SH relocation differences in older versions of the binutils

[Jeff Baker]

> Given that the old relocs only happen with -relax, and gcc doesn't use
> the new ones at all, have you even encountered such a collision yet?

I haven't but I can't guarantee our customers that they never will until I
know more about this.  I'm trying to prepare for the inevitable flood of
questions.
 
> Otherwise, I don't know of any way to test these.  The old numbering
> was wrong; if anything the old ones would need the special comment.
> The new numbering reflects compatibility with Hitachi's tools, which
> we cannot control.  So, even if we added a flag, you would not be able
> to rely on it.

As far as this goes I'm thinking about the more distant future.  Personally
I can't see any harm in having this information in the binary but I can see
it having potential benefits.