Disconnected on every attempt to reply

[P. Durante]

Hi,
trying to figure out the problem, I enabled debug output (at last!),
here's what I see when I send the reply (sorry for the length)

27025:   LOCK: dbus_connection_send
27025: Message 0x806b998 (2 (null) no interface no member 's') for
:1.21 added to outgoing queue 0x8065ae8, 1 pending to send

from the docs, only method_call and signal messages are supposed to
have the path/interface/member fields set, right ? so since this is a
method_return (type = 2) everything's fine

27025: Message 0x806b998 serial is 6
....
27025:  wrote 81 bytes of 81
27025: Message 0x806b998 (2 (null) no interface no member 's') removed
from outgoing queue 0x8065ae8, 0 left to send
....
27025: Notifying of change to dispatch status of 0x8065ae8 now 1 (complete)

the message is actually written!!!!
but then

....
27025: Disconnected from remote app

setting the daemon to verbose, after some crawling I saw (this is
output from the daemon, not my program)
27141: have 81 bytes, need body 9 + header 72 = 81
27141: validating body from pos 0 len 81 sig 'yyyyuua(yv)'
27141: p = 0x80b46bf end = 0x80b46e1 claimed_len 1
27141: initially caching field 6
27141: initially caching field 5
27141: Field 5 was invalid
27141: Failed to load header for new message code 16
27141:    0     0x80b4690:  'l' 0x02 0x01 0x01 BE: 1812070657 LE: 16843372
27141:    4     0x80b4694: 0x09 0x00 0x00 0x00 BE: 150994944 LE: 9
u64: 0x90101026c dbl: 1.91063e-313
27141:    8     0x80b4698: 0x06 0x00 0x00 0x00 BE: 100663296 LE: 6
27141:   12     0x80b469c:  '7' 0x00 0x00 0x00 BE: 922746880 LE: 55
u64: 0x3700000006 dbl: 1.1671e-312
27141:   16     0x80b46a0: 0x06 0x01  's' 0x00 BE: 100758272 LE: 7536902
27141:   20     0x80b46a4: 0x04 0x00 0x00 0x00 BE: 67108864 LE: 4 u64:
0x400730106 dbl: 8.49171e-314
27141:   24     0x80b46a8:  ':'  '1'  '.'  '1' BE: 976301617 LE: 825110842
27141:   28     0x80b46ac: 0x00 0x00 0x00 0x00 BE: 0 LE: 0 u64:
0x312e313a dbl: 4.07659e-315
27141:   32     0x80b46b0: 0x05 0x01  'u' 0x00 BE: 83981568 LE: 7667973
27141:   36     0x80b46b4: 0x00 0x00 0x00 0x00 BE: 0 LE: 0 u64:
0x750105 dbl: 3.78848e-317
27141:   40     0x80b46b8: 0x08 0x01  'g' 0x00 BE: 134309632 LE: 6750472
27141:   44     0x80b46bc: 0x01  's' 0x00 0x00 BE: 24313856 LE: 29441
u64: 0x730100670108 dbl: 6.24737e-310
27141:   48     0x80b46c0: 0x07 0x01  's' 0x00 BE: 117535488 LE: 7536903
27141:   52     0x80b46c4: 0x0e 0x00 0x00 0x00 BE: 234881024 LE: 14
u64: 0xe00730107 dbl: 2.97117e-313
27141:   56     0x80b46c8:  'o'  'r'  'g'  '.' BE: 1869768494 LE: 778531439
27141:   60     0x80b46cc:  'e'  'm'  'p'  't' BE: 1701671028 LE:
1953525093 u64: 0x74706d652e67726f dbl: 7.52737e+252
27141:   64     0x80b46d0:  'y'  's'  'p'  'a' BE: 2037608545 LE: 1634759545
27141:   68     0x80b46d4:  'c'  'e' 0x00 0x00 BE: 1667563520 LE:
25955 u64: 0x656361707379 dbl: 5.50772e-310
27141:   72     0x80b46d8: 0x04 0x00 0x00 0x00 BE: 67108864 LE: 4
27141:   76     0x80b46dc:  'S'  'U'  'X'  '!' BE: 1398102049 LE:
559437139 u64: 0x2158555300000004 dbl: 4.75755e-148
27141:   80     0x80b46e0: 0x00
27141: Corrupted message stream, disconnecting

so, how do I decipher this output to find the real problem ?

thanks in advance,
Paul

On 8/8/05, Havoc Pennington <hp at redhat.com> wrote:
> On Mon, 2005-08-08 at 16:24 +0200, P. Durante wrote:
> >
> > here you see, I get the message, craft a reply ( using nothing more
> > than dbus_message_new_return_message(original_message); ) and send it
> > back over the same connection I read it from, and after
> > dbus_connection_send, BANG! I'm out..
> > I can't recompile dbus with more verbosity since, if I do, neither
> > HAL, nor the dbus-utils won't work anymore (don't know why, seems the
> > ubuntu version was built with some arcane settings I can't figure
> > out), so my question is, what are the possible reasons for getting
> > disconnected ?
> >
> 
> Normally disconnection is done when the bus decides you are looking
> "malicious" by sending things that are not well-formed.
> 
> We try to instead have libdbus print a warning when you do this, so it's
> easier to debug, but the warning-printing code isn't as rigorous as the
> bus's validation code.
> 
> Typical ways to send a not-well-formed message are to break the naming
> rules for methods/interfaces/paths/etc. or to omit a required message
> field. Another possibility is a bug in libdbus.
> 
> You get kicked off rather than an error for security paranoia reasons.
> 
> If you figure out what the problem is we can add a warning on the client
> side to help future programmers.
> 
> Havoc
> 
> 
>