Re: Recent change causing regression for bfin-elf

[Alan Modra <amodra at gmail dot com>]

On Mon, Oct 28, 2019 at 12:33:03PM -0600, Jeff Law wrote:
> My tester spit this out a couple days ago:
> 
> Tests that now fail, but worked before (1 tests):
> 
> Build pr22471

Yes, saw that.  score-elf too.

> A quick bisection lands on:
> 
> > 30fe183248b2523ecff9da36853e2f893c4c4b91 is the first bad commit
> > commit 30fe183248b2523ecff9da36853e2f893c4c4b91
> > Author: Alan Modra <amodra@gmail.com>
> > Date:   Wed Oct 23 17:40:51 2019 +1030
> > 
> >     PR4499, assign file positions assumes segment offsets increasing
> 
> I have no idea how serious this is :-)

The error is
./ld-new: tmpdir/pr22471: error: PHDR segment not covered by LOAD segment
which was the case before this patch but ld didn't diagnose the error.
So not exactly a regression.

In more detail, we have a bunch of targets that have this nastiness in
their setup: eg. from ld/emulparams/elf32bfin.sh
GENERATE_SHLIB_SCRIPT=yes
EMBEDDED=yes

EMBEDDED means no image space wasted on the ELF file header or ELF
program headers.  Which is sad if you also want dynamic objects, as
implied by GENERATE_SHLIB_SCRIPT, since the headers are usually needed
to load dynamic objects.  I had a patch to tidy all this, but haven't
committed yet since I'm fairly sure there are loaders that *don't*
need headers in memory.

-- 
Alan Modra
Australia Development Lab, IBM