[JDEV] Re: jabber; what would you like to see?

Thu Sep 25 06:40:37 CDT 2003

I had exactly the same problem with PostgreSQL.

The solution was that the sql table definition was too restrictive,
resulting in errors from the DB which jabber did not have a method for
handling. Why it segfaulted I did not investigate, but it doesnt any more
after alterations of create tables.

I got on tracks because of the -d not -d - ie. going off debug mode means
that some faults are not handled properly, ie. I think that jabber assumes
it has data when it has NULL and thats usually the way to make a divide by
zero or something similar...

Mysql does not have foreign key and thus no problems there - even if the
result is nothing it is still different than NULL since no error... I
think xdb_sql people are developing under mysql and then porting to
postgres etc. and that they havent tested the foreign keys properly.

Best regards Bernino Lind

> The jabberd process gives me a segmentation fault when run with xdb_sql
> and Oracle.  I'm able to run jabberd fine under an identical setup with
> xdb_sql and mysql.  I'm running jabberd 1.4.2, the latest CVS version of
> xdb_sql, and oracle client version 9.2.0.3.0 on Debian linux with libc
> version 2.3.2.
>
> If I run the daemon without the debug switch the daemon will segault
> immediately after I try to connect to the server.  Here's a backtrace:
>
> (gdb) r
> Starting program: /home/jsmith/jabber/jabber-1.4.2/jabberd/jabberd [New
> Thread 16384 (LWP 5372)]
> xdboracle_backend_new: initialized
> xdboracle_backend_new: initialized
> xdboracle_connect: Logged to Oracle, db=jabber, user=jabberuser,
> pass=pass xdboracle_connect: Logged to Oracle, db=jabber,
> user=jabberuser, pass=pass oracle_query: select SYSDATE from DUAL type 1
> oracle_query: select SYSDATE from DUAL type 1
> xdboracle_is_connected: connected
> xdboracle_is_connected: connected
> xdboracle_free_result: done
> xdboracle_free_result: done
> 20030918T08:11:19: [notice] (-internal): initializing server
>
> Program received signal SIGSEGV, Segmentation fault.
> [Switching to Thread 16384 (LWP 5372)]
> 0x40bd9426 in __pthread_internal_tsd_get () from /lib/libpthread.so.0
> (gdb) bt
> #0  0x40bd9426 in __pthread_internal_tsd_get () from
> /lib/libpthread.so.0 #1  0x400a0de8 in malloc () from /lib/libc.so.6
> #2  0x400a51cc in strdup () from /lib/libc.so.6
> #3  0x0805a5e7 in jlimit_check (r=0x81ce660, key=0x4014e52c "localhost",
> points=1)
>     at rate.c:82
> #4  0x08050c4b in _mio_accept (m=0x81ce5c0) at mio.c:400
> #5  0x080519da in _mio_main (arg=0x0) at mio.c:704
> #6  0x0807d4af in pth_spawn_trampoline ()
> #7  0x40077014 in makecontext () from /lib/libc.so.6
>
>
> And when I try to run with debugging information, it will segfault when
> the daemon starts up, when it tries to write the debugging output to the
> console:
>
> (gdb) r -D
> [ snip ... ]
> 20030918T08:06:26: [notice] (-internal): initializing server
> Thu Sep 18 10:06:26 2003  jabberd.c:255 main load check of 1.00 with 3
> total threads
>
> Program received signal SIGSEGV, Segmentation fault.
> [Switching to Thread 16384 (LWP 5065)]
> 0x4005832e in __ctype_b_loc () from /lib/libc.so.6
> (gdb) bt
> #0  0x4005832e in __ctype_b_loc () from /lib/libc.so.6
> #1  0x0805b708 in format_converter (odp=0x8113ca8, fmt=0x808326d "s:%d",
>     ap=0x8113cf4 " \027\b\bl\002") at snprintf.c:579
> #2  0x0805c097 in strx_printv (ccp=0x8113cdc, buf=0x808d580
> "jabberd.c:256", len=62,
>     format=0x808326c "%s:%d", ap=0x8113cf4 " \027\b\bl\002") at
> snprintf.c:904
> #3  0x0805c0e9 in ap_snprintf (buf=0x808d580 "jabberd.c:256", len=63,
>     format=0x808326c "%s:%d") at snprintf.c:918
> #4  0x0805cd03 in zonestr (file=0x8081720 "mio.c", line=620) at
> str.c:343 #5  0x08051545 in _mio_main (arg=0x0) at mio.c:620
> #6  0x0807d4af in pth_spawn_trampoline ()
> #7  0x40077014 in makecontext () from /lib/libc.so.6
>
>
> In both cases, it segfaults when trying to make a libc call.  Also, in
> the first case I noticed the server is making a libpthread call.  Should
> the jabber daemon be using both libpthread and pth?
>
> Does anyone have any suggestions as to what the problem could be or
> could be able to offer any tips on debugging?
>
> Thanks,
>
> ----- Original Message -----
> From: "Bernino Lind" <lind at hlgroup.dk>
> To: <jdev at jabber.org>
> Sent: Thursday, September 25, 2003 4:39 PM
> Subject: RE: [JDEV] Re: jabber; what would you like to see?
>
>
>> Dear sirs,
>>
>> I would like to see jabber move in a direction where LDAP and RDBM is
>> a robust integrated part of the infrastructure.
>>
>> Also and in extension I would like to see a non-agent'ified middleware
