Breakage with bleeding edge host toolchains (Fedora)
dbn.lists at gmail.com
Wed Feb 28 20:30:12 PST 2007
On 2/28/07, Matthew Burgess <matthew at linuxfromscratch.org> wrote:
> On Wednesday 28 February 2007 18:16, Dan Nicholson wrote:
> > In gcc-2.95.3, if you pass -B and it's not
> > needed for anything, it will issue a warning on stderr
> In which case, I'd say just bump the GCC version requirement.
Attached is a diff which adds -B/usr/bin/ for the configure statements
for binutils-pass1 and gcc-pass1. For binutils, the -B/usr/bin/ isn't
strictly needed until after `make install' since an alternate ld
doesn't exist yet. However, there's no harm in setting it immediately,
and then the preferred way to call gcc is recorded in the Makefile.
I built gcc-pass1 a few times with different options. This seems to be
doing the right thing (Greg said the same). /tools/bin/ld is used for
the feature tests, but /usr/bin/ld is set as LD in the Makefile. This
LD is used to build xgcc and the stage1 gcc of the bootstrap. After
that, a wrapper script executing /tools/i686-pc-linux-gnu is called
for stage2 and stage3.
One difference between LFS and DIY is that DIY adds
--with-ld=/tools/bin/ld. AFAICT, this only affects the default linker
recorded in gcc and collect2. It also means /tools/bin/ld is called
from the wrapper script mentioned above. This could be nice, but
doesn't appear to affect this situation.
So, I think the attached patch should work on all hosts and have no
effect on the purity of the bootstrap. But, someone on a Fedora host
needs to check this out. Anyone got an FC6 partition?
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 3144 bytes
Desc: not available
More information about the lfs-dev