> Brad Proctor wrote:
> > "warning: using `%eax' instead of '%ax' due to `l' suffix."
> > by changing all the 'l' suffixes to 'w' and adding 'w' where it is
> > missing.
> I think the reason for using these "incorrect" constructs was that some
> versions of binutils fail if using the "correct" ones. I also seem to
> recall having seen some mention of general kernel build problems with
> certain recent versions of binutils, so the usual "oh, just upgrade to
> the latest version from CVS" may not help.
The plan (and part of the sinister reason that the new boot code was put
into place) is to gradually phase out the older binutils versions and
migrate to 2.9.5. The older versions just have too many problems, which
although might seem small (and most, truthfully, are) but some are truly
enormous in very subtle ways :)
As the first step, the boot code as is currently stands compiles with all
versions of binutils, 22.214.171.124.7 and up (lower might work, untested). It
does happily whine when compiled with newer binutils, though, as we've
The next step, once the documentation has been updated and people have
been given a good warning, is to fix up the few warts remaining and have
some 2.9.5 version be the absolute required minimum (at the moment its
only a *very* strong recommendation). Might not even be until 2.5, who
knows for sure.
> Isn't it ironic that the whole exercise started as an attempt to
> improve the esthetic value of the boot code ? (-:C
> - Werner
That was a nice side effect (appearance), but not the sole purpose.
Consistency in kernel usage of asm, leverage in making users update
binutils (because it *needs* to be done eventually), reduction in tools
needed to build kernel, my name in lights, to name a few. :)
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to firstname.lastname@example.org
Please read the FAQ at http://www.tux.org/lkml/