[Reproducible-builds] Bug#785742: dh-strip-nondeterminism: Causes package-contains-timestamped-gzip in lintian test suite

Niels Thykier niels at thykier.net
Tue May 26 19:24:48 UTC 2015


On 2015-05-26 20:35, Andrew Ayer wrote:
> tags 785742 + pending
> thanks
> 
> On Tue, 19 May 2015 20:56:17 +0200
> Niels Thykier <niels at thykier.net> wrote:
> 
>> Debugging the lintian FTBFS on reproducible.d.n[1], I have found that
>> if I override on dh_strip-nondeterminism in the test's rules file, the
>> unexpected package-contains-timestamped-gzip tag goes away.
>>
>> In general, all gzip files in the lintian test suite are generated
>> with "gzip -n" exactly to avoid this tag.
> 
> Thanks for the report, Niels.  strip-nondeterminism was actually adding
> back the timestamp using the time from the first entry in
> debian/changelog. I've just fixed it to not add a timestamp if the gzip
> file doesn't already have one.
> 

Thanks,

> That said, I'm now concerned about how strip-nondeterminism interacts
> with the package-contains-timestamped-gzip tag.  At some point after
> package-contains-timestamped-gzip was first proposed, we reproducible
> builds folks decided that instead of eliminating timestamps, we should
> normalize them with the time from the first entry in debian/changelog,
> to satisfy developers who like timestamps.  That means that timestamped
> gzips aren't bad, as long as the timestamp is reproducible.
> 
> Other reproducible folks, is this still our position on timestamps?  If
> so, what should become of package-contains-timestamped-gzip?
> 
> -- Andrew
> 

There is an open bug in lintian about the tag being a bit zealous.  I
believe it will be fixed in the next version of lintian.  That said, I
am very happy that strip-nondeterminism will now respect and retain the
results of the "gzip -n" option.

Thanks,
~Niels



More information about the Reproducible-builds mailing list