Re: [avr-gcc-list] 'relocation truncated to fit' diagnosis ideas?

[Georg-Johann Lay]

Jim Brain schrieb:

> Received this error tonight after enabling a new chunk of code to 
> compile on a MEGA168 target
>
> I noticed some fixes in newer AVR GCC versions for things like this, so 
> I pulled GCC 4.7.0:
>
> http://www.avrfreaks.net/index.php?name=PNphpBB2&file=viewtopic&t=117504&postdays=0&postorder=asc 
>
> I got further, but after copying back in all the code I had pulled out, 
> it still complains about R_AVR_13_PCREL errors.
>
> -lm and relax have been on all the time.
>
> Is there a tutorial that describes how to restructure your code or your 
> Makefile to overcome this issue?  I understand it's a RJMP 13 bit 
> address overrun issue during linking, but I am lost on how to address 
> the issue.  Essentially the same code compiles fine for MEGA162, and the 
> current codebase compiles fine if I leave off a few calls in main.c.

It's not about -lm "on" or "off", it must be at the right position in
the linker command line to have the desired effect.

Besides that we don't see the error message and there is o clue if the
problem comes from library code from libgcc, from AVR-LibC, other lib,
home-brew assembler-code or gcc-generated assembler code.

The mentioned error might also be caused by AVR-LibC issue
http://savannah.nongnu.org/bugs/?33698

FYI, there are also some binutils issues, amongst them
http://sourceware.org/PR13899 so if you use binutils built from
trunk you may want to disable relaxing.

Notice that you don't need to use binutils shipped with a particular
distribution or snapshot; you can just as well use binutils from some
different package. But be aware that the older it is, the more features
might be missing; e.g. plugin support needed from LTO.

From what I can tell avr-binutils 2.21 works fine.

However, your issue is not a binutils issue except of the rare case of
http://sourceware.org/PR13410
Or it is a new problem never reported before.

Johann

> My code is GPL, available to anyone who is keen to see it fail themselves.
>
> Jim