Re: [PATCH] Built kernel without -O2 option

[Michael Snyder <msnyder@vmware.com>]

Américo Wang wrote:
> On Mon, Nov 29, 2010 at 11:12:16AM +0100, Mark Wielaard wrote:
>> On Mon, 2010-11-29 at 10:59 +0100, Andi Kleen wrote:
>>> Hui Zhu <teawater@gmail.com> writes:
>>>
>>>> Now, there are a lot of ways to debug the Linux kernel with GDB, like
>>>> qemu, kgtp or kgdb and so on.
>>>> But the developer more like add a printk. It have a lot of reason, a big one is:
>>>> (gdb) p ret
>>>> $3 = <value optimized out>
>>>> And the code execution order is not right.
>>> Really the right place to fix a lot of this would be gcc (and perhaps
>>> gdb). I suspect with some careful work the debugging experience
>>> for -O2 could be improved a lot.
>>> I also believe the latest gccs already have improvements in this area.
>> The VTA branch was merged for GCC 4.5
>> http://gcc.gnu.org/wiki/Var_Tracking_Assignments
>> Which helps a lot with generating dwarf for previously "optimized out"
>> values. There were also a couple of papers on generating even better
>> debuginfo at the recent GCC Summit: http://gcc.gnu.org/wiki/summit2010
>>
>> Jakub Jelínek, Improving debug info for optimized away parameters
>> http://gcc.gnu.org/wiki/summit2010?action=AttachFile&do=get&target=jelinek.pdf
>> Alexandre Oliva, Consistent Views at Recommended Breakpoints
>> http://gcc.gnu.org/wiki/summit2010?action=AttachFile&do=get&target=oliva.pdf
>>
> 
> Cool! It would be much more convenient if this is solved by gcc.

Those features mentioned at the GCC summit are not real close to 
completion.  I wouldn't put off the discussion just because they're
(possibly) in the pipeline.