Re: Bug#1025708: bullseye-pu: package debootstrap/1.0.123+deb11u2

[Cyril Brulebois]

Simon McVittie  (2023-10-28):
> I believe dpkg-source defaults to the equivalent of `dpkg-source -I`
> for 3.0 (native) format packages, which ignores some files that would
> normally appear in git, notably .gitignore.
> 
> I normally use
> DEBUILD_DPKG_BUILDPACKAGE_OPTS="-us -uc -I.*.sw? -I.sw? -I.git" which
> disables the default `-I` and instead excludes .git but not .gitignore,
> making the uploaded source package exactly equivalent to what's in git
> (and as a result, more dgit-friendly).

Alright, that explains it then.

> If you would prefer any subsequent uploads of d-i-related components
> to always exclude the .gitignore, I'll try to remember that for the
> future.

Until 3.0 (git) was used everywhere, it was very customary to have some
differences in successive uploads, depending on who was uploading, and
whether -i/-I was used; it's not a huge deal, and only means a little
noise when reviewing diffs.

Whatever is fine with SRMs is fine with me. (It just happened to
surprise me a little when I compared a local source build with what was
uploaded and is available on coccia, since I built from my local repo
as usual instead of thinking about downloading your source packages from
the get-go.)


Cheers,
-- 
Cyril Brulebois (k...@debian.org)
D-I release manager -- Release team member -- Freelance Consultant


signature.asc
Description: PGP signature

Re: Bug#1025708: bullseye-pu: package debootstrap/1.0.123+deb11u2

[Simon McVittie]

On Sat, 28 Oct 2023 at 14:54:45 +0200, Cyril Brulebois wrote:
> Compared to what I get from a `dpkg-buildpackage -S` run locally (using
> the bullseye branch at tag debian/1.0.123+deb11u2), the source package
> available on coccia adds the debian/.gitignore file

I believe dpkg-source defaults to the equivalent of `dpkg-source -I`
for 3.0 (native) format packages, which ignores some files that would
normally appear in git, notably .gitignore.

I normally use
DEBUILD_DPKG_BUILDPACKAGE_OPTS="-us -uc -I.*.sw? -I.sw? -I.git" which
disables the default `-I` and instead excludes .git but not .gitignore,
making the uploaded source package exactly equivalent to what's in git
(and as a result, more dgit-friendly). If you would prefer any subsequent
uploads of d-i-related components to always exclude the .gitignore,
I'll try to remember that for the future.

smcv

Re: Bug#1025708: bullseye-pu: package debootstrap/1.0.123+deb11u2

[Cyril Brulebois]

Hi,

Simon McVittie  (2023-08-31):
> [ Reason ]
> The same changes proposed for bookworm in #1050868, but for bullseye.
> Because official buildds that build trixie/sid are not yet all running
> bookworm, we'll need this change in bullseye too.
> 
> I also included the changes that Luca previously proposed on this bug,
> which are backports from bookworm's debootstrap:
> 
> - no longer including usrmerge and its dependencies in the installed
>   system if usr-is-merged would be sufficient, saving ~ 50MB on a minbase
>   image and effectively fixing a regression caused by making
>   usrmerge|usr-is-merged transitively Essential in bookworm (#1025657)
> - enabling merged-/usr on Hurd
> 
> These are technically a behaviour change for bullseye, but we're making
> a larger behaviour change here already, and it aligns the behaviour
> with what we have in bookworm. We could revert those if required, but
> they're really small changes and seem desirable to me: in particular,
> they make the whole merged-/usr code path into the same tested code
> that's in trixie and proposed for bookworm.

Test results look good to me too, feel free to go ahead.

Compared to what I get from a `dpkg-buildpackage -S` run locally (using
the bullseye branch at tag debian/1.0.123+deb11u2), the source package
available on coccia adds the debian/.gitignore file; this is merely
intriguing and not something that should block processing this upload,
possibly linked to dgit's having been used at some point?


Cheers,
-- 
Cyril Brulebois (k...@debian.org)
D-I release manager -- Release team member -- Freelance Consultant


signature.asc
Description: PGP signature

Re: Bug#1025708: bullseye-pu: package debootstrap/1.0.123+deb11u2

[Simon McVittie]

On Thu, 31 Aug 2023 at 00:12:53 +0100, Simon McVittie wrote:
> I did the same testing as for bookworm's #1050868, summarized on
> .
> As with #1050868, all differences between the output of a reference
> version (from Debian 12.1) and the output of this version were expected
> or ignorable.

I have now also tried the proposed debootstrap_*.udeb via d-i, using
essentially the same steps as in
.
This time I installed XFCE rather than GNOME. Installation was successful.

Errata:
- The steps I wrote were incomplete, installing git and fakeroot on the
  build machine was also necessary
- As previously noted on #1050868, the proposed udeb needs to be copied
  into debian-installer_bullseye/build/localudebs/
- d-i 11 doesn't include "less", so the step involving viewing
  /usr/share/debootstrap/functions needs to use "more" instead

An additional note is that if you send Ctrl+Alt+F2 sufficiently early in
the "Installing the base system" step, and "ls -l /target" repeatedly, you
can see that as a result of Helmut's code changes, /target is initially
non-merged-/usr then gets converted to merged-/usr after unpacking the
Essential set.

smcv

Re: Bug#1025708: bullseye-pu: package debootstrap/1.0.123+deb11u2

[Simon McVittie]

On Thu, 31 Aug 2023 at 00:12:45 +0100, Simon McVittie wrote:
> [ Reason ]
> The same changes proposed for bookworm in #1050868, but for bullseye.
> Because official buildds that build trixie/sid are not yet all running
> bookworm, we'll need this change in bullseye too.
> 
> I also included the changes that Luca previously proposed on this bug,
> which are backports from bookworm's debootstrap:
> 
> - no longer including usrmerge and its dependencies in the installed
>   system if usr-is-merged would be sufficient, saving ~ 50MB on a minbase
>   image and effectively fixing a regression caused by making
>   usrmerge|usr-is-merged transitively Essential in bookworm (#1025657)
> - enabling merged-/usr on Hurd
> 
> These are technically a behaviour change for bullseye, but we're making
> a larger behaviour change here already, and it aligns the behaviour
> with what we have in bookworm. We could revert those if required, but
> they're really small changes and seem desirable to me: in particular,
> they make the whole merged-/usr code path into the same tested code
> that's in trixie and proposed for bookworm.
> 
> [ Impact ]
> If not accepted, trixie will continue to be stuck in a
> mostly-but-not-entirely merged-/usr limbo, with the moratorium from #1035831
> remaining in place (until all buildds can be upgraded to bookworm).

I'm aware that we're getting close to the deadline for 12.2 and 11.8,
so I've uploaded the proposed version to bullseye-proposed-updates for
easier testing and review. Luca: the proposed version and a signed tag
are available from my fork on salsa (I am not able to push to the d-i
repository for debootstrap). I uploaded with dgit, so the git tree and
the .dsc have been verified to be identical.

If this version is not accepted for whatever reason, then I think we
should treat version 1.0.123+deb11u2 as having been used, and skip
ahead to 1.0.123+deb11u3 for any subsequent bullseye update.
(And if there is a problem with having this version in bullseye-pu for
whatever reason, I'm happy to upload a +deb11u3 that is identical to
1.0.123+deb11u1 except for the changelog.)

Thanks,
smcv

Re: Bug#1025708: bullseye-pu: package debootstrap/1.0.123+deb11u2

[Simon McVittie]

Control: tags -1 = bullseye d-i

[ Reason ]
The same changes proposed for bookworm in #1050868, but for bullseye.
Because official buildds that build trixie/sid are not yet all running
bookworm, we'll need this change in bullseye too.

