Home Original page

Does the LD --wrap feature work for library internal references?

Sebastian Huber sebastian.huber@embedded-brains.de
Sat Jan 5 18:51:00 GMT 2019 
----- Am 4. Jan 2019 um 20:45 schrieb Sebastian Huber sebastian.huber@embedded-brains.de:

> Hello Nick,
> 
> ----- Am 4. Jan 2019 um 15:27 schrieb Nick Clifton nickc@redhat.com:
> 
>> Hi Sebastian,
>> 
>>> I don't use link-time optimization. I am sure the functions ether_input() and
>>> ether_output() are present in the executable.
>> 
>> OK, but my guess is that the compiler has already done something clever
>> so that those symbols are never referenced, and hence wrapping the symbols
>> has no effect.  (You could look in the object file and see if the symbols
>> are used in one or more relocations...)
>> 
>> Alternatively maybe there really is a bug in the linker's wrapping code.
>> But I have no idea what it might be, or why this particular situation is
>> causing problems.  If you do think that it is a linker bug though, please
>> could you put together a small test case to demonstrate the problem ?
>> Ideally using assembler source files rather than C/C++ sources, so that
>> we do not have to worry about compiler optimizations.
> 
> attached is a self-contained example with Makefile and C code. It seems the
> reference via function pointers is the problem.

I slightly modified the test program. With a main() function like this:

int main(void)
{
	func gg;
	gg = f();
	(*gg)();
	g();
	return 0;
}

I get this output:

__wrap_f
f
g
__wrap_g
g

The f() function returns a function pointer to g(). The indirect (*gg)() call is not wrapped, the direct g() call is wrapped. Is this a bug or a feature?
More information about the Binutils mailing list