0.14rc1 coredump in QScreen::handle() #9814

I just went to shut down my computer and noticed the bitcoin-qt 0.14rc1 had crashed. I was running it inside gdb, so can show the stack traces. I had opened up the debug console window, but hadn't typed anything in it. I think I had left it on the first ("info"?) tab.

• Stack traces

• Last 100 lines of debug.log (I use debug=1):

• Last lines from grep GUI debug.log

Thanks for the report! I'll have a closer look.

@dooglus: Can you tell me about your operating system? The stack trace tells me you are using linux? Ubuntu? Which version? Did you self compile or used the pre-compiled binaries? If self-compiled, what Qt version?

@jonasschnelli: It's Debian Stable:

Description:	Debian GNU/Linux 8.6 (jessie)
Release:	8.6
Codename:	jessie

I built it myself from git from the 0.14 branch, commit 2afefeade6e668cab0643a5c22ff203ce1e509b0. I ran these commands:

./autogen.sh
./configure --with-incompatible-bdb
cd src
make qt/bitcoin-qt bitcoind bitcoin-cli

The configure step mentioned QT:

checking for QT... yes
checking for QT_TEST... yes
checking for QT_DBUS... yes
checking for static Qt... no
checking whether -fPIE can be used with this Qt config... no
checking for moc-qt5... no

Options used to compile and link:
  with wallet   = yes
  with gui / qt = yes
    qt version  = 5
    with qr     = yes

The link line used to create bitcoin-qt:

libtool: link: g++ -std=c++11 -pthread -Wl,-z -Wl,relro -Wl,-z -Wl,now -pie -o qt/bitcoin-qt qt/qt_bitcoin_qt-bitcoin.o -pthread  qt/libbitcoinqt.a libbitcoin_server.a libbitcoin_wallet.a libbitcoin_zmq.a -lzmq libbitcoin_cli.a libbitcoin_common.a libbitcoin_util.a libbitcoin_consensus.a crypto/libbitcoin_crypto.a univalue/.libs/libunivalue.a leveldb/libleveldb.a leveldb/libmemenv.a -L/usr/lib/x86_64-linux-gnu -lboost_system -lboost_filesystem -lboost_program_options -lboost_thread -lboost_chrono -lQt5Network -lQt5Widgets -lQt5Gui -lQt5DBus -lQt5Core -lqrencode -lprotobuf -lpthread -ldb_cxx -lssl -lcrypto secp256k1/.libs/libsecp256k1.a -levent_pthreads -levent -pthread

Checking which versions of Qt5Core I have:

$ locate libQt5Core
/home/chris/Downloads/TeamSpeak3-Client-linux_amd64/libQt5Core.so.5
/usr/lib/x86_64-linux-gnu/libQt5Core.prl
/usr/lib/x86_64-linux-gnu/libQt5Core.so
/usr/lib/x86_64-linux-gnu/libQt5Core.so.5
/usr/lib/x86_64-linux-gnu/libQt5Core.so.5.3
/usr/lib/x86_64-linux-gnu/libQt5Core.so.5.3.2

Checking the versions that are installed:

$ dpkg -l | grep 'qt.*5.*-dev[ :]'
ii  libqt5opengl5-dev:amd64             5.3.2+dfsg-4+deb8u1    amd64                  Qt 5 OpenGL library development files
ii  qtbase5-dev:amd64                   5.3.2+dfsg-4+deb8u1    amd64                  Qt 5 base development files
ii  qttools5-dev:amd64                  5.3.2-3                amd64                  Qt 5 tools development files

Also note that this isn't the first time I saw the 'does not have a handle' warnings. My debug.log goes back quite a long way and I see a bunch of the same errors on the 15th/16th Feb.

Edit: although it appears that on the 16th I did a clean shutdown rather than it crashing.

I just had it crash again.

Here's the stack trace.

I was in the debug console. I had run getblockhash 10, double-clicked the resulting hash, and hit Control-C to copy it when it froze.

I done this shortly before the crash, while bitcoin-qt was still running:

chris@chris:~/Programs/bitcoin.git/src/qt$ touch libbitcoinqt.a
chris@chris:~/Programs/bitcoin.git/src/qt$ cd ..
chris@chris:~/Programs/bitcoin.git/src$ V=1 make qt/bitcoin-qt
make  -C secp256k1 libsecp256k1.la
make[1]: Entering directory '/home/chris/Programs/bitcoin.git/src/secp256k1'
make[1]: Leaving directory '/home/chris/Programs/bitcoin.git/src/secp256k1'
/bin/bash ../libtool  --tag CXX  --mode=link g++ -std=c++11    -pthread  -Wl,-z,relro -Wl,-z,now -pie    -o qt/bitcoin-qt qt/qt_bitcoin_qt-bitcoin.o   qt/libbitcoinqt.a libbitcoin_server.a libbitcoin_wallet.a libbitcoin_zmq.a -lzmq  libbitcoin_cli.a libbitcoin_common.a libbitcoin_util.a libbitcoin_consensus.a crypto/libbitcoin_crypto.a univalue/libunivalue.la leveldb/libleveldb.a leveldb/libmemenv.a -L/usr/lib/x86_64-linux-gnu -lboost_system -lboost_filesystem -lboost_program_options -lboost_thread -lboost_chrono -lQt5Network -lQt5Widgets -lQt5Gui -lQt5Core  -lQt5DBus -lQt5Core  -lqrencode  -lprotobuf -pthread -lpthread  -ldb_cxx -lssl -lcrypto  -lcrypto   secp256k1/libsecp256k1.la -levent_pthreads -levent  -levent  
libtool: link: g++ -std=c++11 -pthread -Wl,-z -Wl,relro -Wl,-z -Wl,now -pie -o qt/bitcoin-qt qt/qt_bitcoin_qt-bitcoin.o -pthread  qt/libbitcoinqt.a libbitcoin_server.a libbitcoin_wallet.a libbitcoin_zmq.a -lzmq libbitcoin_cli.a libbitcoin_common.a libbitcoin_util.a libbitcoin_consensus.a crypto/libbitcoin_crypto.a univalue/.libs/libunivalue.a leveldb/libleveldb.a leveldb/libmemenv.a -L/usr/lib/x86_64-linux-gnu -lboost_system -lboost_filesystem -lboost_program_options -lboost_thread -lboost_chrono -lQt5Network -lQt5Widgets -lQt5Gui -lQt5DBus -lQt5Core -lqrencode -lprotobuf -lpthread -ldb_cxx -lssl -lcrypto secp256k1/.libs/libsecp256k1.a -levent_pthreads -levent -pthread