>> component - ie. that a programmer can pick any packet or piece of
>> packet and apply whatever program to this packet with no restrictions
>> - aka the way FORMs are to http->html and php is to html.
>>
>> Right now I have tweaked the current xdb_sql so that it meets my
>> requirements (ie. I have some processing in the C code related to my
>> and only my project which manipulates the packets I receive, for
>> example "if username == bernino { printf 'woooo'; }" - what I call
>> middleware, since it is not backend nor frontend).
>>
>> I do not like prototyping in C - it reminds me too much of my days in
>> astrophysics doing models of the universe in Fortran 77 on a
>> supercomputer - damn it takes so much time. I like prototyping in Perl
>> or other highlevel language.
>>
>> So I have made my own xdb method, namely opening for an external
>> component, which then handles packets in the namespaces I have told it
>> to, and passing the data the packets might contain to forexample DBI.
>> Just like the xdb_perl project.
>>
>> In this work I have met one really irritating limitation: passing data
>> along is not very easy in a iq-get packet. I dont know if it is me who
>> is lame, but I simply cant find any examples of a iq-get packet that
>> sends more data than the username and xmlns.
>>
>> Lets say I would like to make a query where I fetch entries in my
>> roster groups that have been online within the last day. This is where
>> the current limiation lies within Jabber - its not possible in any
>> other way than creating an external component, a specific xmlns and
>> have this specific xmlns coded with static information, in this case
>> the distance in time from last login.
>>
>> What I picture is that one could have a scripting languague within the
>> packets, for example:
>>
>> <iq type="get">
>> <query xmlns="bla bla">
>> <script>
>> @users=fetchroster(1,2,3);
>> for ($i=0; $i<$@#users) {
>>    echo "<message to=@user[$i]> In my new roster bla bla ";
>> }
>> createrostergroup(@users, "newrostergroup");
>> return @users;
>> </script>
>> </query>
>> </iq>
>>
>> And a backend language, jabberlang:
>>
>> <iq type="get">
>> <query xmlns="jabberlang">
>> <url>test.jab</url>
>> <value1>blabla</value1>
>> <value2>123</value2>
>> </query>
>> </iq>
>>
>> When jabber sees the jabberlang packet it know to fire off the
>> test.jab file, which is say a python program or whatever and it passes
>> the value1 and value2 to the program. The test.jab does whatever to
>> this data with the full access to jabber at hand, ie. xdb etc. is
>> available, and then it returns whatever result to the client.
>>
>> In this way one it would be trivial to create a groupware project,
>> voip ... whatever utility out of the infrastructure, since in this way
>> what we have achieved is completely generalised generic xml socket
>> server - in difference to what it is now: a more or less specific
>> project aimed at messengers.
>>
>> It is not that it is not possible doing tricks with jabber - I myself
>> have created a multiuser gameplatform via flash and web and with a
>> PostgreSQL backend - its just that when one works out such a project
>> it becomes evident that the rationale behind jabber is not a generic
>> set of building blocks, as stated in the DJ Adams book, but rather a
>> specific messenger project which can be tweaked to behave as you would
>> like it to.
>>
>> Best regards,
>> Bernino Lind
>>
>>
>>
>> -----Original Message-----
>> From: jdev-admin at jabber.org [mailto:jdev-admin at jabber.org] On Behalf
>> Of GuruJ
>> Sent: Wednesday, September 24, 2003 12:16 PM
>> To: jdev at jabber.org
>> Subject: [JDEV] Re: jabber; what would you like to see?
>>
>>
>> Personally, I'd love to see Jabber making inroads into the
>> groupware/collaboration scene (maybe along the lines of Groove).
>>
>> -- GuruJ.
>>
>> milk wrote:
>>
>> > out of interest, i was just wondering what everyone thinks about
>> what  the
>>  > most important current/upcoming jabber features are, what features
>> do
>> > you personally think need to be developed/expanded, and what
>> tasks/applications
>>  > would you like to see jabber doing/working with in say, one or two
>> >
>> years time?
>>
>>
>> _______________________________________________
>> jdev mailing list
>> jdev at jabber.org
>> http://mailman.jabber.org/listinfo/jdev
>>
>> _______________________________________________
>> jdev mailing list
>> jdev at jabber.org
>> http://mailman.jabber.org/listinfo/jdev
>>
>>
> _______________________________________________
> jdev mailing list
> jdev at jabber.org
> http://mailman.jabber.org/listinfo/jdev