aborts due to bad lexical cast

[Igor Vlasenko]

qbittorrent 1.3.2 + rasterbar 0.14.2
qbittorrent sometimes crashes
when seeding a lot of torrents (about 100)
terminate called after throwing an instance of 'boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<boost::bad_lexical_cast> >'
  what(): bad lexical cast: source type value could not be interpreted as target

Program received signal SIGABRT, Aborted.

I built a debug versions of qbittorrent 1.3.2 + rasterbar 0.14.2,
and tried ddd and kdbg. Unfortunately, it seems to be not good for multithread debuging,
so I didn't find the exact location :(

[Christophe Dumez (hydr0g3n)]

What version of boost are you using?

I have absolutely no idea of where this bug could be in qbittorrent's code. I will try to ask libtorrent author in case he has an idea (could likely be in his code).

a gdb backtrace would also be more than welcome :) Did you compile it yourself or did you use packages?
Would you be able to recompile libtorrent with debug enabled?

[Christophe Dumez (hydr0g3n)]

Forwarded to libtorrent-rasterbar author. Libtorrent-rasterbar *does* use lexical_cast from boost library. qBittorrent does not.

[Christophe Dumez (hydr0g3n)]

Libtorrent's author would really need a backtrace. Could you please do the following:
$ gdb qbittorrent
> catch throw (if it fails, try: "break _cxa_throw")
Make breakpoint pending on future shared library load? (y or [n]) y
> run
... Wait until it hits the breackpoint
> bt full

Then paste the backtrace here. Thanks in advance.

[Igor Vlasenko (viy-altlinux)]

1) boost = 0.36.0
2) arch = x86_64
libtorrent-rasterbar 0.14.2 built with Og, not stripped:
file /usr/lib64/libtorrent-rasterbar.so.2.0.0
/usr/lib64/libtorrent-rasterbar.so.2.0.0: ELF 64-bit LSB shared object, x86-64, version 1 (SYSV), not stripped

thanks for help, I run qbittorrent again under gdb as requested,
now I am waiting for bug to happen.
When I have no or a few uploads, nothing bad happens :(

[Igor Vlasenko (viy-altlinux)]

how I started gdb.
-------------------------------------
GNU gdb 6.6-alt3 (ALT Linux)
...
This GDB was configured as "x86_64-alt-linux"...
Using host libthread_db library "/lib64/libthread_db.so.1".
(gdb) catch throw
Function "__cxa_throw" not defined.
(gdb) break _cxa_throw
Function "_cxa_throw" not defined.
Make breakpoint pending on future shared library load? (y or [n]) y

Breakpoint 1 (_cxa_throw) pending.
(gdb) run
Starting program: /home/archive/gitrepo/qbittorrent.git/qbittorrent/qbittorrent
[Thread debugging using libthread_db enabled]
[New Thread 139788082653008 (LWP 30680)]
[New Thread 1113073984 (LWP 30683)]
[New Thread 1084275008 (LWP 30684)]
[New Thread 1092667712 (LWP 30689)]
[New Thread 1101060416 (LWP 30690)]

[Igor Vlasenko (viy-altlinux)]

nothing interesting was in backtrace.
Maybe I should try to built
libtorrent-rasterbar as a static library and link it to qbittorrent stratically.
....
QLayout: Attempting to add QLayout "" to properties "properties", which already has a layout
[New Thread 1121466688 (LWP 30717)]
[Thread 1121466688 (LWP 30717) exited]
QLayout: Attempting to add QLayout "" to properties "properties", which already has a layout
[New Thread 1121466688 (LWP 30718)]
[Thread 1121466688 (LWP 30718) exited]
QLayout: Attempting to add QLayout "" to properties "properties", which already has a layout
[New Thread 1121466688 (LWP 30727)]
[Thread 1121466688 (LWP 30727) exited]
QLayout: Attempting to add QLayout "" to properties "properties", which already has a layout
[New Thread 1121466688 (LWP 30730)]
[Thread 1121466688 (LWP 30730) exited]
QLayout: Attempting to add QLayout "" to properties "properties", which already has a layout
[New Thread 1121466688 (LWP 30738)]
[Thread 1121466688 (LWP 30738) exited]
QLayout: Attempting to add QLayout "" to properties "properties", which already has a layout
[New Thread 1121466688 (LWP 30739)]
[Thread 1121466688 (LWP 30739) exited]
QLayout: Attempting to add QLayout "" to properties "properties", which already has a layout
[New Thread 1121466688 (LWP 30740)]
[Thread 1121466688 (LWP 30740) exited]
QLayout: Attempting to add QLayout "" to properties "properties", which already has a layout
[New Thread 1121466688 (LWP 30741)]
[Thread 1121466688 (LWP 30741) exited]
QLayout: Attempting to add QLayout "" to properties "properties", which already has a layout
[New Thread 1121466688 (LWP 30742)]
[Thread 1121466688 (LWP 30742) exited]
QLayout: Attempting to add QLayout "" to properties "properties", which already has a layout
[New Thread 1121466688 (LWP 30743)]
[Thread 1121466688 (LWP 30743) exited]
QLayout: Attempting to add QLayout "" to properties "properties", which already has a layout
[New Thread 1121466688 (LWP 30744)]
[Thread 1121466688 (LWP 30744) exited]
QLayout: Attempting to add QLayout "" to properties "properties", which already has a layout
[New Thread 1121466688 (LWP 30749)]
[Thread 1121466688 (LWP 30749) exited]
QLayout: Attempting to add QLayout "" to properties "properties", which already has a layout
[New Thread 1121466688 (LWP 30750)]
[Thread 1121466688 (LWP 30750) exited]
terminate called after throwing an instance of 'boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<boost::bad_lexical_cast> >'
  what(): bad lexical cast: source type value could not be interpreted as target

