Re: [rfa:symtab] deprecate inside_entry_func

[Daniel Jacobowitz <drow@mvista.com>]

On Fri, Nov 21, 2003 at 07:37:31PM -0500, Andrew Cagney wrote:
> 
> >>That's not what I'm asking.
> >>
> >>With the call to inside_entry_func removed, from get_prev_frame, can you 
> >>point me at any remainng _legitimate) uses of that function?
> >
> >
> >The other call to it in legacy_frame_chain_valid, which wants to know
> >the same thing?  I imagine the third caller, in frv-tdep.c, is bogus
> >and could be removed somehow.  But if it's going to be left there then
> >it seems reasonable to update it also.
> 
> I see Kevin's finally removed that bogus call.

Mistakenly, as you may have seen from his followup.

> >Conceptually the patch you just posted sees to be:
> > - Change the implementation of inside_entry_func
> > - Inline the new inside_entry_func into the one caller you're fond of
> > - Add a deprecated copy of the old implementation for the other
> >   callers
> >
> >>From the man who is always telling us how unimportant performance is
> >compared to clarity, I don't see the point.  Also, this leaves an old
> >implementation and a new implementation around for no visible reason.
> 
> "Indirection can be helpful, but needless redirection is irritating."
> 
> Note that the implementation of the two mechanisms is very different. 
> Better to leave old code using the old mechanism, and new code using the 
> new mechanism.

I believe the point of this discussion is that Kevin and I disagree
with your conclusion.

It's a mechanism, it doesn't matter if the implementation is different;
and one of them works and the other doesn't, so why keep two
implementations of the same thing if only one works?  In any case,
Kevin's patch to resolve this issue looks reasonable to me.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer