Robert Millan 写道:
Hi,
Thanks for looking into this.
On Thu, Apr 16, 2009 at 10:08:56PM +0800, yanhua wrote:
This error is caused by our patch to gcc. It generates a instruction
which causes overflow exception. Instead, we should use a exception free
instruction.
So this usually works in userland, because Linux catches the exception,
but not in kernel mode?
Normally yes, but there are still some risks. This instruction must
act as nop which should raise any exception.
What you describe sounds like something that can be fixed with some
flag (e.g. like -msoft-float). Are you sure this isn't possible?
Thanks very much for your report. We are now correcting the GCC, and
will release it. Then you need to recompile your kernel and modules, and
all programs you use the wrong gcc compiled.
We're using stock gcc 4.3.2-1.1. Could you point me to the commit that
fixed this issue? I'd rather backport a patch than using a development
branch of GCC, if possible.
This error is caused by our own patch to gcc. If you use gcc other than
lemote source, this issue should not occur.
We are now repatch the binutils and gcc. Then you can use lemote
released version again. Sorry for this.
--
晏华
_______________________________________________
gNewSense-dev mailing list
[email protected]
http://lists.nongnu.org/mailman/listinfo/gnewsense-dev
