change in how ubiformat works

change in how ubiformat works

[twebb]

I use a sequence as follows: " flash_erase - ubiformat - nanddump -
verify " to confirm that a UBI image was properly stored in NAND
flash.  I realize that flash_erase is not recommended because it
destroys erase counters, but this is only done on virgin flash so
should not be an issue.

However, I recently upgraded the mtd-utils, and particularly ubiformat
from 1.4 to 1.5, and now I see that what I read back (via nanddump)
does not match the original UBI image.  Can anyone confirm whether
ubiformat.c changes sometime between 1.4 and 1.5 would result in this
behavior?  I have looked at ubiformat.c changes and am wondering if it
has to do with image sequence number support.

Thanks,
twebb

Re: change in how ubiformat works

[Artem Bityutskiy]

On Fri, 2010-11-12 at 13:47 -0500, twebb wrote:
> I use a sequence as follows: " flash_erase - ubiformat - nanddump -
> verify " to confirm that a UBI image was properly stored in NAND
> flash.  I realize that flash_erase is not recommended because it
> destroys erase counters, but this is only done on virgin flash so
> should not be an issue.
> 
> However, I recently upgraded the mtd-utils, and particularly ubiformat
> from 1.4 to 1.5, and now I see that what I read back (via nanddump)
> does not match the original UBI image.  Can anyone confirm whether
> ubiformat.c changes sometime between 1.4 and 1.5 would result in this
> behavior?  I have looked at ubiformat.c changes and am wondering if it
> has to do with image sequence number support.

Try to bisect it and find the commit id which changed the behavior.

-- 
Best Regards,
Artem Bityutskiy (Артём Битюцкий)

Re: change in how ubiformat works

[Artem Bityutskiy]

On Fri, 2010-11-12 at 13:47 -0500, twebb wrote:
> I use a sequence as follows: " flash_erase - ubiformat - nanddump -
> verify " to confirm that a UBI image was properly stored in NAND
> flash.  I realize that flash_erase is not recommended because it
> destroys erase counters, but this is only done on virgin flash so
> should not be an issue.
> 
> However, I recently upgraded the mtd-utils, and particularly ubiformat
> from 1.4 to 1.5, and now I see that what I read back (via nanddump)
> does not match the original UBI image.  Can anyone confirm whether
> ubiformat.c changes sometime between 1.4 and 1.5 would result in this
> behavior?  I have looked at ubiformat.c changes and am wondering if it
> has to do with image sequence number support.


Ok, just define your own sequence number. Use, say, --image-seq=100 both
with ubinize and ubiformat.

-- 
Best Regards,
Artem Bityutskiy (Артём Битюцкий)