It's been a LONG day: two point releases in a day takes of the order of twelve or thirteen hours of fairly solid work on behalf of those doing the releases and testing.Thanks firstly to the main Debian release team for all the initial work.Thanks to Isy, RattusRattus, Sledge and egw in Cottenham, smcv and Helen closer to the centre of Cambridge, cacin and others who have dropped in and out of IRC and helped testing.

I've been at home but active on IRC - missing the team (and the food) and drinking far too much coffee/eating too many biscuits.

We've found relatively few bugs that we haven't previously noted: it's been a good day. Back again, at some point a couple of months from now to do this all over again.With luck, I can embed a picture of the Cottenham folk below: it's fun to know _exactly_ where people are because you've been there yourself.

Motivation
Andy (RattusRattus) and I have been formalising instructions for using Pete Batard's version of Tianocore (and therefore UEFI booting) for the Raspberry Pi 4 together with a Debian arm64 netinst to make a modified Debian installer on a USB stick which "just works" for a Raspberry Pi 4.
Thanks also to Steve McIntyre for initial notes that got this working for us and also to Emmanuele Rocca for putting up some useful instructions for copying.

Recipe

Plug in a USB stick - use dmesg or your favourite method to see how it is identified.

Make a couple of mount points under /mnt - /mnt/data and /mnt/cdrom

1. Grab a USB stick, Partition using MBR. Make a single VFAT
partition, type 0xEF (i.e. EFI System Partition)

For a USB stick (identified as sdX) below:
$ sudo parted --script /dev/sdX mklabel msdos $ sudo parted --script /dev/sdX mkpart primary fat32 0% 100% $ sudo mkfs.vfat /dev/sdX1 $ sudo mount /dev/sdX1 /mnt/data/

Download an arm64 netinst.iso

https://cdimage.debian.org/debian-cd/current/arm64/iso-cd/debian-12.2.0-arm64-netinst.iso

2. Copy the complete contents of partition *1* from a Debian arm64
installer image into the filesystem (partition 1 is the installer
stuff itself) on the USB stick, in /

$ sudo kpartx -v -a debian-12.2.0-arm64-netinst.iso # Mount the first partition on the ISO and copy its contents to the stick $ sudo mount /dev/mapper/loop0p1 /mnt/cdrom/ $ sudo rsync -av /mnt/cdrom/ /mnt/data/ $ sudo umount /mnt/cdrom

3. Copy the complete contents of partition *2* from that Debian arm64
installer image into that filesystem (partition 2 is the ESP) on
the USB stick, in /

# Same story with the second partition on the ISO

$ sudo mount /dev/mapper/loop0p2 /mnt/cdrom/

$ sudo rsync -av /mnt/cdrom/ /mnt/data/ $ sudo umount /mnt/cdrom

$ sudo kpartx -d debian-testing-amd64-netinst.iso $ sudo umount /mnt/data

4. Grab the rpi edk2 build from https://github.com/pftf/RPi4/releases
(I used 1.35) and extract it. I copied the files there into *2*
places for now on the USB stick:

/ (so the Pi will boot using it)
/rpi4 (so we can find the files again later)

5. Add the preseed.cfg file (attached) into *both* of the two initrd
files on the USB stick

- /install.a64/initrd.gz and
- /install.a64/gtk/initrd.gz

cpio is an awful tool to use :-(. In each case:

$ cp /path/to/initrd.gz .
$ gunzip initrd.gz
$ echo preseed.cfg cpio -H newc -o -A -F initrd

$ gzip -9v initrd

$ cp initrd.gz /path/to/initrd.gz

If you look at the preseed file, it will do a few things:

- Use an early_command to unmount /media (to work around Debian bug
#1051964)

- Register a late_command call for /cdrom/finish-rpi (the next
file - see below) to run at the end of the installation.

- Force grub installation also to the EFI removable media path,
needed as the rpi doesn't store EFI boot variables.

- Stop the installer asking for firmware from removable media (as
the rpi4 will ask for broadcom bluetooth fw that we can't
ship. Can be ignored safely.)

6. Copy the finish-rpi script (attached) into / on the USB stick. It
will be run at the end of the installation, triggered via the
preseed. It does a couple of things:

- Copy the edk2 firmware files into the ESP on the system that's
just been installer

- Remove shim-signed from the installed systems, as there's a bug
that causes it to fail on rpi4. I need to dig into this to see
what the issue is.

