• Make d-i packages usrmerge-ready?

    From Holger Wansing@21:1/5 to All on Thu Dec 19 20:00:01 2024
    Hi all,

    I just noticed that many d-i packages currently get a lintian error "aliased-location", complaining about binaries in locations, which are
    now under the concept of usrmerge (aka /bin, /sbin and /lib).

    Since udebs are sometimes a little bit different from "normal" Debian packages, I would like to ask, if this is an issue for these packages or not.

    Should we change this, or is the lintian error a false-positive?
    And do we want to change it *now*?

    Pascal 'pham' already mentioned, that d-i under trixie has merged-usr
    in initrd.gz, but I want to make sure, before pushing changes to several
    d-i packages, thus this mail...

    See for example https://salsa.debian.org/installer-team/grub-installer/-/merge_requests/23 https://salsa.debian.org/installer-team/anna/-/merge_requests/5

    Testing of images with such changes did not show any issues.


    Holger


    --
    Holger Wansing <hwansing@mailbox.org>
    PGP-Fingerprint: 496A C6E8 1442 4B34 8508 3529 59F1 87CA 156E B076

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Cyril Brulebois@21:1/5 to All on Fri Dec 20 02:30:02 2024
    Hello,

    Holger Wansing <hwansing@mailbox.org> (2024-12-19):
    I just noticed that many d-i packages currently get a lintian error "aliased-location", complaining about binaries in locations, which are
    now under the concept of usrmerge (aka /bin, /sbin and /lib).

    Since udebs are sometimes a little bit different from "normal" Debian packages, I would like to ask, if this is an issue for these packages
    or not.

    Important topics:
    - does the package build?
    - does it get accepted when it reaches the archive?
    - is the runtime within d-i OK?

    If you get a triple yes, the rest doesn't matter.

    Should we change this, or is the lintian error a false-positive?

    At this point, I don't know, and I don't want to know. This move has
    been a HUGE mess for MANY years. And if we don't run into troubles as it
    is, then I don't see a need to change anything.

    If that doesn't lead to an autoreject, it's perfectly fine to ignore
    those errors, and not to add any overrides. It's best to think about
    the topic first, then decide whether we need/want to do something on the
    udeb side, get lintian to skip those checks on udebs, and/or use
    overrides.

    (My gut feeling would be we should have udebs close to their counterpart
    debs, but should it happen right now? Absolutely not. :))

    And do we want to change it *now*?

    Even if we needed, or wanted, to do anything about that, it would make
    sense to get everything at the same time, not one package here and one
    other package there.

    Ideally that kind of change should get discussed on the list first, and
    if there's an agreement about needing/wanting to implement the changes
    vs. the associated risks, then we can talk about a timetable and
    proceed. So thank you very much for starting this thread, that's
    perfect.

    Pascal 'pham' already mentioned, that d-i under trixie has merged-usr
    in initrd.gz, but I want to make sure, before pushing changes to
    several d-i packages, thus this mail...

    See for example https://salsa.debian.org/installer-team/grub-installer/-/merge_requests/23 https://salsa.debian.org/installer-team/anna/-/merge_requests/5

    Testing of images with such changes did not show any issues.

    That's a useful datapoint, thanks for that as well.


    Over the past few weeks I've spotted a number of issues that'd be
    blockers for a release, and most if not all of them have been fixed in
    the last few days.

    If you want to include more l10n changes at this point that's still
    fine, but delaying other code changes would have my preference at this
    point. I'll have to check a few more things before mailing the usual
    reminder to dda@ about freezing udeb-producing packages temporarily,
    check with the images team, and wade through the website and other
    places to prepare for Trixie Alpha 1.

    (There's some bit of uncertainty regarding the unshare thing on the
    buildd side — I'm not sure I looked into the proper puppet bits — but
    we'll see what happens when src:debian-installer gets uploaded next
    time.)


    Cheers,
    --
    Cyril Brulebois (kibi@debian.org) <https://debamax.com/>
    D-I release manager -- Release team member -- Freelance Consultant

    -----BEGIN PGP SIGNATURE-----

    iQIzBAABCgAdFiEEtg6/KYRFPHDXTPR4/5FK8MKzVSAFAmdkxxQACgkQ/5FK8MKz VSAsdg//bJk2QQsLW5VLwpEeGtcp+KI/Lga/+sGEZ/9A7YnDDj6rJkj8gKDpfc3A Vep8mAW8eSfj/12UaVQAqce2LwsCBXmELhxmvw60c082zr34LUVsDkO5kCVHRAmq AijpWLUNUDObASjsZZofg7Y5qPg2BSm1e2b3YHQflQy2Ia/hr5oVCntRyn4lsMaP KjwgDuQbBjFbVWnEEJ6s9S2xnA0OzN9Iu9k5VqdtmeoEMBWYsYohQfbbgSnK8iOZ OsODwtcn+G+W6as+ogp+qOMHZN7Dj2ezGk18n6zF6+mVpIBvC7VCrvXXSeyBedGF 6852YuYNB/KnoSyDDWp0eWzTepl0eM2kHzh+Ecvx5KHAqDH++JE3w9VmSmLHFeGx 1EsrOru3W7L8ez9KJ+HA0QMfsEGD4yGml3zNgMOjFm2MeA6LFcI9QCXv9KaTj7Ov watc1n4OgPUIOBV4us4WhpB5FtDYaespcid8vrfpFc0Mke/vwK78dKki9YQQXb1R 9VpNARUzeTkC7vka5vdeBHJn6Gs5uc0ftLKGAs6jYcyx8QGwzI4UiLsB0zHyE9mo vqGaSGtzdaUW12KveUnT2nl3wt1c93eM+/D+FniRaTYD4/Jwh7XdvsrxIwZLULVv DS2BbMhYWwlPCqp0o08D1M6YXIv+j0H8h+G3mHC5QZFuFN7QMLw=
    =O97D
    -----END PGP SIGNATURE-----

    --- SoupGate-Win32 v1.05
    *
  • From Cyril Brulebois@21:1/5 to All on Fri Dec 20 12:30:03 2024
    Helmut Grohne <helmut@subdivi.de> (2024-12-20):
    Assuming that there are no conflicts, my understanding is that moving
    is reasonably safe since d-i has been merged. On the flip side, I also
    do not see urgency in doing the move.

    The “no urgency” was my bottom line as well.

    (https://lists.debian.org/debian-boot/2024/12/msg00121.html if you
    haven't seen it already.)

    Now let me turn your question back: Should I exempt udebs in lintian?

    If you don't mind adding a check/exemption now, and reverting it “later” (either in a few days/weeks or after trixie is published, we wouldn't
    touch that during the freeze) once we agree we want to start doing the
    work (possibly modifying most if not all udebs at the same time), that
    would be perfect I think: people wouldn't have to worry about this when
    they touch udeb-producing packages (which I suppose is what just
    happened to Jonathan when he prepared rootskel-gtk).

    Thanks!


    Cheers,
    --
    Cyril Brulebois (kibi@debian.org) <https://debamax.com/>
    D-I release manager -- Release team member -- Freelance Consultant

    -----BEGIN PGP SIGNATURE-----

    iQIzBAABCgAdFiEEtg6/KYRFPHDXTPR4/5FK8MKzVSAFAmdlU8MACgkQ/5FK8MKz VSBIsw//SSmIaEOPsz/DV3JDdMJgVXS5wGCKqy8/wxpL/1R3Z0UNlsXnvE/q9/X8 BHF1NkWdaBsYBf3Uj3j34KQG/NFln4twPiHDiBSbR1a7f5uQkBqStoaR28yqDr2s A3EJOzJlI8pWyLjTegRwJMgAAKyJeDSLY+m40nleAcLFJuy6Pow8rl54pPDHEROs ymRTNZ/opmY0JYE30H0M8LLP3M4J9ecsNUAm3BL+5K2TRHXsGBiO/IUsLX9t7qhs Ls0sEO82ntiHpNK5aPmPr0a6SJ4a0V4Kf9CFBk6F5fkQ3bc2ObYoUs1o/B7v8aND 6FQvKhUApSaYTtXMaG3cwQHskoU6R8puuWCqWjDnymJAN5KkiaG9iUksIj39129Z FWNZ0M/Tq0BFwTOQWRz3yu2h12KCuW3vayL0YwpJjVegtHY9siQTsuYZcMi2YPma Y1ZMx9WCWgPoWuJ4PFVzAFZpcavewBXis8aR1on4u1RyLmvuxtj+Vi3AXUWPAY9j 884y0T+NxHX+E8UxT6yRvv22DrtCE94a/5Tn9DA7EMvPpPNaTFelZkmCSSY77LKN eW/DUXeSBBO3Y/U4/Io6AODwMvMIpYvZDUmvu7bNvSOj/+VocfDQksK4Pub1ofeE Pyqyw6X8r9LEx7XCOWrppBkZPdSUm1uo0NpFqClXSm4MW5OLtSo=
    =9i9c
    -----END PGP SIGNATURE-----

    --- SoupGate-Win32 v1.05
    *
  • From Helmut Grohne@21:1/5 to Holger Wansing on Fri Dec 20 12:20:02 2024
    Hi Holger,

    Please Cc me in replies as I am not subscribed to d-boot.

