I am seeing malloc and free errors in the kstar and kstar-bounded examples with a "bad energy" confspace that I have added. This causes the program to crash. When I switch to the CPUConfEnergyCalculator, the errors do not occur, leading me to think that there's a bug in the NativeConfEnergyCalculator. If you encounter this issue, and until we fix the bug, a reasonable work around is to swap out the use of the NativeConfEnergyCalculator with the java-based CPUConfEnergyCalculator.
Example errors include:
Apr 02, 2023 6:46:48 PM com.hazelcast.spi.impl.servicemanager.impl.ServiceManagerImpl
WARNING: [192.168.1.39]:5701 [COFFEE-job-470742957] [4.0] Replacing ServiceInfo{name='ClusterZMatrix', service=edu.duke.cs.osprey.coffee.zmat.ClusterZMatrix@20134094} with ServiceInfo{name='ClusterZMatrix', service=edu.duke.cs.osprey.coffee.ClusterMember$UnregisteredService@6c15e8c7}
COFFEE-0: G [-1060.386,-1059.820] width 0.565875 of 0.000000 nodedb 0.0% time 510.04 ms
COFFEE-0: G [-1060.386,-1060.313] width 0.073162 of 0.000000 nodedb 0.0% time 1.01 s
COFFEE-0: target: [-1060.385844,-1060.312682] (w 0.073162) kcal/mol
COFFEE-0: Processing state design, sequence [1171 ASN=asn 1172 GLY=gly 1173 THR=thr 1174 THR=thr 1175 GLY=gly 1176 ILE=ile 301 HIS=his 341 HIS=his]
COFFEE-0: G [-Infinity,Infinity] width Infinity of Infinity nodedb 3.9% time 504.15 ms
free(): invalid pointer
and
reading energy matrix from file: emat.complex.dat
computing K* scores for 1 sequences to epsilon = 0.683 ...
WARNING: Using incubator modules: jdk.incubator.foreign
Fatal glibc error: malloc assertion failure in _int_malloc: (unsigned long) (size) >= (unsigned long) (nb)
Fatal glibc error: malloc assertion failure in _int_malloc: (unsigned long) (size) >= (unsigned long) (nb)
Fatal glibc error: malloc assertion failure in _int_malloc: (unsigned long) (size) >= (unsigned long) (nb)
Fatal glibc error: malloc assertion failure in _int_malloc: (unsigned long) (size) >= (unsigned long) (nb)
> Task :Main.main() FAILED
The compiled conformation spaces causing these failures in the NativeConfEnergyCalculator are attached.
confspaces.zip
I am seeing
mallocandfreeerrors in the kstar and kstar-bounded examples with a "bad energy" confspace that I have added. This causes the program to crash. When I switch to the CPUConfEnergyCalculator, the errors do not occur, leading me to think that there's a bug in the NativeConfEnergyCalculator. If you encounter this issue, and until we fix the bug, a reasonable work around is to swap out the use of the NativeConfEnergyCalculator with the java-based CPUConfEnergyCalculator.Example errors include:
and
The compiled conformation spaces causing these failures in the NativeConfEnergyCalculator are attached.
confspaces.zip