The future of i486 arch in Th

[Witold Filipczyk]

On Mon, Dec 15, 2014 at 09:01:58PM +0100, Jan Rękorajski wrote:
> TL;DR Future of i486 is bleak, it's a strong candidate for removal.
> 
> Proposal: remove i486 from Th.
> 
> Rationale:
> 
> It's increasingly more problematic to get everything built on i486, some
> packages explicitly say "no" to this arch, some require hacks and
> strange workarounds, and other just silently fail in mysterious ways.
> All this makes maintainability cost of i486 very high, and combined
> with the size of userbase, overly high.
> 
> Some usage stats from the Sun Nov 16 - Sun Dec 14 period, from
> ftp1.pld-linux.org:
> 
> # zgrep x86_64 /var/log/archive/xferlog*gz  | grep -v <builder-ip> | grep poldek | awk '{ print $7 }' | sort | uniq -c | sort -n | wc -l
> 533
> 
> # zgrep i686 /var/log/archive/xferlog*gz  | grep -v <builder-ip> | grep poldek | awk '{ print $7 }' | sort | uniq -c | sort -n | wc -l
> 730
> 
> # zgrep i486 /var/log/archive/xferlog*gz  | grep -v <builder-ip> | grep poldek | awk '{ print $7 }' | sort | uniq -c | sort -n | wc -l
> 20
> 
> # zgrep i486 /var/log/archive/xferlog*gz  | grep -v <builder-ip> | grep poldek | awk '{ print $7 }' | sort | uniq -c | sort -n
> [exact IPs removed to protect privacy]
>       < 10		4 IPs
>       < 50		4 IPs
>       < 100		4 IPs
>       < 1000		7 IPs
>       > 1000		1 IP
> 
> As you can see there are 20 distinct users of i486, out of which
> maybe 5 really use this arch.
> 
> Due to higher workload i486 arch requires compared to anything else and
> symbolic userbase, I propose to drop it from Th.
> 
> Comments?
> 
> PS. we can use former i486 builder for x32 port if there will be takers.

Kernel build for non-PAE i686 is what I want.