Program received signal SIGABRT, Aborted.
[Switching to Thread 1084275008 (LWP 30684)]
0x00007f22f035f665 in raise () from /lib64/libc.so.6
(gdb)

[Christophe Dumez (hydr0g3n)]

Backtrace with "bt full".

Did you try? What you posted is not a backtrace.

[Igor Vlasenko (viy-altlinux)]

sorry, I forget. shame on me :(
from another run:
....
Program received signal SIGABRT, Aborted.
[Switching to Thread 1096878400 (LWP 3701)]
0x00007f831d9f5665 in raise () from /lib64/libc.so.6
(gdb) bt full
#0 0x00007f831d9f5665 in raise () from /lib64/libc.so.6
No symbol table info available.
#1 0x00007f831d9f6ae0 in abort () from /lib64/libc.so.6
No symbol table info available.
#2 0x00007f831e24ebb4 in __gnu_cxx::__verbose_terminate_handler ()
   from /usr/lib64/libstdc++.so.6
No symbol table info available.
#3 0x00007f831e24cd76 in std::basic_streambuf<char, std::char_traits<char> >::setbuf () from /usr/lib64/libstdc++.so.6
No symbol table info available.
#4 0x00007f831e24cda3 in std::terminate () from /usr/lib64/libstdc++.so.6
No symbol table info available.
#5 0x00007f831d3a396f in thread_proxy ()
   from /usr/lib64/libboost_thread-mt.so.1.36.0
No symbol table info available.
#6 0x00007f831f643fb7 in start_thread () from /lib64/libpthread.so.0
No symbol table info available.
#7 0x00007f831da8278d in clone () from /lib64/libc.so.6
No symbol table info available.

[Igor Vlasenko (viy-altlinux)]

and before is
(gdb) break _cxa_throw
Function "_cxa_throw" not defined.
Make breakpoint pending on future shared library load? (y or [n]) y

Breakpoint 1 (_cxa_throw) pending.
(gdb) run
Starting program: /home/archive/gitrepo/qbittorrent.git/qbittorrent/qbittorrent
[Thread debugging using libthread_db enabled]
[New Thread 140201161394000 (LWP 3697)]
[New Thread 1113901376 (LWP 3700)]
[New Thread 1096878400 (LWP 3701)]
[New Thread 1105271104 (LWP 3706)]
Warning: Local domain socket was not closed cleanly, deleting file...
[New Thread 1122294080 (LWP 3723)]
terminate called after throwing an instance of 'boost::exception_detail::clone_i
mpl<boost::exception_detail::error_info_injector<boost::bad_lexical_cast> >'
  what(): bad lexical cast: source type value could not be interpreted as targe
t

Program received signal SIGABRT, Aborted.
[Switching to Thread 1096878400 (LWP 3701)]
0x00007f831d9f5665 in raise () from /lib64/libc.so.6

[Igor Vlasenko (viy-altlinux)]

the program is still running, other threads still alive.

[Igor Vlasenko (viy-altlinux)]

after crash main window becomes empty, but exists, there is also a tray icon, and some network activity continues.

[Christophe Dumez (hydr0g3n)]

what do you mean it is still running? You can't type anything in gdb?

backtrace for all threads is like this:
thread apply all bt

see https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/286205 more more information.

