Exit with Return

Hi

I am looking for a “termination” routine that executes finalization/cleanup code, registered via atexit() and global/static object deconstructors, except it doesn’t end the process. Instead returns to the caller (which in turn should fully exit the process) or performs an assignable callback (and then terminates).

So far I have only encountered two such functions: _cexit() for windows and _exithandle() for solaris (which I am not quite sure performs as expected). But I also seek such implementations for Linux, any ideas? “Macros Controlling Initialization Routines” in GCC might help if I could make sense of them.

Any help is appreciated.

> Instead returns to the caller
What does this mean? Returns to main?

If you want to get back to main from anywhere, then use setjmp() and longjmp().

But this won't execute anything registered with atexit(), and won't execute any global/static object deconstructors either. You need to specifically return from main or call exit() for these things to happen.

Quote:

---

> Instead returns to the caller
What does this mean? Returns to main?

---

No not main but the function the "terminating" routine was called from. Instead ending the program and never returing, such as the case with exit().

To further clarify:
I am looking for a function that performs cleanup as exit(), but doesn’t terminate the program, and instead returns or does some sort of callback.

> No not main but the function the "terminating" routine was called from.
You mean like every other function returns to the place it was called from.

Or do you mean to the place where the terminating routine was registered from?

Quote:

---

Originally Posted by Salem

>You mean like every other function returns to the place it was called from.

---

Yes, precisely.

A I read this, you're looking for a library function that you can call that will close all open file descriptors, call the chain of registered exit handlers, and do all the other cleanup that exit() does, then return to the caller so that some other exit processing might be done.

There is no standard entry point for this. It is compiler run-time library dependent, and may even be different between operating systems using the same compiler.

If you have access to glibc (or the C runtime support library for your compiler, whatever it is) sources, you might want to look at the source for the function _exit().

I've no idea what he wants. Calling the global destructors before main() exits makes no sense - what if you then use a global object on you way back to main?

Perhaps a description of the problem, rather than how to implement an imagined solution would help.