Hi Dominique, On 15.08.2017 15:41, Dominique Leuenberger / DimStar wrote:
On Tue, 2017-08-15 at 11:53 +0200, Richard Biener wrote:
SLES 15 will change the architecture level for 32bit x86 package builds to x86-64 to match the 64bit x86 builds as the only use for the 32bit packages are to run on a 64bit SLES install.
I propose to do the same for Tumbleweed.
So you basically send our 32bit users (i586) away - what do we actually WIN by this?
Only very few people are actually using i586, most are using later i686 processors.
Requiring x86-64 means (ontop of the current i586 level) requiring MMX, SSE, SSE2 and FXSR.
That cuts off (from a look at GCC internals) CPUs with nickname pentium-mmx, c3-2, i686, pentiumpro pentium2 pentium3, pentium3m, geode, k6, k6-2, k6-3, athlon, athlon-tbird, athlon-4, athlon-xp, athlon-mp. Most of those will have difficulties running Tumbleweed due to low memory (I remember my Athlon box having 16MB of ram which was plenty at the time).
A i686 with a couple GB of ram is nothing special I'd say.
These machines (pentium M dothan machines for example) will not be discontinued IIUC, only earlier machines (pentium3 with less than ~1GHz and probably a few Athlon class with less than ~1.5GHz CPU frequency).
Advantages of raising the architecture level is better performance of 32bit applications on 64bit and a simpler repository setup (no more glibc.i686).
Bonus points for somebody finding a table mapping architecture features to products rather than the other way around.
Thus - check your /proc/cpuinfo for the lack of MMX/SSE/SSE2/FXSR and report back.
Sadly, we will not get a feedback from all machines running TW - so we frankly have no idea what machine will 'stop working from one day to another'
Of course this is no issue for SLE/Leap, where there is simply no i586 support - I see it slightly different for TW and before accepting any such thing, we should get an understanding of what we break and how many users/machines we 'send away'
People who really need to use such old and (for todays standards) very low powered hardware are likely to not use openSUSE on them anyway but instead will probably be looking into custom embedded-class distributions / meta-distributions (Yocto poky anyone?) already. I'm all for keeping the old stuff working, but on the other hand I could not even be bothered to boot my old Pentium M dothan class machiens for over a year, even less try if current TW still runs, so I can understand why Richard from a "reduce complexity" standpoint wants to get rid of the old stuff and concentrate on the machines that are actually used by people. -- Stefan Seyfried "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." -- Richard Feynman -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org