Bug#805321: [Reproducible-builds] Bug#805321: debian-installer: builds unreproducible netboot images

[Steven Chamberlain]

Hi,

Cyril Brulebois wrote:
> I've cherry-picked 3 patches from there onto master locally and I'm
> currently running diffoscope to see how that goes (and it's taking
> ages…):

I'm guessing the initrd would differ if the Linux tool to generate it
stores timestamps.  If its compressed size varies much due to those
differences, the .iso block numbers may vary as a result.

> FWIW, I'm not exactly entirely convinced by the exporting of the
> SOURCE_DATE_EPOCH variable from debian/rules; all other variables have
> been passed without exporting so I'm wondering if we shouldn't adapt
> this to behave like other variables, reducing possible surprise for
> users.

Just to explain that -- if it's defined in the environment, it requires
no special handling and doesn't need to be (re-)exported.  I think this
is maybe the case now for dpkg-buildpackage in sid?

If the dpkg-buildpackage environment doesn't have SOURCE_DATE_EPOCH
(e.g. jessie), debian/rules sets it to the correct value, and so must
export that.

Or (for jessie or sid), in case build/Makefile is used directly, outside
of a package build, we set SOURCE_DATE_EPOCH to a dummy value ("now") if
undefined (since ../debian/changelog may not exist), which we need
when calling makefs from within that Makefile.  We export it for use by
gen-tarball to avoid duplication there.

Regards,
-- 
Steven Chamberlain
ste...@pyro.eu.org


signature.asc
Description: Digital signature

Bug#805321: [Reproducible-builds] Bug#805321: debian-installer: builds unreproducible netboot images

[Cyril Brulebois]

Cyril Brulebois  (2015-11-26):
> I've cherry-picked 3 patches from there onto master locally and I'm
> currently running diffoscope to see how that goes (and it's taking
> ages…):
>   c182491b05fec16497f2bf1290cac16773d175f9
>   5d59fd1813e794d0821c00757dd56fd9ca25ed16
>   d126622567cfbe10d7f8a207a292eaab622ef73e

The following files are different after two builds with debuild -b once
the 3 commits are applied:
./installer-amd64/20151024/images/cdrom/gtk/debian-cd_info.tar.gz
./installer-amd64/20151024/images/cdrom/gtk/initrd.gz
./installer-amd64/20151024/images/cdrom/initrd.gz
./installer-amd64/20151024/images/hd-media/boot.img.gz
./installer-amd64/20151024/images/hd-media/gtk/initrd.gz
./installer-amd64/20151024/images/hd-media/initrd.gz
./installer-amd64/20151024/images/netboot/debian-installer/amd64/initrd.gz
./installer-amd64/20151024/images/netboot/gtk/debian-installer/amd64/initrd.gz
./installer-amd64/20151024/images/netboot/gtk/netboot.tar.gz
./installer-amd64/20151024/images/netboot/netboot.tar.gz

I see at least timestamp issues in initramfses; this might be due to the
fact that pigz (installed in my development chroots) seems to need both
-n and -T to behave like gzip's -n.

I might poke a bit more around this before pushing.

Mraw,
KiBi.


signature.asc
Description: Digital signature

Bug#805321: [Reproducible-builds] Bug#805321: debian-installer: builds unreproducible netboot images

[Cyril Brulebois]

Hi,

Steven Chamberlain  (2015-11-22):
> I rewrote the patches according to KiBi's feedback and they are
> now uploaded to our jessie-kfreebsd suite, and this Git branch:
> https://anonscm.debian.org/cgit/d-i/debian-installer.git/log/?h=jessie-kfreebsd

I've cherry-picked 3 patches from there onto master locally and I'm
currently running diffoscope to see how that goes (and it's taking
ages…):
  c182491b05fec16497f2bf1290cac16773d175f9
  5d59fd1813e794d0821c00757dd56fd9ca25ed16
  d126622567cfbe10d7f8a207a292eaab622ef73e

> In my own testing on ZFS, file ordering was still an issue for the
> makefs tool that builds the initrd.  But if I were to try again
> on UFS, I hope to be able to reproduce the entire
> netboot-installer-images tarball as built by the buildds.
> 
> This tarball includes bits that are bundled onto the official release
> images by debian-cd tools.  Making this reproducible is a prerequisite
> for someday having reproducibly-built official release images.
> 
> I could merge these patches into sid if they seem okay?  The only
> commit that should not be merged is this one, which is specific to
> jessie-kfreebsd and must be slightly changed for sid:
>   kfreebsd: use makefs -T to clamp timestamps

I suppose your time is better spent actually working on kfreebsd so
that's why I decided to cherry-pick the patches myself.

FWIW, I'm not exactly entirely convinced by the exporting of the
SOURCE_DATE_EPOCH variable from debian/rules; all other variables have
been passed without exporting so I'm wondering if we shouldn't adapt
this to behave like other variables, reducing possible surprise for
users.

I don't think that's a showstopper for a push to master though; just
thinking out loud.

> I expect that Linux d-i builds will have some reproducibility issues
> in whatever generates the initrd or ISOs, but I may look into that
> after the jessie-kfreebsd release is done.

Sure thing, thanks again!

Mraw,
KiBi.


signature.asc
Description: Digital signature

Bug#805321: [Reproducible-builds] Bug#805321: debian-installer: builds unreproducible netboot images

[Steven Chamberlain]

Hi,

I rewrote the patches according to KiBi's feedback and they are
now uploaded to our jessie-kfreebsd suite, and this Git branch:
https://anonscm.debian.org/cgit/d-i/debian-installer.git/log/?h=jessie-kfreebsd

In my own testing on ZFS, file ordering was still an issue for the
makefs tool that builds the initrd.  But if I were to try again
on UFS, I hope to be able to reproduce the entire
netboot-installer-images tarball as built by the buildds.

This tarball includes bits that are bundled onto the official release
images by debian-cd tools.  Making this reproducible is a prerequisite
for someday having reproducibly-built official release images.

I could merge these patches into sid if they seem okay?  The only
commit that should not be merged is this one, which is specific to
jessie-kfreebsd and must be slightly changed for sid:
kfreebsd: use makefs -T to clamp timestamps

I expect that Linux d-i builds will have some reproducibility issues
in whatever generates the initrd or ISOs, but I may look into that after
the jessie-kfreebsd release is done.

Regards,
-- 
Steven Chamberlain
ste...@pyro.eu.org


signature.asc
Description: Digital signature