Alias-ing problem

[Sergio]

Denys Vlasenko wrote:
> 
> So, if newer compiler works for you, what is the problem?

That's the situation:

I have gcc 3.3.6 on my PC, on the server there are gcc 3.3.1 and 3.4.3
For the final build I must use the server but the newer (3.4.3) compiles a 
BB that segfaults on the target board (perhaps I should spend my time on 
this problem) so I have to use the older 3.3.1 which works.

> Let's do it this way.
> 
> * Try 1.13.3, there is not much sense in debugging custom ash.c.

ash.c from 1.13.3 works with both gcc 3.3.1 and 3.3.6 (segfault with 3.4.3) 
but the custom ash.c is needed for another problem posted on this list in 
date 12/15/2008.

> * If using new, working gcc is somehow not acceptable for you,
>   do this: build ash from bbox 1.13.3 with gcc 3.3.1,
>   verify that it is broken, then build ash from bbox 1.13.3 with successively
>   newer gcc's. find a pair of closest gcc versions which made it (dis)appear.
>   Send ash.s from both of them.

Not so simple to set up such a test.

> If you will find out that *all* gcc 3.3.x (up to latest)
> produce broken ash, it would point strongly to a gcc bug.

As I wrote, compiler 3.3.6 on my PC creates a good BB so there is at least 
one "good" 3.3.x

Thanks
	Sergio

-- 
Free software is a matter of liberty, not price. To understand the concept, 
you should think of free as in free speech, not as in free beer
- Richard Stallman

http://www.gnu.org/philosophy/free-sw.html