[prev in list] [next in list] [prev in thread] [next in thread] 

List:       pcc-list
Subject:    Re: stack corruption with -xtemps
From:       Anders Magnusson <ragge () ludd ! ltu ! se>
Date:       2010-05-27 9:45:50
Message-ID: 4BFE3F4E.5090509 () ludd ! ltu ! se
[Download RAW message or body]

Hi,

Iain Hibbert wrote:
> On Thu, 27 May 2010, Anders Magnusson wrote:
>
>   
>> Interesting.  Hm, n_stalign must be set for structure copies, so starg()
>> should always be called.  It's set in p2tree() ny calling talign(), so
>> there should be an error if it would ever become 0.  This is the
>> interesting thing; how n_stalign can end up unset.
>>     
>
> probably cross pollination related to the __attribute__?
>
> 	union {
> 		int	_label;
> 		int	_stalign;
> 		int	_flags;
> 		struct	suedef *_sue;
> 	} n_6;
>   
the bug was because of alignment that was set to 1 bit, and when 
converted to
pass2 it was divided with SZCHAR which is 8.  Fixed by setting alignment
to at least 1 when converting trees for pass2.

Note:  the frontend uses bits for size and alignment, but the backend 
use bytes.

-- Ragge
[prev in list] [next in list] [prev in thread] [next in thread]