[U-Boot] February community conf call minutes

Hey all,
While I'm not the best at taking notes, and for next time I'll see how badly the record function slows down my connection, here's my notes / action items from the call:
- Attendees: Myself, Philipp Tomsich, Michal Simek, Stefano Babic, Vignesh R, