Hello Assaf > It is a long tradition not to bother freeing memory in such cases
Yes. I remind the distinction between "not serious" and "serious" memory leaks (as defined in https://en.wikipedia.org/wiki/Memory_leak ), and I don't intend to do anything about the not serious ones. Why? 1) Because there are cases where one-time allocation of memory is normal. For example the functions textdomain() and bindtextdomain() do allocate memory that stays allocated until the application ends. This is not a bug, because it does not lead to a serious memory leak. 2) In Gnulib we want to have a large number of unit tests. To achieve this, we should not make it too annoying to create unit tests. Therefore some sloppiness is allowed in unit tests that is not allowed in library code. > to help more automated/ASAN testing? It is the duty of the reporters to report as a bug only serious memory leaks. If you want to have automated testing against serious memory leaks, you need to distinguish not serious and serious memory leaks. For example, by passing --suppression options to valgrind, or by running test cases in a loop that repeats them a 1000 times. Bruno