to get the link line I pasted in my previous comment. I thought that replacing a running binary with a new copy was supposed to be safe, but am not 100% sure of that, so maybe the crash was caused by me rebuilding bitcoin-qt while it was running.

Edit: I've tried repeating the same steps again and wasn't able to get the crash to happen again.

I just tried building rc2 but it failed:

$ make qt/bitcoin-qt bitcoind bitcoin-cli
  CXX      qt/qt_bitcoin_qt-bitcoin.o
  CXX      qt/qt_libbitcoinqt_a-bitcoingui.o
  CXX      qt/qt_libbitcoinqt_a-clientmodel.o
  CXX      qt/qt_libbitcoinqt_a-optionsdialog.o
  CXX      qt/qt_libbitcoinqt_a-optionsmodel.o
  CXX      qt/qt_libbitcoinqt_a-peertablemodel.o
  CXX      qt/qt_libbitcoinqt_a-splashscreen.o
  CXX      qt/qt_libbitcoinqt_a-addresstablemodel.o
  CXX      qt/qt_libbitcoinqt_a-coincontroldialog.o
  CXX      qt/qt_libbitcoinqt_a-paymentserver.o
  CXX      qt/qt_libbitcoinqt_a-sendcoinsdialog.o
  CXX      qt/qt_libbitcoinqt_a-signverifymessagedialog.o
  CXX      qt/qt_libbitcoinqt_a-transactiondesc.o
  CXX      qt/qt_libbitcoinqt_a-transactionrecord.o
  CXX      qt/qt_libbitcoinqt_a-transactiontablemodel.o
  CXX      qt/qt_libbitcoinqt_a-walletmodel.o
  CXX      qt/qt_libbitcoinqt_a-walletmodeltransaction.o
  GEN      qt/locale/bitcoin_en.qm
  GEN      qt/locale/bitcoin_fr_FR.qm
  GEN      qt/locale/bitcoin_zh_CN.qm
  GEN      qt/qrc_bitcoin_locale.cpp
  CXX      qt/qt_libbitcoinqt_a-qrc_bitcoin_locale.o
  AR       qt/libbitcoinqt.a
  CXX      libbitcoin_server_a-blockencodings.o
In file included from /usr/include/boost/mpl/aux_/include_preprocessed.hpp:37:0,
                 from /usr/include/boost/mpl/aux_/comparison_op.hpp:35,
                 from /usr/include/boost/mpl/greater_equal.hpp:19,
                 from /usr/include/boost/mpl/comparison.hpp:22,
                 from /usr/include/boost/serialization/version.hpp:76,
                 from /usr/include/boost/multi_index/detail/serialization_version.hpp:18,
                 from /usr/include/boost/multi_index_container.hpp:54,
                 from txmempool.h:24,
                 from blockencodings.cpp:12:
/usr/include/boost/mpl/aux_/preprocessed/gcc/greater_equal.hpp:78:1: internal compiler error: Segmentation fault
 BOOST_MPL_AUX_NA_SPEC2(2, 2, greater_equal)
 ^
Please submit a full bug report,
with preprocessed source if appropriate.
See <file:///usr/share/doc/gcc-4.9/README.Bugs> for instructions.
The bug is not reproducible, so it is likely a hardware or OS problem.
Makefile:5262: recipe for target 'libbitcoin_server_a-blockencodings.o' failed
make: *** [libbitcoin_server_a-blockencodings.o] Error 1

Sounds like there's something wrong my end. Maybe faulty RAM or hdd. So the crashes I saw could be symptom of same.

Repeating the make command a 2nd time worked.

Repeating the make command a 2nd time worked.

Indeed, gcc is extremely stable and actual crashes in gcc are almost always CPU overheating/memory corruption related. Especially if it goes away when you repeat it.

Actually the 2nd time I ran 'make' it got past the .cpp file that previously crashed, but then hung for 10 minutes on a different .cpp file, not using any CPU or disk I/O. The compiler just didn't ever exit.

The 3rd time I ran it everything worked and I ended up with a working bitcoin-qt executable.

It's odd. I never see my web browser crashing, and it is usually using almost all available RAM. But I've seen gcc crash and/or hang many times since I got this laptop. I do think that this invalidates my bug report though, since nobody else is seeing the same crash and my computer is pretty clearly unreliable.

I agree. Thanks for reporting that, it could save hours of chasing ghosts :) Would probably make sense to close this until we get at least one other report.