[Igor Vlasenko (viy-altlinux)]

it is a multithread program,
...
[New Thread 140201161394000 (LWP 3697)]
[New Thread 1113901376 (LWP 3700)]
[New Thread 1096878400 (LWP 3701)]
[New Thread 1105271104 (LWP 3706)]
...
so when one of them crashed, other threads was still alive
and the program as a whole was still running.
I did
(gdb) kill
Kill the program being debugged? (y or n) y
(gdb)
The program is not being run.
(gdb) quit

[Igor Vlasenko (viy-altlinux)]

thanks for help, now I will try
thread apply all bt

[Igor Vlasenko (viy-altlinux)]

[New Thread 1088903488 (LWP 24269)]
terminate called after throwing an instance of 'boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<boost::bad_lexical_cast> >'
  what(): bad lexical cast: source type value could not be interpreted as target

Program received signal SIGABRT, Aborted.
[Switching to Thread 1108678976 (LWP 24263)]
0x00007fa4a1673665 in raise () from /lib64/libc.so.6
(gdb) thread apply all bt
[empty lines where I cut the message (---Type <return> to continue, or q <return> to quit---)]
Thread 5 (Thread 1088903488 (LWP 24269)):
#0 0x00007fa4a16f86a6 in poll () from /lib64/libc.so.6
#1 0x00007fa4975c14c4 in std::basic_streambuf<char, std::char_traits<char> >::setbuf () from /lib64/libresolv.so.2
#2 0x00007fa4975c05c6 in __libc_res_nquery () from /lib64/libresolv.so.2
#3 0x00007fa4975c0837 in std::basic_streambuf<char, std::char_traits<char> >::setbuf () from /lib64/libresolv.so.2
#4 0x00007fa4975c0b60 in __libc_res_nsearch () from /lib64/libresolv.so.2
#5 0x00007fa4977cf44f in _nss_dns_gethostbyname3_r ()
   from /lib64/libnss_dns.so.2
#6 0x00007fa4977cf6c4 in _nss_dns_gethostbyname2_r ()
   from /lib64/libnss_dns.so.2
#7 0x00007fa4a17142e5 in gethostbyname2_r () from /lib64/libc.so.6
#8 0x00007fa4a16ed2fc in std::basic_streambuf<char, std::char_traits<char> >::setbuf () from /lib64/libc.so.6
#9 0x00007fa4a16eddcf in getaddrinfo () from /lib64/libc.so.6
#10 0x00007fa4a3b6b1d7 in boost::asio::detail::socket_ops::getaddrinfo ()
   from /usr/lib64/libtorrent-rasterbar.so.2
#11 0x00007fa4a3bfe242 in boost::asio::detail::resolver_service<boost::asio::ip::tcp>::resolve_query_handler<boost::_bi::bind_t<void, boost::_mfi::mf2<void, libtorrent::http_connection, boost::system::error_code const&, boost::asio::ip::basic_resolver_

e<boost::shared_ptr<libtorrent::http_connection> >, boost::arg<1>, boost::arg<2> > > >::operator() () from /usr/lib64/libtorrent-rasterbar.so.2
#12 0x00007fa4a3bfe3c1 in boost::asio::asio_handler_invoke<boost::asio::detail::resolver_service<boost::asio::ip::tcp>::resolve_query_handler<boost::_bi::bind_t<void, boost::_mfi::mf2<void, libtorrent::http_connection, boost::system::error_code const&, boost::asio::ip::basic_resolver_iterator<boost::asio::ip::tcp> >, boost::_bi::list3<boost::_bi::value<boost::shared_ptr<libtorrent::http_connection> >, boost::arg<1>, boost::arg<2> > > > > ()
   from /usr/lib64/libtorrent-rasterbar.so.2
#13 0x00007fa4a3bfe3e9 in boost_asio_handler_invoke_helpers::invoke<boost::asio::detail::resolver_service<boost::asio::ip::tcp>::resolve_query_handler<boost::_bi::bind_t<void, boost::_mfi::mf2<void, libtorrent::http_connection, boost::system::error_code const&, boost::asio::ip::basic_resolver_iterator<boost::asio::ip::tcp> >, boost::_bi::list3<boost::_bi::value<boost::shared_ptr<libtorrent::http_connection> >, boost::arg<1>, boost::arg<2> > > >, boost::asio::detail::resolver_service<boost::asio::ip::tcp>::resolve_query_handler<boost::_bi::bind_t<void, boost::_mfi::mf2<void, libtorrent::http_connection, boost::system::error_code const&, boost::asio::ip::basic_resolver_iterator<boost::asio::ip::tcp> >, boost::_bi::list3<boost::_bi::value<boost::shared_ptr...

