avr-gcc doesn't know address space wraps?

[Georg-Johann Lay]

Ralph Doncaster schrieb:
> On Sun, Jan 3, 2016 at 1:22 PM, Georg-Johann Lay <avr@gjlay.de> wrote:
>> Ralph Doncaster schrieb:
>>> avr-gcc 4.9.2 doesn't seem to know that the address space wraps, so
>>> that an rjmp in the last 2KB of the address space can reach code in
>>> the first 2KB.  The following code works fine with a jmp, but if I
>>> change the jmp ResetVector to rjmp, I get:
>>> (.bootloader+0x4): relocation truncated to fit: R_AVR_13_PCREL against
>>> `no symbol'
>>>
>>> I'm compiling with:
>>> avr-gcc -mmcu=atmega328p -nostartfiles
>>> -Wl,-section-start=.bootloader=0x7E00   picobootSTK500.S   -o
>>> picobootSTK500
>>
>> I don't quite get what your question has to do with the compiler proper.
>> As you are programming in assembly, the only thing that avr-gcc does is
>> calling the assembler and the linker for you.
> 
> I don't quite get your point.  Are you suggesting I should report this
> on the binutils list instead of here?

Well, you are using RJMP with an address that's out of scope, hence the 
linker/locator complains.  The compiler would not generate code like

>>>     rjmp ResetVector

The compiler would just emit "JMP <address>", hence you can use the same 
code sequence which always works.

Or are you having trouble with the compiler driver?  This is actually 
the only part of the compiler that's involved here...

If you prefer RJMP over JMP provided address is in the range of RJMP, 
you would usually link with --relax or call the driver with -mrelax.

If it's vital for your code to RJMP instead of JMP provided address is 
not in the range of RJMP but address+0x8000 happens to be, you can make 
explicit use of wrap around capability of RJMP by RJMP <address>+0x8000.

Third approach is to link with --pmem-wrap-around= so that no explicit 
offset and no explicit RJMP is needed in the code.