ARM946E-S attempt (gdb patches / insight)

[Smith, Gene SEA]

How do you set the compile option to -O0 so it takes effect everywhere
(building rtems and your app)?  

-----Original Message-----
From: rtems-users-bounces at rtems.org
[mailto:rtems-users-bounces at rtems.org] On Behalf Of Joel Sherrill
Sent: Monday, March 24, 2008 4:24 PM
To: gds
Cc: rtems-users at rtems.com
Subject: Re: ARM946E-S attempt (gdb patches / insight)

gds wrote:
> In my attempt to build an ARM946E-S bsp I hijacked the existing csb337
> bsp as a starting point. I have built an arm-rtems4.8-insight debugger
> in addition to the standard gdb. However, I never applied the rtems
> gdb6.6 patches to the insight 6.6 tree. Are they needed with insight?
>
> Anyhow, when running insight debugger I seems that, in "mixed mode", I
> don't always see the assembly instructions associated with the correct
> source line. The order seems somewhat random and jumps around when I
> step. In the attached example you can see that the instructions
> associated with a c source line are not always sequentially
increasing.
> Would that be expected with arm architecture, which I have not worked
> with before? Would lack of rtems patches affect this?
>   
It isn't an RTEMS issue.  With optimized code on RISC CPUs,
the assembly from multiple source lines might be intermixed
for performance reasons.  I know the PowerPC and SPARC are
pretty aggressive about this and it can be VERY frustrating.

Some people compile at -O0 when things to make this phase
of debugging easier.  But remember, your code may break
again at -O2 especially in the case of missing volatiles.
> The code shown in ex.txt is from bspstart.c.
>
> -gene
>