somebody else who may have the same issue.
Caine
Hardcore Member
I was just about to post the same. For me the first remote update stalled but closing down the application and retrying worked.
Tempel
Active Member
I don't suppose you have the .out files from when it stalled, do you? I'm not seeing such stalls on my desktop. It could just be network flakiness, which I understand to be prevalent on Pandoras, but if it's something more serious, it should show up in the .out, I think.
. Just haven't needed to change it.
Don't worry, I'm not giving you special treatment
. Just haven't needed to change it.
