telegram transport not working

[silberzwiebel]

I tried to setup a transport to telegram with my raspberry pi. I'm using the spectrum version that comes in the apt repository (2.0.3, I guess) and the git version from telegram-purple.
While other transports work well, the telegram transport does not work.
Here is my telegram.cfg, my spectrum2.log and my backend.log.

I know there is #118, but my problem is different (I do not use any [purple] settings).

I appreciate any help.

[vitalyster]

does this plugin works for you with different purple "ui" (pidgin, finch, adium, bitlbee, etc)?

[silberzwiebel]

Yes, the plugin does work on a different machine (Fedora) with pidgin. I
do receive an authorization code in telegram while using pidgin, but I
do not receive anything (except an error message in the XMPP client)
when I'm using it with spectrum2.

Am 01.06.2016 um 09:19 schrieb vitalyster:

does this plugin works for you with different purple "ui" (pidgin, finch, adium,
bitlbee, etc)?

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#146 (comment), or mute
the thread
https://github.com/notifications/unsubscribe/AEUuZixYGdr4ncUokgUwVXSD-mHSrBMcks5qHTJwgaJpZM4IpEzR.

[thundergreen]

I'm facing the same issue

[herndlm]

I seem to have the same problem as @silberzwiebel and @thundergreen.. I guess you two still have no solution?
My spectrum2.log, backend-1.log and additionally I managed to get the following coredump:

PID: 817 (spectrum2_libpu)
           UID: 0 (root)
           GID: 0 (root)
        Signal: 6 (ABRT)
     Timestamp: Fri 2016-10-21 13:53:08 CEST (10min ago)
  Command Line: /usr/bin/spectrum2_libpurple_backend --host 127.0.0.1 --port 30902 --service.backend_id=1 -j telegram.<DOMAIN> /etc/spectrum2/transports/telegram.cfg
    Executable: /usr/bin/spectrum2_libpurple_backend
 Control Group: /user.slice/user-1001.slice/session-c9.scope
          Unit: session-c9.scope
         Slice: user-1001.slice
       Session: c9
     Owner UID: 1001 (martin)
       Boot ID: <ID>
    Machine ID: <ID>
      Hostname: raspi2
      Coredump: /var/lib/systemd/coredump/core.spectrum2_libpu.0.d5fa9cafcd94492aa9c6bdd27b4b14f6.817.1477050788000000000000.lz4
       Message: Process 817 (spectrum2_libpu) of user 0 dumped core.

                Stack trace of thread 817:
                #0  0x00000000767d01bc raise (libc.so.6)

gdb backtrace is

(gdb) bt
#0  0x767d01bc in raise () at /usr/lib/libc.so.6
#1  0x767d1520 in abort () at /usr/lib/libc.so.6
#2  0x767c8fa4 in __assert_fail_base () at /usr/lib/libc.so.6
#3  0x767c9080 in __assert_fail () at /usr/lib/libc.so.6
#4  0x701dc3d4 in bn_factorize () at /usr/lib/purple-2/telegram-purple.so
#5  0x701b1bfc in  () at /usr/lib/purple-2/telegram-purple.so

Also I noticed that I get an error message from Telegram if I enter my phone number not as expected (Unable to sign on as xxx: phone number lacks country prefix. Numbers must start with the full international prefix code, e.g. +1 for USA.), so it seems the communication with Telegram in general is working, but maybe Spectrum/libpurple is crashing because of an unexpected response or smth like that?

I would like to help you out with this, please guide me to compile it with debug info if you need that for testing

[herndlm]

After recompilling telegram-purple I got this from the backtrace:

