| 1 |
2016-01-02 14:25 GMT-06:00 <waltdnes@××××××××.org>: |
| 2 |
> On Sat, Jan 02, 2016 at 12:55:56PM -0600, Jc García wrote |
| 3 |
> |
| 4 |
>> Then why the recently introuced multilib method of bulding 32bit |
| 5 |
>> libraries for packages that need it on 64 bit works? I don't think the |
| 6 |
>> devs would have bothered to introudce the variable ABI_X86 and a |
| 7 |
>> mulitib eclass just to compile a 32bit Hello World. |
| 8 |
>> |
| 9 |
>> I'm not trying to make a flame here, but don't blame the compiler, |
| 10 |
>> when in this case is more likely you the user are doing something |
| 11 |
>> wrong. |
| 12 |
>> My guess is you are blaming the effects of CPU_FLAGS_X86 on CFLAGS. |
| 13 |
> |
| 14 |
> The fact that I use "no-multilib" profiles on my 64-bit machines |
| 15 |
> probably doesn't help. The example I was using involved a manual build |
| 16 |
> of Pale Moon from source. I manually specified in the build script... |
| 17 |
> |
| 18 |
> ac_add_options --enable-optimize="-O2 -march=bonnell -mfpmath=sse -pipe \ |
| 19 |
> -fomit-frame-pointer -fno-unwind-tables -fno-asynchronous-unwind-tables \ |
| 20 |
> -mmmx -msse -msse2 -mssse3 -mmovbe -mfxsr" |
| 21 |
> |
| 22 |
> ...which are the options I use on my netbook client for outsourcing the |
| 23 |
> build process. So the host's CPU_FLAGS_X86 setting should not be a |
| 24 |
> problem. |
| 25 |
> |
| 26 |
A bit unrelated to the thread, but not really as a solution, here's my |
| 27 |
experience: |
| 28 |
I use Gentoo on a netbook too, but I'm not using distcc, it won't help |
| 29 |
much anyway, my approach is to make a 32 bit container(using |
| 30 |
systemd-nspawn for the sake of booting it fot testing, but chroot can |
| 31 |
get it done) and make sure to have the same world and /etc/portage/ in |
| 32 |
both, with the only difference that the container has |
| 33 |
FEATURES="buildpkg", and the netbook has FEATURES="getbinpkg", I use |
| 34 |
git to keep the changes in sync, and source make.local.conf (ignored |
| 35 |
by git) on each the container). I serve the binpkg host from my |
| 36 |
desktop to my LAN with nginx but I'm considering git from the booted |
| 37 |
container. I also mount $PORTDIR via NFS to have the same tree( |
| 38 |
bandwidth is expensive for me, and I also don't want to have tons of |
| 39 |
portage tree's around as I use the same method for another amd64 pc |
| 40 |
with an older processor, that has only 2 threads) |
| 41 |
|
| 42 |
PD: The next step is to make the upgrades be handled by ansible but I |
| 43 |
haven't sat down to make it happen. |
Archives