SCIENTIFIC-LINUX-USERS Archives

July 2014

SCIENTIFIC-LINUX-USERS@LISTSERV.FNAL.GOV

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Lamar Owen <[log in to unmask]>
Reply To:
Lamar Owen <[log in to unmask]>
Date:
Wed, 2 Jul 2014 13:15:23 -0400
Content-Type:
text/plain
Parts/Attachments:
text/plain (24 lines)
On 07/02/2014 08:41 AM, Brett Viren wrote:
> To add, if deterministic builds were not possible it would mean this 
> could not exist: http://nixos.org/nix/ Regards, -Brett. 

I don't see anywhere on that site where they claim that any given source 
package, when build, will always produce a bit-for-bit identical binary 
package. The hash that's mentioned is of the dependency graph, not the 
actual package binary.

They seem to define a 'reproducible' build as "Nix builds packages in 
isolation from each other. This ensures that they are reproducible and 
don’t have undeclared dependencies, so if a package works on one 
machine, it will also work on another." This is 'functionally 
compatible.' (It also sounds a lot like how mock is used to build rpms 
from source rpms; this is very much an improvement over how builds were 
done in the days before mock and its predecessor, mach.)

Russ mentions parallel builds; I will have to try that on my 30 CPU 
Altix box at some point; I'll take a source package (I'll probably do 
the kernel) and rebuild it, wait a day, rebuild the same source RPM 
again (with the same buildroot of the same binaries), and compare the 
binary packages. With a -j 28 argument to make it would show this sort 
of artifact most easily.

ATOM RSS1 RSS2