M68K (mcpu=cpu32)issues with latest tool chain
chrisj at rtems.org
Fri Apr 1 21:38:48 UTC 2011
On 2/04/11 8:15 AM, Kirspel, Kevin wrote:
> The statement below if true will try and do 32-bit compares/copies (fast). If false, it will compare/copy byte by byte (slow).
> if (!UNALIGNED (s1, s2))
> The first instance of my crash is on a strcmp() call in find_handler() (mount-mgr.c). I have not tracked down why its failing. I will probably have to take a look at the disassembly to do so. I think you may be right about GCC. In my system, I don't need my m68k processor to run on RTEMS version 4.10. I'll probably revert back to 4.9 for that processor.
I run 4.10 on a Coldfire all the time and it is fine. It is not a CPU32
but they are not that different. I would like to see what the issue if
you do not mind helping dig in.
The compiler could a problem but this could be related to other code.
One area that may have changed is stack usage. Do you have plenty of
stack space allocate ? Would you be able to increase them for testing ?
More information about the users