That's it! Run the installer as normal, all should Just Work (TM).

BlueTooth didn't quite work : raspberrypi-firmware didn't install until adding a symlink for boot/efi to /boot/firmware

20231127 - This may not be necessary because raspberrypi-firmware path has been fixed

Preseed.cfg
# The preseed file itself causes a problem - the installer medium is
# left mounted on /medis so things break in cdrom-detect. Let's see if
# we can fix that!
d-i preseed/early_command string umount /media true

# Run our command to do rpi setup before reboot
d-i preseed/late_command string /cdrom/finish-rpi

# Force grub installation to the RM path
grub-efi-arm64 grub2/force_efi_extra_removable boolean true

# Don't prompt for missing firmware from removable media,
# e.g. broadcom bluetooth on the rpi.
d-i hw-detect/load_firmware boolean false

Finish.rpi
!/bin/sh

set -x

grep -q -a RPI4 /sys/firmware/acpi/tables/CSRT
if [ $? -ne 0 ]; then
echo "Not running on a Pi 4, exit!"
exit 0
fi

# Copy the rpi4 firmware binaries onto the installed system.
# Assumes the installer media is mounted on /cdrom.
cp -vr /cdrom/rpi4/. /target/boot/efi/.

# shim-signed doesn't seem happy on rpi4, so remove it
mount --bind /sys /target/sys
mount --bind /proc /target/proc
mount --bind /dev /target/dev

in-target apt-get remove --purge --autoremove -y shim-signed

At Arm for two days before the mini-Debconf this weekend.First time at Arm for a few years: huge new buildings, shiny lecture theatre.Arm have made us very welcome. A superb buffet lunch and unlimited coffee plus soft drinks - I think they know what Debian folk are like.
Not enough power blocks laid out at the beginning - only one per table - but we soon fixed that
The room is full of Debian folk: some I know, some new faces. Reminiscing about meeting some of them from 25 years ago - and the chance to thank people for help over a long time.
Andy (RattusRattus) and I have been working out the bugs on an install script using UEFI for a Raspberry Pi 4. More on that in the next post, maybe.As ever, it's the sort of place where "I can't get into the wiki" is sorted by walking three metres across the room or where an "I can't find where to get X for Raspberry Pi" can be solved by asking the person who builds Raspbian. "Did you try and sign up to the Debian wiki last week - you didn't follow the instructions to mail wiki@ - I _know_ you didn't because I didn't see the mail ... "

My kind of place and my kind of people, as ever.

Thanks again to Arm who are one of our primary sponsors for this mini-Debconf.

Over in Cambridge with RattusRattus, Sledge, egw and Isy. Andy is very kindly putting us up.

We're almost all of the way through testing 12.2 and some of the way through testing 11.8.

It's a LONG day - heads down into laptops and relatively quiet - I think we're all tired and we've a way to go yet.

There's been a very successful barbeque running in the garden: burgers, sausages, beer, vegetarian dishes and then ice cream.The chance to catch up with people you only meet in IRC. Talking and laughter - and probably a couple of games of Mao.Thanks also to our sponsors - Collabora, Codethink and RattusRattus for contributions to food and drink.

And so we're back at Steve's in Cambridge for release testing. A few testers here: we're now well into testing the various iso images.So we have Sledge, RattusRattus, Isy, smcv and myself. We've also been joined here by Helen who has just done her first install. Online we've got a couple of new folk - luna, The_Blode. Thanks to everyone involved - we always need all the help we can get.
Cold and grey outside: usual warmth in here. Also the usual snake of cables to trip over across the floor.

We've all met up in Cambridge - so there's an egw_, amacater, kibi who has travelled over to join us, Isy, RattusRattus and Sledge mostly sat round a table. The usual number of laptops, three monitors, Rattus' tower machine.Network running well and we're all ready to go, I think - there's normally a flurry of activity to get things started then a wait for a while for the first images
Coffee and tea at the ready - bacon sandwiches are on the way

[And the build process is under way - and smcv has joined us]

RattusRattus, Isy, smcv have all just left after a very long day. Steve is finishing up the final stages. The mayhem has quietened, the network cables are coiled, pretty much everything is tidied away. A new experience for two of us - I just hope it hasn't put them off too much.The IRC channels are quiet and we can put this one to bed after a good day's work well done.

