Working ELF GCC

Nick Burrett nick at dsvr.net
Mon Oct 6 02:26:55 PDT 2003


Peter Naulls wrote:
> In message <3F812F8D.7080100 at dsvr.net>
>           Nick Burrett <nick at dsvr.net> wrote:
> 
> 
>>Peter Naulls wrote:
>>
>>>See: http://www.drobe.co.uk/riscos/artifact834.html
>>>And: http://www.riscos.info/gcc/32bit.html
>>>
>>>Don't forget to read the notes.
>>>
>>>Although there are obvious short term problems, this proves the
>>>concept, and I suggest we move all GCCSDK development to ELF based ASAP
>>>- especially since we aren't currently looking at any kind of release
>>>for GCC 3.3.
>>
>>I disagree.  The thing holding up GCC 3.3 is a bug in the symbol/area 
>>creation in 'as'.  Actually fixing this is quite difficult without an 
>>AOF dumper.
> 
> 
> Indeed - but going to ELF would side step these problems.   As for an
> AOF dumper - do you have access to DecAOF in Acorn C/C++? 

No.  I haven't had a copy of Acorn C since version 3.

> IF there's a
> minimal case, then someone else could have a look at it.

The problem was detailed in 'Common AREAs and libfile'.

The minimal test case is:

---- snip ----
	AREA |__common|, DATA, COMMON
	EXPORT	|__common|

	% 12
	END
----- snip ----

'as' doesn't export the __common symbol properly


> Additionally, on 30th July, you referred to a code generation bug with
> alloca - what is the status of this?

The code gen bug is fixed in my local copy of the tree.  I haven't fully 
tested it due to the 'as' bug.

> The reason I suggest we are a way from a release is that we have yet to
> produce a working GCC 3.3 in RISC OS :-|
> 


-- 
Nick Burrett
Network Engineer, Designer Servers Ltd.   http://www.dsvr.co.uk





More information about the gcc mailing list