I don't think that the bug is related to threading/locking failures within mia, i.e. compiling the package by using g++-5* and -fsanitize=thread doesn't show any locking problems (specifically no double un-locking of a mutex).
Specifically, the error hints at ./mia-3dmaskseeded --help-xml <output> not generating the required output that is later used by the python program that actually fails. And this command is run without reported errors, i.e. it doesn't crash like it seems to be what would be expected from threading errors that are revealed when TSX is enabled. In addition, the only place in the code where mutex locking is applied when running the command like given above is a piece of code that is shared by all other command line tools in mia, and running these tools with the the same command line option (--help-xml) don't fail. It would be nice to know whether the failure was exactly at the same point when the build was tried, or whether there were differences there? Are there any other options, apart from using the thread sanitizer, that can help me to debug this problem? Best, Gert * g++-5, because for g++-6 on Debian the thread sanitizer is currently broken (#796246).