Real purpose for failure in "Too restrictive sanity check"

[Jan Hubicka]

> It won't even apply to the current sources.
OK, I will try to re-check and update the patch once again.
> 
> Worse yet, it looks like this change is also now causing code generation
> problems on PA64.  The constant breakage in the calls.c code is starting to
> get annoying.  Unfortunately, I don't have the time to track it down, and
> with the constant breakage I can't get the code into any kind of shape that
> other people can work on it.
> 
> *please* try to be more careful with this code.  It is very complex and easy
> to break.  IMHO, if you're tweaking this code, you need to be bootstrapping
> and testing on a variety of platforms, not just ia32.
I am bootstrapping at least two platforms for each calls.c change.  I am sorry
for causing such breakage, but in fact all my patches so far are bugfixes,
so the calls.c is already broken.  I am trying my best to understand that code
and get it working right in all cases. 

I would happily bootstrap on PA (since this one seems to be causing most problems)
if I had one available.

Honza
> 
> jeffk