This applie to memusagestat.c of the latest git version. I'm not sure if the usage of memusagestat.c is supported and if I'm seeing a bug and if patches are welcome. Using the -T option on a memusage "trace" from an application that was halted and didn't write out the header entry results in a SIGFPE due "maxsize_total" being 0. If a patch is welcome I would change the "repair" logic to update maxsize total header entry as well, write the whole headent structure to disk. And i would move the repair above assigning the maxsize_* variables. I can upload the data file, my simple test application and a gdb backtrace if that is required/wanted/wished. The other observation is that the -t option is not doing anything, from looking at the trace (this is not from glibc git... but the distro version) is that the time sometimes jump backwards. Does this classify as opening a new bug?
I've checked in a patch. There were a number of other issues. Wrt -t: I don't know what you refer to. You have to be more concrete.
Thanks a lot. I will recompile and create another report for the time going backwards in the profile (at least this is what I think I see).