Details
Description
Hi, we get some "python: free(): invalid next size (fast)" in PyROOT 6.06.02 and patches.
Unfortunately this is there since LCG84, I thought I had tested it but apparently I was wrong. This has also gone unnoticed in the nightlies because the test suite we run in the nightlies is tuned to be fast, so many important tests are disabled. Now I have reenabled a short test that should show this also in the next dev4 nightlies.
The issue happens in the pycool ifolder tests, it looks like this
Using QMTEST_CLASS_PATH=/afs/cern.ch/sw/lcg/app/nightlies/dev4/Sat/COOL/3_1-patches/x86_64-slc6-gcc49-dbg/CoolTest/qmtest
|
|
|
Using CORALCOOL_QMTEST_WORKDIR=/tmp/avalassi/Sat/COOL/3_1-patches/tmp.x86_64-slc6-gcc49-dbg
|
|
|
Launch tests - results will be in /tmp/avalassi/Sat/COOL/3_1-patches/logs/qmtestCool/x86_64-slc6-gcc49-dbg.qmr
|
|
|
Launch 'qmtest run pycool.sqlite.ifolder'
|
|
|
*** glibc detected *** python: free(): invalid next size (fast): 0x0000000009fc3ee0 ***
|
|
|
======= Backtrace: =========
|
|
|
/lib64/libc.so.6[0x3971675f3e]
|
|
|
/lib64/libc.so.6[0x3971678dd0]
|
|
|
/afs/cern.ch/sw/lcg/app/nightlies/dev4/Sat/ROOT/v6-06-00-patches/x86_64-slc6-gcc49-dbg/lib/libPyROOT.so(_ZN5Cppyy10DeallocateElPv+0x1c)[0x7fdb07a35017]
|
|
|
/afs/cern.ch/sw/lcg/app/nightlies/dev4/Sat/ROOT/v6-06-00-patches/x86_64-slc6-gcc49-dbg/lib/libPyROOT.so(_ZN6PyROOT17op_dealloc_nofreeEPNS_11ObjectProxyE+0xb0)[0x7fdb07a28d1c]
|
|
|
/afs/cern.ch/sw/lcg/app/nightlies/dev4/Sat/ROOT/v6-06-00-patches/x86_64-slc6-gcc49-dbg/lib/libPyROOT.so(+0xbe4c0)[0x7fdb07a294c0]
|
|
|
/afs/cern.ch/sw/lcg/releases/Python/2.7.10-806f9/x86_64-slc6-gcc49-dbg/bin/../lib/libpython2.7.so.1.0(+0xb7bc2)[0x7fdb10239bc2]
|
Note that this even breaks my test driver (qmtest), it is a rather catastrophic failure.
I also got a similar signature in oracle tests, where the issue is intercepted by the oracle client signal handler
test_listChannelsWithNames (__main__.TestIFolder) ... ORA-24550: signal received: [si_signo=6] [si_errno=0] [si_code=-6] [si_int=0] [si_ptr=(nil)] [si_addr=0x38d200005a3b]
|
|
|
kpedbg_dmp_stack()+360<-kpeDbgCrash()+192<-kpeDbgSignalHandler()+119<-skgesig_sigactionHandler()+218<-__sighandler()<-gsignal()+53<-malloc_printerr()+190<-_int_free()+1579<-_ZN5Cppyy10DeallocateElPv()+23<-_ZN6PyROOT17op_dealloc_nofreeEPNS_11ObjectProxyE()+176
|
<-_ZN6PyROOT12_GLOBAL__N_110op_deallocEPNS_11ObjectProxyE()+24<-subtype_dealloc()+1122<-00007FC738D2FF80<-00007FC738D53450<-00007FC74C8A5680
|
<-0000000009DE98B0<-00007FC738D317C0<-00007FC738D31818<-00000000022BB510<-00007FC738D34660<-00007FC738D40B20<-00007FC738D346A8
|
<-00007FC738D2F5D0<-00007FC738D31870<-00007FC738D3EBF0
|
Let's see in the nightlies tonight. If confirmed I will send you a reproducer.
Thanks
Andrea
Attachments
Issue Links
- is triggered by
-
ROOT-8273 TClass::Size() vs TEmulatedCollectionProxy::Sizeof()
-
- Open
-