# 21th April 2022, 6:00 - 7:00 am UTC+0
Attendees: Thomas, David, Edward, Nico, Anastasia, Nikolai
## Decisions Summary
* Anastasia will be doing next release. She has never done release before, and needs help from more experienced people.
* [FelixS] Redmine is ready to use * If you wonder, new registrations require an unlock by an administrator. This is to prevent spamming and bots. Maybe we can work out something better in the future. * I added additional fields to issues, so that important information is easily visible. For example affected hardware, affected OS, affected version. Please tell me if you find these useful or if you need other fields, trackers, .. * aklm: let’s start using, I will try to create an account and see how it goes * [FelixS] Created a proposal for closing GitHub issues and pull requests automatically * Closes newly created issues and pull requests and adds a comment using GitHub Actions hook. Old issues and pull requests stay untouched. * [flashrom patch](https://review.coreboot.org/c/flashrom/+/63671), based on GitHub Actions hook and [repo- lockdown](https://github.com/dessant/repo-lockdown). * My playground [https://github.com/felixsinger/awesome-repo%5D(https://github.com/felixsinge...) * [Nico] ChromeOS fork and attempted convergence * (in case witnesses of that time are present) What led to the fork? * [PatrickG] Velocity requirements: stuff needed to happen to make progress on the products using flashrom and it didn’t happen in time. Not much of a problem when it’s temporary but stuff just piled up over time. * Nico: lots of things went wrong. And it took around 8 months to converge * David: Fork happened when Carl-Daniel was the maintainer, and he was really busy. Tools were worse that time. * We had irreconcilable differences at that time * Edward: Stefan asked me to unfork flashrom. I had to work out how things were done, how to re-write parts of the system. It was just me, on my own, no other resources. I took upstream and tried to make chrome os work with upstream. Except for a few patches: mapping hack inside of sb600spi, I am trying to chase AMD to get documentation , so that I can improve it. * David: Timer stuff I was trying to deal locally. There were a bunch of things that had to be dealt locally. * David: It was uglier downstream than it should have been. Trying to minimize the number of hacks. There are things that can be tr * Edward: Where we are today: I was trying to have a team for flashrom, to spend dedicated time to work on flashrom. We had good results on that, but it takes time. Hard to allocate people. * (in case somebody knows) What is Google’s current strategy to avoid repeating mistakes of the past? * [PatrickG] Generally CrOS is willing to go upstream-first (as with coreboot) after divergence has been resolved. Depends on the upstream, so probably need to work on the questions raised by Edward et al. * [Nico] Upstream-first of a dominant vendor made it harder for others to work upstream on coreboot. It’s not what I had in mind, but this would be repeating a mistake, IMHO. * Nico: There were very good and very bad contributions during last years * Edward: how do you quantify that nothing is happening? * Nico : sb600 is bricking machines , but nothing is happening * Edward: maybe it is a visibility problem, also some things are really hard * Nico: maybe devs in google need to talk to each other? * Edward: original patch came from AMD, not from google * Nico: if it takes a long time to find a solution, maybe just revert? * Edward: we didn’t know it would take so long * David: board bringup. There is 3-4 months time before chromebooks get into users hands, and then bugs get filed. Bugs in the code which is 6 months old, but people are already reassigned to other tasks. * Edward: maybe have a parameter which turns on the behaviour, default is turned off. The problem exists, but it is not trivial. * Nico: The problem with the “upstream-first” idea: when something is not trivial, you push it upstream first, and then try to figure out how to fix it. * Edward: that was exceptional situation, it is better now * Nico: most of the issues in the list are from google. When someone looks at the code, it is clear that code it is not understood. * Edward: we opened issues about non-documented programmer, and added documentation * Nico: one of the drivers is mixing all layers of flashrom. * Nikolai: what are concerns with the code? We will try to remove them. * Edward: The datasheet was in mandarin, and lots of questions were discussed in mandarin with the vendor. Vendors were not giving documentation, despite us asking them many times. * Nico: Who is doing reviews for google? We need reviewers who know all layers of flashrom. * David: Can we simplify the process of adding a new chipset? So that it would be easier to contribute without introducing bugs? * Thomas: It would be nice to have more structure in the code repository. It is hard to understand where the code needs to go. Having a specific place for every component is a good idea * David : flash.h and flashrom.c are dumping grounds * Anastasia: Nico do you think it is possible so that we all work together normally and peacefully? If we resolve issues from the past, and also write guidelines so that new issues won’t appear and development goes normally? >> Yes it is possible. * Let's make a release! * Nico: I spent a lot of time on this already, I can’t spend even more time. Anyone else? * Nikolai: What if the vendor does not give documentation, do you think it’s better to have a driver which is not documented, or no driver at all? * David: for early chromebooks , drivers were not documented. That’s why an option was called “I want brick” * Edward: some of the changes in the chromium tree we are not sending upstream, because there is no documentation * Nikolai: Can we have a flag for situations like sb600spi ? Finer granularity than the whole driver. Feature flags? * Nico: other people work on the sb600spi code, affects everyone who works on the code. * Edward: something similar in kernel hardware quark framework. * Nico: this needs documentation, whatever is needed to understand the quark. Man pages are for the users, but there also should be documentation for developers. In what product is this chip? You can document what you know * [dhendrix] Is this the kind of finger pointing that Edward mentioned above? At this point, CrOS has proved more than happy to move upstream. The Github mirror is probably more of a concern. Perhaps upstream should consider how to get developers to choose upstream rather than the Github mirror or some other fork. * [Nico] Pointing at Google, actually. If one company forks and later wants their code upstream without proper review, something seems wrong. As if they never reflected on the problems. * [aklm] I am not sure we are on the same page on what are “mistakes of the past”. Everyone has their own idea on what the mistakes were. I can start from myself. From my observations, converging the code was considered a purely technical task, and so it was done as a technical task. However, converging the code means converging the people/the teams, converging very different workflows and work cultures. This does not seem to be thought through or ever discussed, and to me this is a mistake. * [quasisec] I am also unclear how CrOS convergence had really much of an impact on upstream considering >99% of the work was adjusting the CrOS fork to realign with upstream. Only an extremely small amount of alignment was in the form of upstream commits, e.g., missing raiden_debug and some unfortunate part of sb600spi. Is there a specific area that stands out? * [Nico] No idea if related to the original reasons for the fork, but that the convergence is stalling new releases for years stands out pretty much. * Anastasia: We currently have more GSOC project proposals than mentors (5 vs 3). If someone wants to be a mentor, it is not too late. Contact Anastasia or Felix.