Bug#878817: closed by Gianfranco Costamagna <locutusofb...@debian.org> (Re: Bug#878817: luajit: Please drop mips64el again)

2017-10-23 Thread James Cowgill 
Control: forwarded -1 https://github.com/LuaJIT/LuaJIT/pull/362
Control: tags -1 patch

Hi,

On Fri, 20 Oct 2017 16:23:18 +0300 Adrian Bunk  wrote:
> Control: reopen -1
> 
> 16:20 < jcowgill> bunk, LocutusOfBorg: i have a patch for luajit which fixes 
> neovim - i haven't tested it on any other lua code yet 
>   though
> 16:21 < bunk> jcowgill: so that's actually a luajit bug?
> 16:21 < jcowgill> bunk: yes it's a jit miscompile

I've submitted my fix in the above PR. It has some gory details of how
it causes the segfault in neovim.

Thanks,
James



signature.asc
Description: OpenPGP digital signature

Bug#878817: closed by Gianfranco Costamagna <locutusofb...@debian.org> (Re: Bug#878817: luajit: Please drop mips64el again)

2017-10-20 Thread Adrian Bunk 
Control: reopen -1

16:20 < jcowgill> bunk, LocutusOfBorg: i have a patch for luajit which fixes 
neovim - i haven't tested it on any other lua code yet 
  though
16:21 < bunk> jcowgill: so that's actually a luajit bug?
16:21 < jcowgill> bunk: yes it's a jit miscompile


cu
Adrian

-- 

   "Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
   "Only a promise," Lao Er said.
   Pearl S. Buck - Dragon Seed

Bug#878817: luajit: Please drop mips64el again

2017-10-18 Thread James Cowgill 
Control: affects -1 - src:knot-resolver

Hi,

On 16/10/17 21:23, Adrian Bunk wrote:
> Source: luajit
> Version: 2.1.0~beta3+dfsg-5
> Severity: serious
> Control: affects -1 src:neovim src:knot-resolver
> 
> neovim and knot-resolver FTBFS with luajit 2.1.0~beta3+dfsg-5:
> https://buildd.debian.org/status/logs.php?pkg=neovim=mips64el
> https://buildd.debian.org/status/logs.php?pkg=knot-resolver=mips64el
> 
> knot-resolver FTBFS on the buildd, but builds for me on eller.
> 
> 21:10 < jcowgill> bunk: not sure, but i notice both failed on loongsons
> 21:10 < jcowgill> gcc has a few workarounds disabling some instructions since 
> loongson doesn't quite implement the mips isa properly
> 21:14 < bunk> jcowgill: that sounds like possibly related to a jit failing on 
> this hw
> 
> Instead of shipping code that is strongly suspected to be nonworking
> on part of the supported hardware of mips64el, it is IMHO better to
> disable luajit on mips64el again.

It turns out the knot-resolver build failure had nothing to do with
luajit (see #878976). I haven't investigated the neovim failure
thoroughly yet.

James



signature.asc
Description: OpenPGP digital signature

Bug#878817: luajit: Please drop mips64el again

2017-10-16 Thread Adrian Bunk 
Source: luajit
Version: 2.1.0~beta3+dfsg-5
Severity: serious
Control: affects -1 src:neovim src:knot-resolver

neovim and knot-resolver FTBFS with luajit 2.1.0~beta3+dfsg-5:
https://buildd.debian.org/status/logs.php?pkg=neovim=mips64el
https://buildd.debian.org/status/logs.php?pkg=knot-resolver=mips64el

knot-resolver FTBFS on the buildd, but builds for me on eller.

21:10 < jcowgill> bunk: not sure, but i notice both failed on loongsons
21:10 < jcowgill> gcc has a few workarounds disabling some instructions since 
loongson doesn't quite implement the mips isa properly
21:14 < bunk> jcowgill: that sounds like possibly related to a jit failing on 
this hw

Instead of shipping code that is strongly suspected to be nonworking
on part of the supported hardware of mips64el, it is IMHO better to
disable luajit on mips64el again.