diff options
Diffstat (limited to 'newlib/libc/stdlib/__call_atexit.c')
| -rw-r--r-- | newlib/libc/stdlib/__call_atexit.c | 36 |
1 files changed, 36 insertions, 0 deletions
diff --git a/newlib/libc/stdlib/__call_atexit.c b/newlib/libc/stdlib/__call_atexit.c index 6dd2ec63a..c99b2744e 100644 --- a/newlib/libc/stdlib/__call_atexit.c +++ b/newlib/libc/stdlib/__call_atexit.c @@ -15,6 +15,42 @@ void free(void *) _ATTRIBUTE((__weak__)); extern _LOCK_RECURSIVE_T __atexit_lock; #endif +/* If "__libc_fini" is defined, finalizers (either + "__libc_fini_array", or "_fini", as appropriate) will be run after + all user-specified atexit handlers. For example, you can define + "__libc_fini" to "_fini" in your linker script if you want the C + library, rather than startup code, to register finalizers. If you + do that, then your startup code need not contain references to + "atexit" or "exit". As a result, only applications that reference + "exit" explicitly will pull in finalization code. + + The choice of whether to register finalizers from libc or from + startup code is deferred to link-time, rather than being a + configure-time option, so that the same C library binary can be + used with multiple BSPs, some of which register finalizers from + startup code, while others defer to the C library. */ +extern char __libc_fini __attribute__((weak)); + +/* Register the application finalization function with atexit. These + finalizers should run last. Therefore, we want to call atexit as + soon as possible. */ +static void +register_fini(void) __attribute__((constructor (0))); + +static void +register_fini(void) +{ + if (&__libc_fini) { +#ifdef HAVE_INITFINI_ARRAY + extern void __libc_fini_array (void); + atexit (__libc_fini_array); +#else + extern void _fini (void); + atexit (_fini); +#endif + } +} + /* * Call registered exit handlers. If D is null then all handlers are called, * otherwise only the handlers from that DSO are called. |