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

Jérémy Bobbio lunar at debian.org
Tue Feb 18 23:35:54 UTC 2014


Stéphane Glondu:
> Sorry for spamming...

Thank you for spamming. It feels really great to see other people
working on the project. :)

> Le 15/02/2014 00:42, Jérémy Bobbio a écrit :
> > Picking a random package and trying to understand why it's not
> > reproducible usually takes me around 3 to 30 minutes. It's sometimes a
> > relaxing pause. Try it! ;)
>
> I've noticed that many packages differ because of so-called build IDs
> that are written to ELF files. Even hello is concerned. Is there any
> package that builds ELF files reproducible? Am I missing something?

With the current scheme, having the proper Build ID require two things:
usage of dpkg-buildflags and usage of debhelper to handle stripping
binaries.

In the case of hello, the build system does not dh_strip. So debugedit
does not get called and so the Build Id is not recomputed over the
stable source path.

With the previous rebuild, using -fdebug-prefix-map, this was less an
issue. One conclusion that can be made is that dpkg-buildflags +
debhelper approach is not the right one to solve the Build Id issue.

Except that most packages that calls strip directly probably don't
implement `DEB_BUILD_OPTIONS=nostrip` correctly. So we could see the
task of fixing these problems (together with dpkg-buildflags) as a good
opportunity to do archive-wide QA work.

But really, since I've started looking at it, I have yet to see/find a
nice solution for the build-path-in-DWARF issue.

-- 
Lunar                                .''`. 
lunar at debian.org                    : :Ⓐ  :  # apt-get install anarchism
                                    `. `'` 
                                      `-   
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.alioth.debian.org/pipermail/reproducible-builds/attachments/20140218/42e6cc11/attachment.sig>


More information about the Reproducible-builds mailing list