    On Thu, Dec 19, 2024 at 07:59:19PM +0100, Holger Wansing wrote:
    I just noticed that many d-i packages currently get a lintian error "aliased-location", complaining about binaries in locations, which are
    now under the concept of usrmerge (aka /bin, /sbin and /lib).

    I introduced this lintian check and did not have udebs in mind at the
    time of writing it. The warning was not intentional and we should figure
    out whether it is better to keep or relax it.

    Since udebs are sometimes a little bit different from "normal" Debian packages,
    I would like to ask, if this is an issue for these packages or not.

    It is much less of an issue that for the rest of Debian, because udebs
    are never upgraded. You only use them to install once. As a result, much
    of the problems that motivated the moving do not apply to udebs.

    Should we change this, or is the lintian error a false-positive?
    And do we want to change it *now*?

    I agree with the questions but provide no answers.

    Pascal 'pham' already mentioned, that d-i under trixie has merged-usr
    in initrd.gz, but I want to make sure, before pushing changes to several
    d-i packages, thus this mail...

    See for example https://salsa.debian.org/installer-team/grub-installer/-/merge_requests/23 https://salsa.debian.org/installer-team/anna/-/merge_requests/5

    Testing of images with such changes did not show any issues.

    I also implemented the /usr-merging of d-i. It technically works the
    same way as debootstrap now. That is all the udebs are extracted into
    the same location and then files are shuffled around if necessary. If
    your udeb contains /bin/foo, it'll be moved to /usr/bin/foo after all
    udebs are extracted. If you do the move before creating the .udeb little
    should change.

    Famous last words, right? There is a semantic change. The moving code is
    very picky about file conflicts. If one udeb installs /bin/foo and
    another installs /usr/bin/foo, that makes it fail hard. If you move
    /bin/foo, it stops failing and chooses an undefined one. If two udebs
    install /bin/foo and you move one, it starts failing. So all those
    file conflicts between udebs now pose to be a problem. I reported some
    of those when I implemented the merge, but they're not all fixed yet. I
    think busybox still conflicts with two other udebs. So in the interest
    of not breaking stuff, you should resolve those conflicts before moving
    the files.

    Resolving those conflicts would be very useful for independent of the /usr-move. At this time, which file happens to be used depends on the
    unpack order and that happens to be poorly defined (but likely stable).

    Assuming that there are no conflicts, my understanding is that moving is reasonably safe since d-i has been merged. On the flip side, I also do
    not see urgency in doing the move.

    Now let me turn your question back: Should I exempt udebs in lintian?

    Helmut

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Holger Wansing@21:1/5 to All on Fri Dec 20 17:00:01 2024
    Hi,

    Am 20. Dezember 2024 02:23:37 MEZ schrieb Cyril Brulebois <kibi@debian.org>: >Important topics:
    - does the package build?
    - does it get accepted when it reaches the archive?
    - is the runtime within d-i OK?

    If you get a triple yes, the rest doesn't matter.

    That was the easy part, and I got Yes for all three (for the
    two it tried - anna + grub-installer).

    Should we change this, or is the lintian error a false-positive?

    At this point, I don't know, and I don't want to know. This move has
    been a HUGE mess for MANY years. And if we don't run into troubles as it
    is, then I don't see a need to change anything.

    For the lintian part there is a follow-up in another thread.

    But I just would like to point out:
    since the d-i initrd.gz is already usr-merged, it seems inconsistent
    to me, if we leave the d-i packages completely unchanged.
    Just my 2 cent


    If you want to include more l10n changes at this point that's still
    fine, but delaying other code changes would have my preference at this
    point. I'll have to check a few more things before mailing the usual
    reminder to dda@ about freezing udeb-producing packages temporarily,
    check with the images team, and wade through the website and other
    places to prepare for Trixie Alpha 1.

    Feel free to upload packages for an alpha, or not. As you wish.
    For a first alpha the translation status is sufficient IMO.


    Holger





    --
    Sent from /e/ OS on Fairphone3

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)