On Sat, Jul 22, 2017 at 05:39:44PM -0400, Renaud OLGIATI wrote:
> On Sat, 22 Jul 2017 23:26:56 +0200
> Adam Borowski <kilobyte@???> wrote:
>
> > Then this looks like a problem that needs to be looked at. There's no way
> > that many people use gear from ≤ 2004 (or a brief throwback of early Atoms
> > from 2008).
>
> Dont we have stats on how many download the 386 version, against how many for 64 ?

Not sure if mirrors provide download stats; popcon is probably good enough.

What I'm talking about is running i386 on 64-bit-capable CPUs. You can
check that by ｢grep '^flags.*\bnx\b' /proc/cpuinfo｣ or checking the op-mode
field in what lscpu says.

There's a long list of reasons why that's a bad idea, especially when kernel
is concerned; the only reason to the contrary is some memory saving in
pointer-heavy code. 32-bit code also sees almost no upstream testing
(at least on x86).

If the machine has >2GB ram, running a 32-bit kernel should be a crime.

--
⢀⣴⠾⠻⢶⣦⠀
⣾⠁⢠⠒⠀⣿⡁ A dumb species has no way to open a tuna can.
⢿⡄⠘⠷⠚⠋⠀ A smart species invents a can opener.
⠈⠳⣄⠀⠀⠀⠀ A master species delegates.