Re: Strange KEEP behavior in linker script

[Denis FERRANTI <denis dot ferranti at st dot com>]

Nick Clifton wrote:
> Hi Denis,
>
> > just because of the symbols attached to it in the linker 
> > script...that are now attached to another section.
>
> Is the section containing the symbols relevant in this case ?
>
> Presumably what is important is the difference in the values of the 
> two symbols.  ie the size of the contents of the .mysec section.  When 
> this section is empty, the two symbols are attached to another 
> section, but the difference between them still accurately reflects the 
> size of the (non-existent) .mysec section.
i agree with you. This is the mainly the case in our implementation.
> Essentially I see the situation like this:
>
>   The bug with the 2.16 linker's behaviour is that it would create 
> empty sections in the output.
>   The bug with the 2.17 linker's behaviour is that it would convert 
> section-relative symbols to absolute symbols.
>   The problem with the 2.18 linker's behaviour is that the symbols in 
> the empty section are being (non-intuitively) attached to another 
> section.
>
> So, where do you think that the symbols should go ?
Reading at old mails it seems indeed that there have been some issues 
with alignment of empty output sections not discarded. I can see the issue.
If the creation of an empty section is a problem, then the best thing to 
do with the symbols is indeed to attach them somewhere.

Thanks for you answers,


> Cheers
>   Nick

--
Denis Ferranti