And I'm over here with the Debian images/media release team in Cambridge.First time together in Cambridge for a long time: several of the usual suspects - RattusRattus, Sledge, Isy and myself. Also in the room are Kartik and egw - I think this is their first time.Chat is now physically in Sledge's sitting room as well as on IRC. The first couple of images are trickling in and tests are starting for Bullseye.
This is going to be a very long day - we've got full tests for Bullseye (Debian 11) and Buster (Debian 10) so double duty. This should be the last release for Buster since this has now passed to LTS.

We're flagging a bit now, I think but close to the end. The standard Debian images caused no problems: Sledge and I are just finishing up the last few live images to test now.Thanks, as ever, to the crew: RattusRattus and Isy, Sledge struggling through feeling awful. No debian-edu testing today, unfortunately, but that almost never breaks anyway.Everyone's getting geared up for Kosovo - you'll see the other three there with any luck - and you'd catch all of us at the BBQ in Cambridge. It's going to be a hugely busy month and a bit for Steve and the others. :)

A lower profile release today: Sledge working in the background as affected by COVID. RattusRattus and Isy doing sterling service on the other side of Cambridge, /me over here.Testing on the standard install media is pretty much done: Isy, Andy and Sledge have moved on to testing the live images.Stupidly hot for UK - it's 28 degrees indoors with windows open.All good so far :)

And we're working through quite nicely.
It's been a long, long day so far and we're about 1/2 way through :)
Shout out to Isy, Sledge and RattusRattus in Cambridge and also smcv.Two releases in a day is a whole bunch :)

We've more or less finished the release of the Debian CD/DVD/Blu-Ray and other media for Bullseye 11.2 release. This is one of the (roughly) quarterly point releases and roll-up releases.Thanks firstly to the developers, users, helpers, bug filers who help to keep Debian moving and working and to the release team and press team who more or less finish their job before the media team start theirs.
Thanks to Sledge and RattusRattus and Isy in Cambridge and to Schweer who single-handedly clears all the Debian-edu testing. This release run seemed to be a bit more slick, a bit faster, a few fewer problems and it's always a joy to work with colleagues who are like family.The last couple of years have been more than a bit difficult: this (should) be more or less the last major action from the release team and others for 2021.I'm going to take the opportunity this gives me to wish all concerned with Debian (and every other Linux) the very best wishes for the year to come. Surely, it can only get better - eventually.

I'm guessing the last glasses will be through the dishwasher (again) and Pepper the dog can settle down without having to cope with so many people.For those who don't know - Steve and his wife Jo (Sledge and Randombird) hold a barbeque in their garden every August Bank Holiday weekend [UK Bank Holiday on the last Monday in August]. The barbeque is not small - it's the dominating feature in the suburban garden, brick built, with a dedication stone, lights, electricity. The garden is small, generally made smaller by forty or so Debian friends and allies standing and sitting around. People are talking, arguing, hugging people they've not seen for (literal) years and putting the world to rights. This is Debian central point - with large quantities of meat and salads, an amount of beer/alcohol and "Cambridge gin" and general goodwill. This year was more than usually atmospheric because for some of us it was the first time with a large group of people in a while. Side conversations abound: for me it was learning something about the high energy particle physics community, how to precision build helicopters, fly quadcopters and precision 3D print anything, the maths of Isy counting crochet stitches to sew together randomly sized squares ... and, of course, obligatory things like how random is random and what's good enough entropy. And a few sessions of the game of our leader.
This is also a place for stuff to get done: I was unashamedly using this to upgrade the storage in my laptop while there were sensible engineers around. A corner of the table, a RattusRattus and it was quickly sorted - then a discussion around the internals of Thinkpads as he took his apart. Then getting a full install - Gb Ethernet to the Debian mirror in the cupboard six feet away is faster bandwidth than a jumbo jet full of tapes. Then getting mail to work again - it's handy when the mailserver owner is next to you, having come in from the garden to help, and finally IRC. And not just me: "You need a GPG key signed - there's three DPLs here, there's a release manager - but you've just missed one of the DAMs." plus an in-depth GPG how-to session on the other side of the table.I was the luckiest one with the most comfortable bed in the house overnight but I couldn't stay for last night. Thanks once again to all involved but especially Steve and Jo who do this for the love of it, and the fun, and the community and the family. Oh, and thanks to Lenovo - not just for being a platinum sponsor of Debconf but also for providing the official laptop of this and most Debian occasions

