printk and side effects

[Myles Watson]

This is a reminder to myself in case it is obvious to everyone else:

It seems very difficult to guarantee that the removal/insertion of a print
statement has no other effects.

I printed out the location of a variable, and that changed where it was
stored (from a register to being placed on the stack)...

This might not be a traditional "side effect", but it changed the behavior
of the code and surprised me.

Thanks,
Myles

[Stefan Reinauer]

On 17.04.2010, at 16:41, Myles Watson  wrote:
> This is a reminder to myself in case it is obvious to everyone else:
>
> It seems very difficult to guarantee that the removal/insertion of a  
> print statement has no other effects.

Thanks for sharing this kind of information. This is definitely not  
obvious and people need to be aware of it.

I have seen similar side effects with printk too which lead to our  
current printk which evaluates the varargs, even if the maximum log  
level is below the printk's ...
> I printed out the location of a variable, and that changed where it  
> was stored (from a register to being placed on the stack)...
>
> This might not be a traditional "side effect", but it changed the  
> behavior of the code and surprised me.

In which way was the behavior changed? Anything we can watch out for?

Stefan

[Myles Watson]

> > This is a reminder to myself in case it is obvious to everyone else:
> >
> > It seems very difficult to guarantee that the removal/insertion of a
> > print statement has no other effects.
>
> Thanks for sharing this kind of information. This is definitely not
> obvious and people need to be aware of it.
>
> I have seen similar side effects with printk too which lead to our
> current printk which evaluates the varargs, even if the maximum log
> level is below the printk's ...

I appreciate that feature more now :)
> > I printed out the location of a variable, and that changed where it
> > was stored (from a register to being placed on the stack)...
> >
> > This might not be a traditional "side effect", but it changed the
> > behavior of the code and surprised me.
>
> In which way was the behavior changed? Anything we can watch out for?

I'm trying to track down a bug that's seems to be sensitive to stack
alignment, but I am having a hard time isolating it.  I think it's specific
to my setup.

Thanks,
Myles

[Myles Watson]

On Sat, Apr 17, 2010 at 2:41 PM, Myles Watson  wrote:
> > > This is a reminder to myself in case it is obvious to everyone else:
> > >
> > > It seems very difficult to guarantee that the removal/insertion of a
> > > print statement has no other effects.
> > > I printed out the location of a variable, and that changed where it
> > > was stored (from a register to being placed on the stack)...
> > >
> > > This might not be a traditional "side effect", but it changed the
> > > behavior of the code and surprised me.
> >
> > In which way was the behavior changed? Anything we can watch out for?
>
> I'm trying to track down a bug that's seems to be sensitive to stack
> alignment, but I am having a hard time isolating it.  I think it's specific
> to my setup.

Luckily it was specific to my hardware setup, not a bug in Coreboot.

Thanks,
Myles