Re: Tail recursion in Java, was Proper Tail Recursive C++

> > Surely the compiler has to apply the same policy statically as would
> > be applied dynamically.
>
> How does a Java VM know that the code it is running is produced by a
> compiler?


Even more to the point, the same class (as compiled) could be secure
in one runtime, and insecure in another - it is a dynamic property of
class loading, not something that can be determined statically.


The solution is that at load time we can make a determination of which
method calls are potentially between trusted and untrusted classes,
and label them as not tail-eliminatable. You also have to do an
analysis for calls from synchronized methods, which are again not, in
general, tail-eliminatable.




__Mark
[ markt@harlequin.co.uk | http://www.harlequin.co.uk/ | +44(0)1954 785433 ]
--

Return to the comp.compilers page.
Search the comp.compilers archives again.