[lfs-dev] Two changes to consider for 8.0
dj at linuxfromscratch.org
Sat Dec 31 12:05:37 PST 2016
On 12/31/2016 08:37 AM, John Frankish wrote:
>>> I've followed LFS to build a toolchain several times, except that I
>>> built gold and used isl as a dep.
>>> With all of the toolchains, wherever other packages automatically use
>>> gold (libreoffice and others) compilation fails for different
>>> non-obvious (to me at least) reasons until gold is disabled.
>> I'll be able to check it out soon. How are you using gold? -B?
> As mentioned, some apps (webkitgtk, not libreoffice as I originally stated) look for gold themselves and will only compile if it is disabled) but I've also used "-fuse-ld=gold" or made a symlink ld -> ld.gold
Hard to say, but most errors encountered with gold are due to implicit
linking. There is a lot of history here, and I'm boiling it down by
saying this, but bfd effectively replicates a lot of the functionality
of the dynamic linker, and this is one of the reasons cited for gold
being faster than bfd (yet to be shown in 1:1 comparison by me) and
being used to identify link time errors. Take a look at this old bug for
a much more complete explanation:
> I just tested with something simple - sed - and in fact "-fuse-ld=gold" works, albeit the resulting sed binary is smaller if gold is not used.
Thanks for doing this. The readelf utility will explain the difference
better than I can. :-)
More information about the lfs-dev