i386 setjmp, asm volatile ("") and -fno-defer-pop

[Geoff Keating]

Etienne Lorrain <etienne@masroudeau.com> writes:

>   Hello,
> 
>  I have redefined my own setjmp()/longjmp() inline functions for
>  a special use (i386 realmode) but keeping same meaning.
>  The special thing about these functions is that they transfer
>  the control to somewhere else - and so any "defer-pop" optimisation
>  is not really welcome across those "asm volatile inline functions".

It doesn't work to transfer control inside an 'asm'.  There will be
other problems besides the one you're seeing.

-- 
- Geoffrey Keating <geoffk@geoffk.org>