[Reproducible-builds] How to help out sorting logs?

[Stéphane Glondu]

Le 22/02/2014 14:07, Stéphane Glondu a écrit :
>  - there is some variation in xz output that is not detected by diffp

I got around that by building only binary (dpkg-buildpackage -b).

> Since the script evolved while I was using it, I am doing a second pass
> on the reproducible packages I've found (sanity check).

That was more eventful than expected. Eventually, I've rebuilt:
 - all packages marked reproducible in [1]
 - all packages with unidentifed status in [1]
 - all packages with timestamps it gzip headers listed in [2]
 - some packages that were marked as reproducible in [3]

In total, I've rebuilt 1188 packages. Among them:
 - 82 were unreproducible
 - 28 failed to build
 - 1078 were reproducible

Among the unreproducible ones, I've identified:
 - 1 contains varying benchmarks results
 - 4 contain varying egg-info
 - 3 contain varying emacs stuff
 - 5 contain varying jar files
 - 11 contain unsorted output
 - 15 contain unsorted Python Depends
 - 24 contain a single varying binary file
 - 12 contain timestamps
 - 7 are still unidentified

I've uploaded the logs to [4]. The tree structure follows the
description above.

[1] <https://wiki.debian.org/ReproducibleBuilds/Rebuild20140126>
[2] <https://wiki.debian.org/ReproducibleBuilds/TimestampInGzipHeaders>
[3] <https://wiki.debian.org/ReproducibleBuilds/GoodPackages>
[4] <http://reproducible.alioth.debian.org/logs-20140222/>


Cheers,

-- 
Stéphane