(gdb) bt
#0  0x767c11bc in raise () at /usr/lib/libc.so.6
#1  0x767c2520 in abort () at /usr/lib/libc.so.6
#2  0x767b9fa4 in __assert_fail_base () at /usr/lib/libc.so.6
#3  0x767ba080 in __assert_fail () at /usr/lib/libc.so.6
#4  0x701cd3d4 in BN2ull (b=b@entry=0x65) at mtproto-utils.c:101
#5  0x701cd3d4 in bn_factorize (pq=pq@entry=0x16ab368, p=p@entry=0x768d2078 <lock>, q=q@entry=0x16ab6a0) at mtproto-utils.c:127
#6  0x701a2bfc in send_req_dh_packet (temp_key=0, pq=0x16ab368, c=0x16ab540, TLS=0x2) at mtproto-client.c:256
#7  0x701a2bfc in process_respq_answer (TLS=0x2, TLS@entry=0x16a7928, c=c@entry=0x16ab540, len=len@entry=84, temp_key=temp_key@entry=0, packet=0x7067ef38 <Response.14694> "")
    at mtproto-client.c:407
#8  0x701a59a4 in rpc_execute (TLS=0x16a7928, c=0x16ab540, op=0, len=84) at mtproto-client.c:1168
#9  0x70193828 in conn_try_read () at /usr/lib/purple-2/telegram-purple.so
#10 0x00052b14 in _start ()

Maybe @BenWiederhake can help?
Btw. I'm using Linux 4.4.24-2 on armv7l

[BenWiederhake]

You're running on a 32-bit platform, I guess? In that case, I recognize the problem, it has been fixed quite a while ago. Try the current dev-1.3.0 branch. Don't forget to git submodule update --init every now and then!
Also, thanks for the attribution, but @majn wrote this fantastic plugin, I only answer a few questions sometimes. :)

EDIT: Specifically, you'll need the "Update tgl (32-bit BN2ull)" commit, which literally just updates tgl: vysheng/tgl@47b571f

[herndlm]

Thank you for the quick help! I re-compiled the latest version from that branch and it is looking better now. But unfortunately there is another segfault now related to libgcrypt:

PID: 14923 (spectrum2_libpu)
           UID: 0 (root)
           GID: 0 (root)
        Signal: 6 (ABRT)
     Timestamp: Fri 2016-10-21 15:29:47 CEST (11min ago)
  Command Line: /usr/bin/spectrum2_libpurple_backend --host 127.0.0.1 --port <PORT> --service.backend_id=1 -j telegram.<DOMAIN> /etc/spectrum2/transports/telegram.cfg
    Executable: /usr/bin/spectrum2_libpurple_backend
 Control Group: /user.slice/user-1001.slice/session-c9.scope
          Unit: session-c9.scope
         Slice: user-1001.slice
       Session: c9
     Owner UID: 1001 (martin)
       Boot ID: d5fa9cafcd94492aa9c6bdd27b4b14f6
    Machine ID: 5890fdfecaaa435a8e17b789ddc2ea1a
      Hostname: raspi2
      Coredump: /var/lib/systemd/coredump/core.spectrum2_libpu.0.d5fa9cafcd94492aa9c6bdd27b4b14f6.14923.1477056587000000000000.lz4
       Message: Process 14923 (spectrum2_libpu) of user 0 dumped core.

backtrace is not really revealing atm I guess...:

(gdb) bt
#0  0x755a64cc in  () at /usr/lib/libgcrypt.so.20
#1  0x796b53e0 in  ()

[herndlm]

I'm seeing some progress.. I tried downgrading libgcrypt from 1.7.3 to 1.6.5 (and also 1.7.2) and it's not segfaulting now, but when trying to register with my number I get: "Unable to sign on as : problem in the underlying library 'tgl'. Please submit a bug report with the debug log."
The relevant log entries are:

10381: 2016-10-21 18:16:48,888 INFO  libpurple: prpl-telegram:Init gcrypt
10381: 2016-10-21 18:16:48,889 INFO  libpurple: prpl-telegram:Init gcrypt: version mismatch!
10381: 2016-10-21 18:16:48,889 INFO  libpurple: prpl-telegram:Call to tgl_init failed.
10381: 2016-10-21 18:16:48,889 INFO  libpurple: connection:Connection error on 0x2100888 (reason: 7 description: Unable to sign on as <PHONE_NUMBER>: problem in the underlying library 'tgl'. Please submit a bug report with the debug log.)

After re-compiling telegram-purple it starts segfaulting again in libgcrypt. Also with the older version. Don''t know what I can do now I guess. @BenWiederhake sorry for bothering again, do you think you can help me out?

[BenWiederhake]

This is really the wrong place to discuss this. Either report it at tgl or telegram-purple.
Apart from that, the error message already tells you what's the problem: gcrypt version mismatch. Go fix that. If there's any more issues, create a new issue at the correct repository. I won't respond in this issue anymore.