mutable members in global const objects

[Jamie Lokier]

I wrote:
> I produced a patch to deal with the bug (mutable global objects in read
> only data section), 2.5 years ago.  I don't why it never made it in.
> Maybe I never submitted it :-)  Probably something to do with GCC 2.8
> taking forever.
> 
> It also fixes a bug -- you're right about the optimiser needing to know
> about mutable.  Otherwise it thinks some values are fixed when they're
> not, and generates the wrong code.  At least, GCC 2.7.2.x did.

I just checked, EGCS 1.0.2 has both bugs.

-- Jamie