I also included the changes that Luca previously proposed on this bug,
which are backports from bookworm's debootstrap:

- no longer including usrmerge and its dependencies in the installed
  system if usr-is-merged would be sufficient, saving ~ 50MB on a minbase
  image and effectively fixing a regression caused by making
  usrmerge|usr-is-merged transitively Essential in bookworm (#1025657)
- enabling merged-/usr on Hurd

These are technically a behaviour change for bullseye, but we're making
a larger behaviour change here already, and it aligns the behaviour
with what we have in bookworm. We could revert those if required, but
they're really small changes and seem desirable to me: in particular,
they make the whole merged-/usr code path into the same tested code
that's in trixie and proposed for bookworm.

[ Impact ]
If not accepted, trixie will continue to be stuck in a
mostly-but-not-entirely merged-/usr limbo, with the moratorium from #1035831
remaining in place (until all buildds can be upgraded to bookworm).

[ Tests ]
I did the same testing as for bookworm's #1050868, summarized on
.
As with #1050868, all differences between the output of a reference
version (from Debian 12.1) and the output of this version were expected
or ignorable. In addition to the ignorable differences noted in #1050868,
/etc/apt/apt.conf.d/01autoremove-kernels reflected the use of a bullseye
rather than bookworm kernel.

The autopkgtest passes, as in #1050868.

[ Risks ]
Same as in #1050868.

Additionally, I suppose someone could conceivably be relying on debootstrap
of bookworm|trixie|sid from a bullseye host installing a non-minimal system
with Perl and various libraries? But that seems tenuous.

[ Checklist ]
  [x] *all* changes are documented in the d/changelog
  [x] I reviewed all changes and I approve them
  [x] attach debdiff against the package in (old)stable
  [x] the issue is verified as fixed in unstable (and trixie)

[ Changes ]

The addition of EXCLUDE_DEPENDENCY in ./debootstrap, and the logic that
uses it (last hunk in ./functions and second hunk in scripts/debian-common),
are for #1025657.

setup_merged_usr() in ./functions now allows merged-/usr to continue on
Hurd too.

In copying setup_merged_usr() from trixie (via bookworm), it seems we've
accidentally enabled correct merged-/usr handling for loongarch64, which
wasn't exactly intentional but seems harmless!

scripts/trixie now exists (it's a symlink, the same as for every other
Debian suite).

The rest is the same as in bookworm's #1050868.

[ Other info ]
To keep bookworm's debootstrap "better" than bullseye's, this update
should not be accepted until after #1050868 is.

smcv
diffstat for debootstrap-1.0.123+deb11u1 debootstrap-1.0.123+deb11u2

 debian/.gitignore  |6 +
 debian/changelog   |   44 +
 debian/gbp.conf|3 
 debian/salsa-ci.yml|1 
 debian/tests/debian-testing|   27 +++
 debian/tests/fake/schroot-1.6.10-3 |2 
 debootstrap|4 +
 debootstrap.8  |5 -
 functions  |  125 +
 scripts/amber  |2 
 scripts/artful |2 
 scripts/bionic |2 
 scripts/cosmic |2 
 scripts/debian-common  |9 +-
 scripts/disco  |2 
 scripts/eoan   |2 
 scripts/focal  |2 
 scripts/gutsy  |2 
 scripts/hardy  |2 
 scripts/intrepid   |2 
 scripts/jaunty |2 
 scripts/karmic |2 
 scripts/lucid  |2 
 scripts/maverick   |2 
 scripts/natty  |2 
 scripts/oneiric|2 
 scripts/precise|2 
 scripts/quantal|2 
 scripts/raring |2 
 scripts/saucy  |2 
 scripts/trixie |   16 
 scripts/trusty |2 
 scripts/utopic |2 
 scripts/vivid  |2 
 scripts/wily   |2 
 scripts/xenial |2 
 scripts/yakkety|2 
 scripts/zesty  |2 
 38 files changed, 246 insertions(+), 50 deletions(-)

diff -Nru debootstrap-1.0.123+deb11u1/debian/changelog debootstrap-1.0.123+deb11u2/debian/changelog
---