[Christophe Dumez (hydr0g3n)]

Thanks a lot. I provided libtorrent author with your backtrace. Now we just have to let him do his magic :)

[Christophe Dumez (hydr0g3n)]

Apparently the break did not work.

Could you please try the following:
gdb qbittorrent
> break 120
Breakpoint 1 at 0x80b5e76: file main.cpp, line 120.
> run
... wait until it breaks
> catch throw (should work now that the program is loaded)
... wait until it catches a throw
> bt full
> c (to continue and make sure the throw was due to a bad_lexical_cast, if not, wait until it break again and type bt full again.)

[Igor Vlasenko (viy-altlinux)]

In my binary function main has no line number information, so I used
break main (It seems the same).

it looks like the proper dump now, as I observe
...
#3 0x00000000004b640b in boost::lexical_cast<long, std::string> ()
#4 0x00007f78543969b4 in libtorrent::detail::bdecode_recursive<char const*> ()
   from /usr/lib64/libtorrent-rasterbar.so.2
...
----------------------------------------------------------------------
(gdb) break main
Breakpoint 1 at 0x479282
(gdb) run
Starting program: /home/archive/gitrepo/qbittorrent.git/qbittorrent/qbittorrent
[Thread debugging using libthread_db enabled]
[New Thread 140154793432912 (LWP 25652)]
[Switching to Thread 140154793432912 (LWP 25652)]

Breakpoint 1, 0x0000000000479282 in main ()
(gdb) catch throw
Catchpoint 2 (throw)
(gdb) continue
Continuing.
[New Thread 1086691648 (LWP 25655)]
[New Thread 1095084352 (LWP 25656)]
[New Thread 1103477056 (LWP 25661)]
[New Thread 1111869760 (LWP 25662)]
Catchpoint 2 (exception thrown)
0x00007f785265de30 in __cxa_throw () from /usr/lib64/libstdc++.so.6
(gdb) bt full
#0 0x00007f785265de30 in __cxa_throw () from /usr/lib64/libstdc++.so.6
No symbol table info available.
#1 0x00000000004b635a in boost::throw_exception<boost::bad_lexical_cast> ()
No symbol table info available.
#2 0x00000000004b63b5 in boost::detail::lexical_cast<long, std::string, false, char> ()
No symbol table info available.
#3 0x00000000004b640b in boost::lexical_cast<long, std::string> ()
No symbol table info available.
#4 0x00007f78543969b4 in libtorrent::detail::bdecode_recursive<char const*> ()
   from /usr/lib64/libtorrent-rasterbar.so.2
No symbol table info available.
#5 0x00007f7854396ce9 in libtorrent::bdecode<char const*> ()
   from /usr/lib64/libtorrent-rasterbar.so.2
No symbol table info available.
#6 0x00007f7854394ebe in libtorrent::http_tracker_connection::on_response ()
   from /usr/lib64/libtorrent-rasterbar.so.2
No symbol table info available.
#7 0x00007f7854395b2c in boost::_mfi::mf4<void, libtorrent::http_tracker_connection, boost::system::error_code const&, libtorrent::http_parser const&, char const*, int>::operator()<boost::intrusive_ptr<libtorrent::http_tracker_connection> > () from /usr/lib64/libtorrent-rasterbar.so.2
No symbol table info available.
#8 0x00007f7854395b82 in boost::_bi::bind_t<void, boost::_mfi::mf4<void, libtorrent::http_tracker_connection, boost::system::error_code const&, libtorrent::http_parser const&, char const*, int>, boost::_bi::list5<boost::_bi::value<boost::intrusive_ptr<libtorrent::http_tracker_connection> >, boost::arg<1>, boost::arg<2>, boost::arg<3>, boost::arg<4> > >::operator()<boost::system::error_code const, libtorrent::http_parser const, char const*, int, libtorrent::http_connection>
    () from /usr/lib64/libtorrent-rasterbar.so.2
No symbol table info available.
#9 0x00007f7854395ba7 in boost::detail::function::void_function_obj_invoker5<boost::_bi::bind_t<void, boost::_mfi::mf4<void, libtorrent::http_tracker_connection, boost::system::error_code const&, libtorrent::http_parser const&, char const*, int>, boost::_bi::list5<boost::_bi::value<boost::intrusive_ptr<libtorrent::http_tracker_connection> >, boost::arg<1>, boost::arg<2>...

[Igor Vlasenko (viy-altlinux)]

I've rebuilt qbittorrent with latest rasterbar svn (r3344) in 0_14 branch.
Everything seems to be ok, they do fixed the bug.

[Christophe Dumez (hydr0g3n)]

Good news. I'm marking this bug as invalid since it does not affect qBittorrent itself.

Thank you for your report.