https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105240
Ian Lance Taylor <ian at airs dot com> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |ian at airs dot com
--- Comment #3 from Ian Lance Taylor <ian at airs dot com> ---
mmap is definitely preferred. The malloc implementation is only there to
support systems without mmap.
That said I'm not sure I understand the valgrind report. Some of that memory
is not lost as long as the program has a pointer to the backtrace_state. It's
true that there is no way to release all memory allocated by the state. That
is not a goal, as I expect that programs will allocate a backtrace_state and
use it until the program is complete. What kinds of reports do you see if you
put the backtrace_state in a global variable?