When so many people involved with postmarketOS and Linux Mobile come together in one place, magic happens. Without even trying to solve problems, we end up finding important solutions to long-term issues as we discuss running mainline Linux on all sorts of devices while walking around in ULB or elsewhere in Brussels, going by public transport or while eating in restaurants. Thanks to everybody who came by, it was amazing!
FOSDEM 2025
We had a separate postmarketOS stand with our Linux on Mobile friends (Mobian, Sailfish OS, AsteroidOS, LINMOB.net, MNT) on one side and our friends from CalyxOS on the other. Usually we team up with the rest of the Linux on Mobile crowd, but given that in 2024 we almost ran into serious space issues, we figured it would be safer to request our own stand for 2025. And it was a blast, thanks to everybody who came by and asked us questions and tried out our (sometimes quite crazy) setups. It was especially nice seeing true tiling WM lovers experiencing Sxmo for the first time and feeling right at home!
Federico had printed out 150 copies of the leaflet (which were all taken by visitors — except for the last two which people apparently didn't want to take since these were the last ones). He also printed the entire devices wiki page in DIN A3 format, which proved to be very useful: people ask us if we support a specific device all the time, so every few minutes we could whip out the entire print out and look for that specific device together with the visitor to see if it is in there or not. Regardless, we always got the same reaction: sooo many devices!
On Saturday afternoon the FOSS on Mobile Devices devroom took place, with many postmarketOS related talks such as Kernel support for Mobile Linux: The missing 20% by Luca, Sxmo: A mobile UI for hackers by Maarten, postmarketOS: what is it and what's new? by Oliver and many more! For most of them, the video recordings are already up by now. Anjan also gave a talk called Introduction to pmbootstrap in the Embedded, Mobile and Automotive devroom. See also the list of related talks in other tracks at LINMOB.net.
On Sunday we recorded a new episode of the postmarketOS podcast. This episode is being released together with the blog post here and has a short collection of stories related to what people experienced this FOSDEM.
Hackathon
As our team becomes bigger, so does our post-FOSDEM hackathon! This year we rented an apartment for 12 people where mostly a mixture of Core & Trusted Contributors got together. We had several rooms with different discussions happening simultaneously for most of the week, which was a great success! It is going to become harder to scale this up if the team keeps growing.
As last year, we also set a wall full of post-its to organize things to hack and discuss together. You can see the photo of completed tasks as part of the header image, and read more on what these were about below.
Reliability
Through many discussions we have concluded that our main goal for 2025 is: Improve the reliability of postmarketOS!
One major part of getting there is automated hardware testing. Once we have that in place, we can continuously verify whether features are still working on devices we test. Regressions will be caught early, ideally before we even merge changes, and we should be able to fix them much more quickly. This should help us get rid of the days on which you upgrade your edge install and have no way of knowing whether audio will still work after the upgrade, for example.
We followed up on the hardware CI project that was started in 2024-09. There we originally intended to pay a developer to design a PCB for connecting the phones in our future automated hardware testing setup. Unfortunately they didn't have time and so we didn't make much progress on that topic up until the hackathon. However having enough skilled people in one place worked magic and we figured that our TCs Federico and Anjan could take this over with support from Caleb. We are happy to report that they have already started and this is now back on track!
The result will of course be under an open license, and we want to make it so that others in the community can get the PCB too and use it to add hardware CI for their favorite device as well.
In addition, we have started preparing for the software work that will be necessary to actually do testing on real hardware. This part of the project involves all the infrastructure that allows the GitLab CI to communicate with the devices, send it the jobs to execute, and receive the artifacts. For this part of the project, we intend to use CI-tron, and will be working with both its maintainer as well as an experienced member of the community to implement the missing pieces and integrate it with our GitLab. This will be funded entirely through our OpenCollective! Once the agreement is in place, we will provide a detailed update on the architecture and our plan for implementing it. Stay tuned!
As a follow-up from this project, and with the idea to improve reliability, Anjan, Federico, together with Dylan planned on how to possibly test calls on CI. They made some progress planning, see #96!
Camera
While they were not at the hackathon, DrGit surprisingly sent patches that made one of the rear cameras and the front camera on the OnePlus 6 and OnePlus 6T work for the first time! Caleb prepared !6148 based on that, and made a call for testing and soon lots of people posted initial photos and even videos of the camera in action in the same Mastodon thread. This feature has been requested for such a long time, it is amazing to see this moving forward!
Caleb and David looked into getting the not-yet enabled C-PHY mode working.
Minecrell worked on the BQ Aquaris X5 camera on top of the work by André on the BQ Aquaris M5.
Device ports
- Luca proposed to clearly differentiate devices in the testing category between ports based on mainline and downstream kernels. (All higher device categories, community and main, already require mainline kernels). This was discussed and turned into actionable items, see milestone 30 for details. Contributions welcome!
systemd
Clayton, together with external support from Robert finished enabling sensor support on systemd (!6147).
Clayton, Pablo, Jane, Oliver, Caleb discussed how to finish upstreaming the abuild systemd split function (#2804). We discussed different alternatives and decided that it is probably best to follow up with abuild maintainers regarding what solution they would prefer the most.