And that's it for another release. A few bugs but nothing show-stopping.Thanks again to Sledge, RattusRattus and Isy, to Linux-Fan and schweer.As the testing page notes, there's a bug in some arm64 installs - the fix should come out via debian-security shortly but you might want to be aware of this. Here's to the next one: only a short time until Bullseye

Late blogging on this one.Even as we wait for the final release of Bullseye [Debian 11], we're still producing updates for Debian 10 [Buster].Today has thrown up a few problems: working with Steve, RattusRattus and Isy in Cambridge, Schweer and Linux-Fan somewhere else in the world.A couple of build problems have meant that we've started later than we otherwise might have been and a couple of image runs have had to be redone. We're there now and happily running tests.As ever, it's good to be doing this. With practice, I can now repeat mistakes with 100% reliability and in shorter time :)More updates later.

And so I'm rather late: Sledge, RattusRattus and Isy and I have all been busy chasing down bits and pieces for images release today. We're well underway through the normal images: live images are still ongoing. We've had help from Schweer and Linux-fan as last time - a couple of other folk have had chats to us. So far, it seems relatively straightforward - we're all experienced at this by now.As ever, it's a joy to be doing this with a smart set of colleagues. We can now catch our mistakes _just_ after we make them, so it's really getting sorted: the next stage is for me to realise _before_ I make them.

Steve is about to sign the release and to begin the push across to the mirrors.Although it sometimes seems like an age, this will be approximately 8 hours rather than 15 hours - a 50% improvement in release timetable means that the behind the scenes work has been well worth while.We always find tweaks, improvements, things we forgot and genuine bugs. Once again: We've found that live images can be significantly memory intensive on older hardware or machines with limited memory. The boot process for any live CD image expands a squashfs so that the image runs entirely in memory. This is particularly noticeable on the 32 bit i386 images. These really require a minimum of 2GB of memory if you are using the heavier weight desktop images like KDE or Gnome: much less and they will either be unreasonably slow or just fail to work.There is also now a note on the download pages warning on this.Thanks once again to maswan early on and Sledge, RattusRattus, Isy, Schweer, linux-fan and sqr not for helping out in testing images. I'm listening in to an impromptu chat explaining the behind the scenes steps to run to actually sign and push the final images so we're a short while away from publishing to the principal CD image machine. At the same time, the torrent seeders will be started and the scripts will push to update the mirrors. And so to the next time :)Handy tip For people running mirrors of debian-cd: If you're low on disk space, perhaps you could remove the 10.7 images by hand before running your next sync scripts to allow space for 10.8 to move in: any ftpsync or rsync process might only remove the old images after copying in the new ones. It's only about 220GB - but you don't want that to be an instantaneous 440GB.

Thanks to the changes behind the scenes, images are now being produced significantly faster than they were: the embarrassingly parallel speed up has worked, though at slight cost to the predictability of when we get each architecture produced for us to test.Thanks, as ever, to Sledge, RattusRattus and Isy over in Cambridge and to schweer who painstakingly tests all the debian-edu releases.83 out of 94 steps done and it's only 13:50 GMT - less popular architectures are coming on apace. There's still a lot of testing to do: we will never be able to test s390x for example. Other architectures - including mips* - we have no machines, we're building as best efforts and we can't guarantee how well they will work for anyone.Anybody installing the less popular architectures please let us know that you've installed them and how you get on. We'd like to see a positive report as well as bugs: we don't see much feedback at all to know how useful they are: evidence of installs on any architecture is always helpful to us.

And we're through the live images as far as we can reasonably go. There remain a couple of MIPS architectures to build and IBM s390x - we have no hardware to test these.It's now 23:30 UTC - so that's 12 1/2 hours since we started and Sledge has another couple of hours to wait before the last images can be signed and the image release can be pushed.Thanks so much to Sledge, RattusRattus, Isy, Schweer and jlsantos - who came back to us after a significant internet outage local to them in California - to those who popped by - hi Kibi and Manty - and to all who labour behind the scenes to make each release "just work"Here's to the next one - in another couple of months, I'm guessing.