Re: [Libmtp-discuss] libmtp error: could not get object handles for some directories

[Andrea V. <and...@un...>]

Hi,
  Il giorno lun, 11/06/2018 alle 20.35 -0400, John Assanti via Libmtp-
discuss ha scritto:
> Hi. This is regarding the original thread involving Andrea, Philip
> and Marcus.
> 
> I am also a victim of Galaxy (S6) and I've done some debugging, and
> this are my 10 cents:
> - Effectively, I hit the 123/124 limit.
> - The thread hangs at io.c:2151 (handle_events, libusb code) and
> libc in poll() on a file descriptor
> - This is the backtrace, if anyone is curious about it:
> [...]
> - Now, after some USB sniffing with Wireshark, I see that the last
> packet for the bulk transfer is initiated from the host but the
> device is not answering. After the timeout of 60s the device answers
> with a ENOENT (-2) error "file or directory not found", but I don't
> know if this is a side effect of the timer.
> - I couldn't go further, but if someone provides an interesting tip,
> I can help and continue debugging.

thank you John for your contribution. Just to summarize, it seems that
Marcus gave us the keys to solve it

"...we might be again hitting a certain USB packetsize border
causing this hang. Like 512/124 is around 4 ... 512-4*124 =
16. getobjecthandles of 124 entries (+ 4 byte count + 12 byte header)
would hit exactly 512 bytes."

, but is cryptic to me :-)

John added some further info from his side (see above), and a
backtrace.

Philip didn't reply to Marcus's hint, but may have some useful ideas
to make us go on on this. Or, Marcus could give again a little help if
possible...

I know and understand that we all have few time. I would just kindly
say that some people are happy to give their help on testing, and it's
a pity if this problem remain unsolved (BTW, I have no idea how far
from the solution we are).

Thank you very much,
Andrea