Oliver and Jane discussed how to make Phosh work again with postmarketOS edge and systemd since it was currently broken there. We decided to replace the current, not working
phosh.serviceimplementation with tinydm (the same way it currently works with OpenRC). Achill had then implemented this change in !6150. The plan is to later on use greetd with phrog instead of tinydm, and to use that for both systemd and OpenRC. Joel and Sam are working on this in !6106.
Organizational
Pablo, Oliver, Stefan, Luca, Bart managed to finish up the Request for Change process we now call postmarketOS Change Request (PMCR). With that we finally have a place to discuss bigger changes to the project with a defined structure. This led to the second PMCR being sent within hours! We also had a fun time trying to figure out an original name of the process (so it is immediately clear that people are not just referring to any RFC, but a postmarketOS specific RFC). We considered PMEP at one point (postmarketOS Enhancement Proposal) but since that is impossible to pronounce if you try to say it like PEP from Python, we went with PMCR :)
During our first hackathon after FOSDEM 2024 we set some projects as priorities for the postmarketOS team that year. Now one year later we have reviewed our progress and improved on how we approach these kinds of mid-term milestones. We will soon provide an update on this blog with further details (either as separate blog post or as part of the upcoming monthly blog post)!
In addition to the mid-term project priorities, we also started discussing a possible 5 year strategic plan. postmarketOS is in a really amazing position with a healthy and productive community. We want to ensure this continues to be the case as the project grows, but we also want to continue pushing the envelope and see where it leads. We have some ideas, and did some initial discussions at the hackathon. However properly answering this will likely require further discussions and planning with the rest of the team and the community, so we can't write much about this yet.
Infrastructure
Pablo added support in aports-qa-bot for pinging maintainers when somebody proposes changes to packages they maintain. This should make it easier to maintain packages. Previously maintainers wouldn't always be notified when changes to their packages were made — even though they were in CODEOWNERS, but notifying through that is a GitLab premium feature that we don't have on our self-hosted instance anymore. The bot is already live (deployed by Luca) and commenting away!
Caleb, Oliver and Clayton tweaked GitLab CI configs and eventually created a custom GitLab runner to fix that QEMU-related jobs didn't run reliably since we had switched to gitlab.postmarketos.org.
Caleb, Pablo and Oliver worked towards integrating marge-bot into our development process. This will allow us to automate manual steps that we currently do before merging (rebase, put MR ID into commit messages, wait for CI), and therefore accelerate our development process. It also takes care of merging ready MRs in order, avoiding conflicts we can currently run into when two developers try to merge MRs at the same time. We talked to David from upstream who patiently answered all of our questions and found a way forward for all features that we need before we can use marge-bot (userfriendly messages (implemented!), and commit signing (TBD)).
Bart and Luca did an upgrade of all our matrix rooms, in order to fix state resets that a couple of people have been getting as well as making permissions consistent across all rooms.
BPO, our build orchestrator got a security audit from Radically Open Security as part of the postmarketOS daemons grant from NGI Zero Core / NLnet. The audit was successfully kicked off during the hackathon (from our end Oliver explained how to run tests locally, answered follow-up questions etc). While we still need to get the final report, it looks like there were no remarkable security concerns.
Other
Jane spent some time looking into a compatible solution for GNOME Clocks to set alarms that will work even on suspend.
We discussed at length how to do taxes for invoices to OpenCollective. Jane has done some initial systemd upstreaming work that we still need to figure out the VAT situation for. We have fortunately gotten some support from KDE e.V. and we will be contacting some tax lawyer in the short term. This work is not very enjoyable, but very important.
Oliver and Clayton finished a MR that has been open for a long time, which moves the contribute page from the wiki to our website and modernizes it. Thanks Ranny for the initial work on it, and for adding all these cute postmarketOS logo variations! Together with Pablo we also made progress moving some important governance information from different scattered places to the website (!335).
We discussed how to better organize ourselves and go to conferences. We should start keeping an inventory for merchandising and marketing materials, but also plan and better support people from the community going to them. We want postmarketOS to have visibility in many more places, and not just very technical conferences. But we need to make sure those spreading the word have the support they deserve! In the following weeks we will be creating a repository to better organize and plan this effort.
For hackathon movie night, we chose For The Plasma (2014) this time (see our reviews) to balance out the GNOME themed movie The GNOME Mobile (1967) from last year. Let us know if you have similar movie suggestions for next year, maybe something that relates to SXMO or tiling window managers! For the last night, Bart had a Wii emulator set up where some of us played a few rounds of Mario Party.
Cooking and shopping for 12 people is a challenge. Thanks to Luca and Federico who cooked some delicious meals that kept all hackers going!
Although this is not a short blog post, this doesn't cover everything that was going on. We didn't make sticky notes for each topic we worked on and we had many, many important conversations that would be hard to fit in here.
Having this hackathon was only possible thanks to your donations! Just like last year, it was incredibly productive to have everybody in the same place, for planning, discussing goals and for just getting things done quickly without delay.
If you appreciate the work we're doing with postmarketOS and want to support us, consider contributing financially via OpenCollective.