Another relocation problem - and patch

[Alan Modra]

On Mon, 5 Mar 2001, Mikulas Patocka wrote:

> It doesn't use optimal length but at least it doesn't create broken code.

This patch is more or less OK, but I'm currently testing another patch
to fix the previous problem you reported that happens to handle this case
as well.

> The relax is actually done in md_estimate_size_before_relax because the
> section of expression symbol is different. Has anybody idea what to do
> with it? Is there some function that evaluates the expression symbol and
> converts it to symbol + constant value?

In general, you can't evaluate expressions at this point.
expr.c:clean_up_expression is probably what you're looking for.

Alan Modra
-- 
Linuxcare.  Support for the Revolution.