Re: [RFA] Improve Sparc epilogue analysis

[Michael Snyder <msnyder at redhat dot com>]

"David S. Miller" wrote:
> 
>    From: Michael Snyder <msnyder@redhat.com>
>    Date: Wed, 24 Apr 2002 17:36:02 -0700
> 
>    "David S. Miller" wrote:
>    > I'd like to ask that you re-review this change, given this, please.
> 
>    The part that's mentioned in the changelog entry is OK, except
>    you can remove sparc_skip_prologue_frameless_p (and say so in
>    the changelog) rather than rewrite it.
> 
> Wait a second, how can I remove sparc_skip_prologue_frameless_p
> when it isn't even there anymore? :-)
> 
> What exists is "sparc_prologue_frameless_p" which is a boolean.
> That is what replaces the functionality for what I removed
> (sparc_skip_prologue_frameless_p, SKIP_PROLOGUE_FRAMELESS_P,
> et al.).
> 
> I think you're still a little confused on the current state of
> the tree.

That comment was unnecessary, thank you.

Whether the function is in or out, it is wrong as written, 
or at the least you have to justify why you have changed
the behavior.

The existing function returns TRUE if and only if the input
address is equal to the address returned by examine_prologue, 
which in this case will be the address of one of a small set
of specific instructions (eg. a SAVE or an add sp).

Your function, in the presence of symbols, will return TRUE
if and only if the input address is equal to the last instruction
in the line (presumably of the prologue).  This is a change.
I don't know whether it was intentional or not, but you can't
make it without explaining it.