| 1 |
On 29.03.2015 23:29, Davide Pesavento wrote: |
| 2 |
> On Sun, Mar 29, 2015 at 9:12 PM, Matthias Schwarzott <zzam@g.o> wrote: |
| 3 |
>> On 29.03.2015 20:58, Matthias Schwarzott wrote: |
| 4 |
>>> Hi there! |
| 5 |
>>> |
| 6 |
>>> I updated my ~amd64 system recently to new hardware (Intel Core i3-4160). |
| 7 |
>>> Since then valgrind did no longer work for 32bit programs because |
| 8 |
>>> "-march=native" did choose instructions that valgrind does not support |
| 9 |
>>> in 32bit mode (even ld.so was unusable). |
| 10 |
>>> |
| 11 |
>>> After some research I put this into make.conf and now it works: |
| 12 |
>>> CFLAGS_x86="${CFLAGS_x86} -mno-avx2 -mno-sse4 -mno-bmi -mno-bmi2" |
| 13 |
>>> CXXFLAGS_x86="${CXXFLAGS_x86} -mno-avx2 -mno-sse4 -mno-bmi -mno-bmi2" |
| 14 |
>>> |
| 15 |
>>> Is this the best solution to the problem? |
| 16 |
>>> If yes, the valgrind ebuild could suggest something like this. |
| 17 |
>>> Either always show it or check cpu-flags first (is this maintainable?). |
| 18 |
>>> |
| 19 |
>> I should add, that it seems to break for exactly one package: mariadb |
| 20 |
>> |
| 21 |
> |
| 22 |
> Not only mariadb, there are other known breakages... see |
| 23 |
> https://bugs.gentoo.org/show_bug.cgi?id=541616#c5 |
| 24 |
> |
| 25 |
> According to mgorny (Cc'ed): |
| 26 |
> "You are not supposed to touch CFLAGS_x86, ever. That's some magic |
| 27 |
> stuff that's used in profiles & multilib.eclass." |
| 28 |
> |
| 29 |
I created this bug to track the issue: |
| 30 |
https://bugs.gentoo.org/show_bug.cgi?id=545052 |
| 31 |
|
| 32 |
Regards |
| 33 |
Matthias |
Archives