Re: [RFC]: New dwarf2 reader

[Daniel Berlin <dan at cgsoftware dot com>]

Andrew Cagney <ac131313@cygnus.com> writes:

> Dan,
> 
> 
> For such a radical change to fly, I think this should be a runtime
> replacement for the existing dwarf2 reader.
I thought of this, and it could be done, I just don't know what to
name the file.
:)

> 
> 
> That way people can test the two side by side.  It will give people
> greater confidence and an easy reference point.
> 
> 
> Some trivia:
> 
>> #ifndef DWARF2_REG_TO_REGNUM
>> #define DWARF2_REG_TO_REGNUM(REG) (REG)
>> #endif
> 
> 
> 
> this shouldn't be needed.  If it is then there is a bug in
> multi-arch.

I can't remember who added this. Some register mapping differences or
something. 
> 
>> #if defined(HAVE_MMAP) && defined(HAVE_GETPAGESIZE)
>> #include <sys/mman.h>
>> #endif
> 
> 
> It should be posible to structure things so those #if #else #endif's
> are largely eliminated.  You want as much of your code to be exposed
> to the C (not C++ :-) compiler at all times.  That mysterious #if
> D2FIX is similar. 
It's something on my list to fix.

> Does bfd already provide an MMAP facility?
No, that's the whole problem.
We've waited many years for this to happen, and it hasn't.
So I gave up waiting.
:)


> 
> 
> You'll eventually want to go over the code with a fine tooth comb -

Yup.
> 
> flush out the naughtness in the old dwarf2reader and eliminate the
> neuances you, your self have introduced.
Yup.

As I said to michael, I won't even consider submitting it for approval
until all known bugs i know of are fixed, and we do better or the same
on the regression tests.
(I've not gotten a chance to run them yet, so i don't know where we
stand on that.)
> 
> 
> 	Andrew
> 
> 

-- 
"Last year I went fishing with Salvador Dali.  He was using a
dotted line.  He caught every other fish.
"-Steven Wright