Re: lua_resume and a C-function as the main coroutine function problem

[David Given <dg@...>]

On 2010-04-14 14:34, Tony Finch wrote:
[...]
> 		vois *osp =&f;
> 		void *nsp = malloc(STACKSIZE) + STACKSIZE - SLOP;
> 		void *dummy = alloca(osp - nsp);

I would be very, very careful doing things like this --- you're utterly 
at mercy of the implementation. This will, for example, trigger any 
stack limit checking that your toolchain might be doing. And woe betide 
you if your alloca() implementations tries to memset() the data!

The Official(TM) way of changing stacks is to use 
setcontext()/getcontext(). However, that's still pretty fraught. I had 
one project that used C-based coroutines using setcontext()/getcontext() 
that produced obscure errors on certain libc versions on certain Linux 
kernels if you happened to link against a library that had been compiled 
with pthreads support, because pthreads was storing magic data at the 
top of the stack, and didn't work properly if you changed stacks...

--
┌─── ｄｇ＠ｃｏｗｌａｒｋ．ｃｏｍ ───── http://www.cowlark.com ─────
│
│ life←{ ↑1 ⍵∨.^3 4=+/,¯1 0 1∘.⊖¯1 0 1∘.⌽⊂⍵ }
│ --- Conway's Game Of Life, in one line of APL