Patch: update -fmessage-length docs

[Tom Tromey]

>>>>> "Gaby" == Gabriel Dos Reis <gdr@cs.tamu.edu> writes:

> Tom Tromey <tromey@redhat.com> writes:
> | It turns out that -fmessage-length defaults to 0 even for g++. 

Gaby> Then, that is a bug. 
Gaby> When I oiginally implemented that for C++, the intended semantics
Gaby> definitely was to have it default to 72 -- but C people wanted
Gaby> something different for C front end.  If it is observed that g++ has
Gaby> it set to zero, then we must set it back to 72.

I don't have a preference on this, just that the docs match the
implementation.

If we change it to 72 then PR 712 should be reopened.  That is how I
noticed this :)

Tom