Cowon D2

[grantmasterflash]

The fact that I can fix it by deleting the file from the D2 and copying it anew from my PC makes me wonder if it has to do with some fragmentation of the D2's internal memory that Rockbox doesn't like...  I don't know - I'm shooting in the dark.

Do these files play reliably if you play them from the SD card?

If so, the problem is most likely due to incorrect file data being read from the internal flash. There are almost certainly bugs left in our implementation of the NAND FTL (flash translation layer), and since FLACs tend to be rather large files they are likely to show this kind of problem more often than other formats.

Edit to add: Now we have semi-reliable SD access it should be much easier to debug NAND issues, however it's unlikely I'll find the time to look into this for a while.

Based on your question about the SD card, I did the following:

1.  Copied one of the 'skipping' files from internal memory to the SD card from within Rockbox using the context menu to copy and paste and played the file on the SD card

Result: File copied to SD still stopped playback at the same point as the original file did on internal memory

2.  Copied the same 'skipping' file from internal memory to the SD card using standard D2 USB mode (so just plugged to my PC), then booted into Rockbox and played the file on the SD card

Result: File copied to SD played without a problem

I would guess the results in #1 back up your theory about the NAND FTL, since copying the file from within Rockbox resulted in a duplicate file that exhibits the same problem.

[shotofadds]

Yes, that's pretty conclusive. Accessing files from the internal flash is "fairly" reliable now (it's certainly much, much better than it used to be) but clearly it isn't quite 100% yet. I just need to find the time to look into it again.

[mbk1969]

grantmasterflash, I have encounted the same problem

[kristinuk21]

grantmasterflash, I have encounted the same problem

Maybe a diff between original file and file copyed from memory would help someone find the issue faster!

[samokhov]

I've copied a skipping file from internal flash to SD by RockBox and then compared to the original. 4 bytes differ. 4 bits, actually:

original value	misread value
77	75
14	16
23	A3
B7	F7

I've uploaded the files, if that helps:
http://samokhov.com/files/itwit/original.flac
http://samokhov.com/files/itwit/misread.flac

[shotofadds]

Ah, it all becomes clear! Those are all single-bit errors, which is almost certainly because the NAND driver does not yet implemented any error correction (ECC). Nothing to do with the FTL, phew!

It should be relatively straighforward to add ECC, but try as I might I couldn't get any errors to show up on my D2 at all, so it makes testing rather difficult! It's another one for the to-do list I'm afraid...

[samokhov]

It should be relatively straighforward to add ECC, but try as I might I couldn't get any errors to show up on my D2 at all, so it makes testing rather difficult!

I've got a 8GB D2+, maybe that matters.

UP: Then again, since grantmasterflash gets the same kind of errors on his 16GB D2, probably it doesn't 

[grantmasterflash]

It should be relatively straighforward to add ECC, but try as I might I couldn't get any errors to show up on my D2 at all, so it makes testing rather difficult! It's another one for the to-do list I'm afraid...

In my experience, I find one of these error-inducing files approx. every 10-12 tracks - for example, one track per flac album on D2 exhibits this problem.  If you rip a few CDs to flac (or convert 20-30 audio files you have, whatever you prefer), transfer these to your D2, load them into a playlist, and start listening, I would bet you'll encounter this issue (or at least I hope so, for the sake of getting to the bottom of the problem).

[kristinuk21]

In my experience, I find one of these error-inducing files approx. every 10-12 tracks

So how about a "how to reproduce"
If a bug is easy reproducible by someone that knows how to fix it, i'm pretty sure that it will be fixed. So if you really want to fix this thing ..why don't you give a hand ??
- steps required:
1) find Rockbox versions used
2) pick n tracks, [free tracks if possible ]
3) clear the player [delete all files except Rockbox - maybe a clean install would help more ]
4) copy files to the player [describe how...if you think it might help]
3) listen to the skip, mark it down somewhere
4) repeat procedure - see if reproducible
5) zip everything
6) if possible ask others to reproduce [with the zip]
6) file a bug (you know ..we have buttons in the left up )
7) then ask what can be done in order to help more the debug

Since you said is easy ... it shouldn't take you too long, besides...do you know how frustrating is to try to fix something that you can't reproduce ?? and it might take a lot of time, to find the bug...time that could be invested in other ....bugs.

[mcuelenaere]

Ah, it all becomes clear! Those are all single-bit errors, which is almost certainly because the NAND driver does not yet implemented any error correction (ECC). Nothing to do with the FTL, phew!

It should be relatively straighforward to add ECC, but try as I might I couldn't get any errors to show up on my D2 at all, so it makes testing rather difficult! It's another one for the to-do list I'm afraid...

You do know there's a patch for this on the tracker?

[shotofadds]

Yes, but there are reasons why I can't use that code. It needs implementing afresh from the datasheet.

[mcuelenaere]

Yes, but there are reasons why I can't use that code. It needs implementing afresh from the datasheet.

Then I suppose that patch should be closed/rejected?

[grantmasterflash]

In my experience, I find one of these error-inducing files approx. every 10-12 tracks

So how about a "how to reproduce"
If a bug is easy reproducible by someone that knows how to fix it, i'm pretty sure that it will be fixed. So if you really want to fix this thing ..why don't you give a hand ??

Have you taken the time to read what I've posted here before lecturing me?  Clearly you have not.

I provided the exact Rockbox build I used in my first post about this (shotofadds' official test build with the SD driver enabled - see my post on 7/30 at 4:26pm), but the build is also unimportant - the problem has occurred with every Rockbox build for D2 for the past several months at least.  My message right before this post from you detailed exactly what can be done to reproduce it (see my post on 7/31 at 4:00pm).

Additionally, shotofadds has indicated he has a pretty good idea of what is causing it.  If more is needed from me, I will be happy to assist.

You are correct that I haven't filed a bug, but since shotofadds feels that this is due to lack of ECC, he can probably file a more intelligent bug about this than I can.  Additionally, I don't believe the bug tracker is for new ports that haven't been released yet, such as D2's Rockbox - my understanding is that it is for reporting bugs on the officially supported and released version of Rockbox.   

If you take the time to read what has already been posted here, you will find that your posting is pretty unnecessary, but thanks for the lecture.  Given your whopping 2 post total in this forum, I hardly think you should be doling out advice as if you're a seasoned veteran.

[markV]

Hi,

I am using the test build version r22056M-090726 on my D2 8Gb + 2Gb SD card.  Everything is working well; I'm so happy Rockbox finally saves my settings !  Thank you so much !!!

Aside this, I've made the following observation: when using a 16Mb (yes, megs!) or 256Mb SD cards with the same test build, Rockbox boots then shuts down and never comes back.  The file system goes corrupted and I even had to reformat these SD cards. 
Why is that ?  At first I thought 16 Mb was not enough but 256Mb should be...

Is it because these cards are not SDHC ?

MarkV

[shotofadds]

Aside this, I've made the following observation: when using a 16Mb (yes, megs!) or 256Mb SD cards with the same test build, Rockbox boots then shuts down and never comes back.  The file system goes corrupted and I even had to reformat these SD cards.

I've certainly not seen that behaviour before. I've tried a 16Mb card (which Rockbox won't recognize since it's FAT12 formatted) but it didn't experience any corruption - it just ignored it. I've also used a 512Mb non-SDHC card plenty of times without any trouble at all. Are you sure these aren't MMC cards?

Can you explain "Rockbox boots then shuts down and never comes back" in a bit more detail?