On Fri, May 29, 2015 at 02:57:16PM +0000, Emmanuel Dreyfus wrote: > I have seen that with older libc: setting breackpoints on > pthread_key_create() and pthread_setspecific() shows that malloc() > calls the later without calling the former
I don't think this is true, just gdb misses the first call or something. There has been a discussion about this recently on some other list and we concluded that the key_create() actually does happen as first thing. Maybe gdb is confused, as _rtld_debug() happens later? Martin
