Re: [avr-libc-dev] Possible hole in the assembler interrupt documentation

[Georg-Johann Lay]

Joerg Wunsch schrieb:
> As Dale wrote:
>
> > The issue then is the case of a library that contains an ISR.
>
> That's correct.  In this case, you have to cause an explicit undefined
> reference to it, in order to make the linker even consider pulling the
> library module.

Doesn't the linker look in the libraries if it finds a non-weak 
definition for a weak symbol?

__vector_8 is defined weak in .vectors (gcrt1.S, i,e, crt*.o) and 
defined there as __bad_interrupt.  Is that sufficient to stop the linker 
from dearching for overrides of __vector_8?

Johann

> In addition to placing another (referenced) symbol
> into the same module, this might be done using the -u linker option:
> suppose, my __vector_8 function where placed in a library, the
> respective (compiler) commandline switch would be "-Wl,-u,__vector_8".
> This makes the symbol "__vector_8" a "global undefined" straight from
> the beginning, so pulling the library module in satisfies that
> reference to the linker.