Re: Data abort on Sansa Zip with latest daily build

[bahus]

My Sansa Clip Zip is also unstable on latest builds with default WPS (my issue). I've just attached to issue my config just in case it makes reproduction easy.

[Bilgus]

can you run this build
https://www.mediafire.com/file/sxd4vdznpuj8f0c/ClipPlus_3bb75e0039-230213_rockbox-full.zip/file
https://www.mediafire.com/file/b5vficmkuszpwps/SansaClipZip_3bb75e0039_rockbox-full.zip/file
 it includes the .map file so we can track it based on the PC in the abort

so far it appears deleting the config file fixes it (in my case even if you later set the same settings)

[bahus]

can you run this build

If it's for me I need sansa clip zip build (provided link for sansa clip+)

[Bilgus]

Sorry I didn't realize you were not the clip+
I split the topic to help keep them straight not sure if your issue is the same or not
but I gave you the same HEAD build as I gave carlo here https://forums.rockbox.org/index.php/topic,54472.msg251745.html
it contains the file rockbox.map and you should be able to find the listed pc (btw the error in your flyspray report)
https://www.rockbox.org/tracker/task/13388?type=2 Data abort at 30 would read Data abort at 3006b48c
I just don't have a reference as to which function it will line up to without that map file
mind you it won't be listed as exactly that address either it will be more like 3006b000 or something else prior or just after

in the supplied map file the pc you gave would line up to lcd_set_viewport_ex

Code: [Select]

                0x000000003006b3d0                lcd_init_viewport
                0x000000003006b458                lcd_set_viewport_ex
                0x000000003006b498                lcd_set_viewport
                0x000000003006b4a0                lcd_drawpixel

but mind you the pc you gave doesn't necessarily line up properly so we will need the next pc it gives you and it should get us close
 

[Carlo]

so far it appears deleting the config file fixes it (in my case even if you later set the same settings)

Am I understanding correctly that the issue may be caused by updating while keeping an old configuration file?

[Bilgus]

Maybe but hopefully not since IM not aware of any changes made to the settings between those builds but i had a problem till i moved my settings your issue disappeared once you tried default settings so anything is possible at this point

[Bilgus]

If you want to try removing a setting at a time till the problem stops it might be a particular setting but at this point if i cant reproduce it  on my clipzip or clip+ there must be something different that we are missing

[Bilgus]

I think I found it its been merged so try the latest daily
https://gerrit.rockbox.org/r/c/rockbox/+/5123

[bahus]

Tested it for few hours - so far everything's good. Thanks!

[bahus]

Just received another data abort on attempt to load bookmark So seems it's not fully fixed just got harder to reproduce

[bahus]

Steps to reproduce that almost always works for me:
1) Create a bookmark for mp3 file. Somewhere closer to the end of file. - file1;
2) Start playing other mp3 file. Wait 10-20 seconds - file2;
3) Shut down player;
4) Power on player and press Resume playback. file2 resumes playback successfully;
5) While file2 is playing go to bookmark file for file1 and try to load last saved bookmark;

I have a data abort on attempt to load bookmark.

P.S. I have another quirk on latest build most likely not related. When I remove some bookmark from bookmarks list - background image is gone (Rockbox sign in background is gone - just black screen)

[Bilgus]

Likely both are different bugs i'll try and reproduce them

[Bilgus]

I couldn't reproduce the bug using your method

but I was able to start a song from disk shutdown
restart resume playback and choose a track from the playlist catalog
and get a data abort centered in the font_cache_get() function

however I reverted my font changes and I still get a crash
so more investigation is needed I'll try to find the commit that broke it
but it might be an existing bug as well..

[Carlo]

(Clip+ user commenting on the Zip thread)

I've been using Rockbox for years on several devices (mostly iPod Classic, Sansa Clip+ and Fuze v2) with different themes and never ever had a single crashing issue, so I believe it wasn't a preexisting bug. I've reverted back to a 20221231 build that I've compiled myself and hadn't a single crash in almost two hours of use, and during the last days I also extensively used a 20221230 build that seemed to be absolutely bug free.

I understand it's like searching for a needle in a haystack, but it seems the bug has been introduced from early january onwards.

[Bilgus]

I think I found the other bug
https://gerrit.rockbox.org/r/c/rockbox/+/5146
this has also been merged

let me know if either of you are still getting crashes after this and i'll send you out another build with
a map file included so we can continue tracking it down

in the above bug when I switched to buflib_get_data_pinned
there had previously been no locking on that buffer afterwards there was a possibility to leave the buffer pinned