Closed GoogleCodeExporter closed 9 years ago
I would hazard a guess that this is not a 64bit issue as I don't encounter it
on my
64bit install of Windows. I can't recall if my Karmic install is 64bit or not,
but i
don't experience it there either.
Original comment by m.joel.m...@gmail.com
on 10 Nov 2009 at 12:12
I dug into this a bit and found, that the client/server-communication is
different
when using e.g. php. So it might not be mythbox's failure, but an error in the
used
mysql connector library. I would as well report this to "upstream" then, if I
knew,
where upstream is? Is it python.org?
Anyway, attached are the wireshark dumps, "mysqlok.txt" is the working
connection,
with following PHP script, "mysqlnok.txt" the non-working connection from
mythbox.
The interesting packets are 9 and following. It seems like mysql wants to send
three
packets after login, and mythbox sends the "use database" directly after
receiving
the first one -- hence "packets not in order".
This is the PHP script I used for testing:
$conn = mysql_connect('127.0.0.1', 'mythtv', '*******');
$res = mysql_select_db("mythconverg", $conn);
mysql_close($conn);
Original comment by markus.d...@gmail.com
on 10 Nov 2009 at 9:20
Attachments:
[deleted comment]
An interesting observation with the wireshark dumps.
In the working scenario, wireshark is not able to "recognize" the request byte
sequence in frame 9 (line 78) and tags it as "Unknown (79) (79)"
In the non-working scenario, wireshark is able to "recognize" the request byte
sequence in frame 9 (line 77) and tags it as "Request Use Database"
Without going into it further, it would seem the 5.1.37 mysql client/server
protocol
has a slight change which neither wireshark or the pure python mysql client that
mythbox uses know about. Will follow up with upstream project @
https://launchpad.net/myconnpy
Original comment by semir.pa...@gmail.com
on 11 Nov 2009 at 12:02
Actually it's not part of the request, that wireshark does not recognize, but
mysql
sends three packets after login. One without payload, the unrecognized one and
then
"ok".
The myconnpy client does not wait for all three packets to arrive, but sends
the "use
database" immediately after the first (the empty one). I guess, that is the
reason
for mysql bailing out with "packets out of order".
Anyway. Could you report this to upstream? I think I do not give a good reporter
there as I dont know python, don't know the version I am using in this case and
how
to upgrade it (as it is the version bundled with xbmc, I guess) and also would
not
know how to test patches to myconnpy.
Original comment by markus.d...@gmail.com
on 11 Nov 2009 at 7:22
No problem. I'm going to run the myconnpy unit tests in a Fedora 11 vm with
mysql
5.1.37 just for grins.
Original comment by semir.pa...@gmail.com
on 12 Nov 2009 at 1:52
Looks like it is already a known issue. I'll update mythbox when the bug is
fixed
upstream. https://bugs.launchpad.net/myconnpy/+bug/380528
Original comment by semir.pa...@gmail.com
on 22 Nov 2009 at 9:32
Geert's guess in https://bugs.launchpad.net/myconnpy/+bug/380528 is right -- I
had
old_passwords=1 in my my.cnf, which seems to be a distribution default.
After setting this to 0 and resetting all passwords, mythbox can connect to
mysql and
now basically works. So I assume, this bug can be closed.
Now it's time to test the rest of mythbox :)
Thanks for the help!
Original comment by markus.d...@gmail.com
on 23 Nov 2009 at 7:05
Added solution as Q15 to the FAQ. You can verify in the attached build.
http://dl.dropbox.com/u/131337/mythbox-svn-1699.tar.gz
Original comment by semir.pa...@gmail.com
on 22 Dec 2009 at 6:37
Original issue reported on code.google.com by
markus.d...@gmail.com
on 9 Nov 2009 at 10:26