[rb-general] Flathub, initial proposal

[Holger Levsen]

Hi,

thanks for reaching out to us reproducible builds folks!

On Wed, Aug 31, 2016 at 08:30:25AM +0200, Alexander Larsson wrote:
> I'm very interested in reproducible builds, and I love what the debian
> reproducible build people are doing.  Flatpak has several features that
> make reproducible builds easier. For instance, we build with a well
> defined SDK, with no access to anything else, plus the builds are
> always in the same directory (under /run/build/) to make debug info the
> same, and we have some .pyc mtime rewrite hackery.

nice!

how's the SDK defined? Will I be able to use this SDK in 2 or 5 years to
reproduce a build from today?

> However, getting fully reproducible builds is all about tweaking the
> individual build of each app. There is no way we can make every build
> automatically reproducible, or require that they will be.
> 
> Basically, our tooling helps make reproducible builds easier, and we
> can add features making it even easier, but we can't guarantee/require
> it for every app.

while I agree with this last paragraph I disagree with the conlusion of
the first paragraph above :) In the long run (think 5-10y) I hope that
all software will be reproducible, because the community will not accept
unreproducible software anymore…

It's really not that hard, just tedious to get done in thousands of
projects ;-)

Most of the GNOME apps are actually reproducile, if you carefully look
at
https://tests.reproducible-builds.org/debian/testing/amd64/pkg_set_gnome.html
:-)

regarding Flatpak, have you tried building something twice and comparing the
results with diffoscope? (https://try.diffoscope.org) - we would *love*
to hear about the results of this simple test!


-- 
cheers,
	Holger
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 811 bytes
Desc: Digital signature
URL: <http://lists.reproducible-builds.org/pipermail/rb-general/attachments/20160903/52241405/attachment.sig>