[CRIU] [PATCH] kdat: allow to override build time

[Pavel Emelyanov]

On 06/30/2017 04:58 PM, Bernhard M. Wiedemann wrote:
> On 2017-06-30 14:45, Pavel Emelyanov wrote:
>> On 06/30/2017 12:59 PM, Bernhard M. Wiedemann wrote:
>>> in order to allow for reproducible builds.
>>> See https://reproducible-builds.org/ for why this is good
>>> and https://reproducible-builds.org/specs/source-date-epoch/
>>> for the definition of this variable.
>>
>> Thanks, Brenhard!
>>
>> Acked-by: Pavel Emelyanov <xemul at virtuoozo.com>
>>
>> Is that the only thing that's required to make criu builds reproducible?
>> Presumably not, so is there anything we can help you with to make it happen?
> 
> actually, it was the only diff I needed to have 2 builds produce
> bit-identical results[1].
> But we have a very normalized build environment with constant locale,
> timezone, user, build path... so others may find more issues.
> 
>> I'm asking because criu's "build system", as you've noticed, is homegrown
>> and is about to be constantly tuned and fixed (probably be replaced) in the
>> future. So I'd like to align those potential changes with yours :)
> 
> It seems, we are closely following upstream with what we have in our
> openSUSE Tumbleweed (aka Factory) package
> because we already had your new line from 2017-05-04
> and I'll keep monitoring it for packages that do not build reproducibly.
> 
> So not much to do right now.
> When you make something new, you can try to get it right in the first
> place by sorting directory listing, sorting hashes (aka associative
> arrays), not including timestamps, hostnames, usernames and other build
> time information into the build output (except things like logs that we
> do not need to package).

Hm... Need to try :) Thanks!

> And if you do not get it right, I might be back... with patches!

:D

> Ciao
> Bernhard M.
> 
> P.S. I still wonder why my mails do not appear in
> https://lists.openvz.org/pipermail/criu/2017-June/date.html
> even though I subscribed to the ML 6h ago.

Indeed :\ But replies are there.

> [1] my
> https://github.com/bmwiedemann/reproducible-faketools/blob/master/bin/date
> even allows to get identical results without that patch.
> .
>