Hi John - I noticed a mutex is being acquired recursively in 
fileManger::tryGetFile, using -DFASTBIT_RECURSIVE_UNLOAD, latest release 2.0.3

line 1069: mutex is acquired
line 1161: call to storage::enlarge
line 2212: call to storage::storage
line 1876: mutex is acquired again

Without -DFASTBIT_RECURSIVE_UNLOAD, there is an intermittent std::bad_alloc 
when maxBytes is reached while trying to load the index.

In addition, ibis::column::loadIndex handles the failure by calling purgeIndex 
at line 5872, deleting the index files.  We don’t distribute the raw data 
files, so subsequent queries fail to reconstruct the deleted index.
_______________________________________________
FastBit-users mailing list
[email protected]
https://hpcrdm.lbl.gov/cgi-bin/mailman/listinfo/fastbit-users

Reply via email to