The new assembler with unwind table bug fix

[Jim Wilson]

On Thu, 2004-02-12 at 23:05, David Mosberger wrote:
> Then I thought I'd do some more sanity checking by compiling the Linux
> kernel with -falign-loops=32.  Unfortunately, it looks like that's
> triggering a bug:

As Alan's message points out, md_estimate_size_before_frag is wrong.

This is a problem with HJ's patch to make .align work.  If we are going
to change the size of a frag when finalizing, then we need to tell the
relaxation code about this before hand, but we don't.  This isn't too
hard to do by copying the first part of ia64_convert_frag into a new
ia64_estimate_size_before_relax function.  Then there is a related
problem that slot_index needs to compute proper sizes before we
finalize, instead of during.  I fixed this by adding a new argument to a
bunch of functions so we could pass this info down from
estimate_size_before_relax.

The result works for your testcase.  I haven't tried it on anything else
yet.  I am not happy with the code duplication between convert_frag and
estimate_size_before_frag.  This needs to be looked at.
-- 
Jim Wilson, GNU Tools Support, http://www.SpecifixInc.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: patch.gas.unwind
Type: text/x-patch
Size: 8783 bytes
Desc: not available
URL: <https://sourceware.org/pipermail/binutils/attachments/20040217/a518f27b/attachment.bin>