[Prev][Next][Index][Thread]
Re: linux and mklinuximage take 2
Vaughan Knight wrote:
> >From reading the configure scripts, I have found that my system does
> support '-oformat binary' but the problem appears to arise over the fact
> that in the configure script, 'flux_cv_HAVE_CODE16=no'. To create the
> mklinuximage script flux_cv_HAVE_CODE16 must equal yes.
What is the output of the configure script? If it cannot figure out how to
create 32 bit offsets by specifying an address size override, it won't
even try the .code16 assembler directive, because the code wouldn't work.
But a dump of the relevant parts of the configure script output (the lines
around the one mentioning ".code16" in some way) can shed some light on.
But please don't send the full dump to the mailing list - it might annoy
some readers.
Maybe the check for the addr32 prefix is wrong for some version of binutils.
I found the binutils-2.9.5 much better than any previous version wrt.
16 bit code support, but as I haven't done anything with the oskit lately,
there might be a small bug.
Oh - as a side note: if you upgraded your binutils and just rerun the
configure script afterwards, you'll end up with the previously cached
results. Delete the config.cache file in your oskit build tree.
Hope this helps
Klaus
Follow-Ups:
References: