[Pkg-giraffe-discuss] FW: FD_SETSIZE

[Mark Dufour]

hi all,

before anyone decides to look further into the FD_SETSIZE issue, please see below for a comment by jan about this. my mistake probably that this went off-list.


cheers,
mark.


-----Original message-----
> From:Jan Engelhardt <jengelh at inai.de>
> Sent: Monday 24th July 2017 22:27
> To: Guido Günther <agx at sigxcpu.org>
> Cc: Mark Dufour <m.dufour at kopano.com>
> Subject: Re: [Pkg-giraffe-discuss] preparations for kopanocore 8.3.x (or 8.4.x)
> 
> 
> On Monday 2017-07-24 20:15, Guido Günther wrote:
> >> >I'm inclinde to drop the madness and not redefine __FD_SETSIZE and
> >> >therefore have hardening for a network facing service.
> >> >
> >> >Patch attached.
> >> 
> >> That will be causing crashes, since we have file descriptors above 1023.
> >
> >Thanks for your brevity ;) I just grepped for RLIMIT_NOFILE and found
> >ec_relocate_fd. That explains why you get away with it without having
> >everything crash. I'll keep the patch in Debian for now and rather drop
> >ec_relocate_fd and see what happens. Maybe I give epoll a whirl should
> >we run into issues.
> 
> select+__FD_SETSIZE has already been removed anyway.
> 
> commit db9a4d3e8d23a98d47866089c3859d76032a7340
> Author: Jan Engelhardt <jengelh at inai.de>
> Date:   Tue Mar 28 22:38:53 2017 +0200
> 
>     common: abolish FD_SETSIZE hack
>     
>     Overriding FD_SETSIZE is only really supported on Windows, not POSIX.
>     Since KC uses no more select(2), the attempt to redefine glibc
>     internals can go away now.
> 
> >What about joining pkg-giraffe-discuss at lists.alioth.debian.org, your
> >input would be very valuable!
> 
> I think I already did. But I just noticed the mbox archives don't
> contain the list address, so the previous reply did not make it to the list
> despite Reply-All.
>