Resetting LMA for new VMAs

[Nathan Sidwell]

Alan Modra wrote:
> On Sat, Apr 05, 2008 at 07:11:43PM +0100, Nathan Sidwell wrote:

>> or rather the manual says the LMA spacing matches the VMA spacing, but does 
>> not mention this surprise if you set the VMA but not the LMA.
> 
> But your example script *does* set LMAs.  I don't see why you think
> that the new linker behaviour is any more surprising than the old.

I guess it depends on one's mental model of the linker internals.  Also, as 
there doesn't appear to be a tescase to check this behaviour, I wondered if it 
was an intentional design consequence.

> Getting back to overlays, suppose I have a script that specifies
> overlays and thus the section lma is greater than section vma for
> every section past the end of the overlay region.  If I happen to set
> vma for one of these sections, say to leave a gap, then your patch
> would set lma back to vma, potentially overlapping some previous
> section.  Note that this imaginary overlay script doesn't currently
> need to set lma anywhere.

Yes, I see that. thanks for the example.  I'll tell the customer to change their 
linker script.

nathan
-- 
Nathan Sidwell    ::   http://www.codesourcery.com   ::         CodeSourcery