I386 strlen "fix" (giveup)

Richard Henderson rth@cygnus.com
Tue Mar 7 18:55:00 GMT 2000

On Tue, Mar 07, 2000 at 06:22:25PM -0800, Richard Henderson wrote:
> ... which is suggestive of stack alignment rather than strlen. 

Which is actually related to this strlen nonsense. 

Given something like strlen(foo(x)) if we abort the expansion we must
remember that we've expanded a call to foo and must undo not only the
rtl but whatever other state we've changed, e.g. pending_stack_adjust
and arg_space_so_far.

I hadn't actually committed the previous patch yet, and I'll wait for
the bootstrap to finish before doing anything else.


More information about the Gcc-patches mailing list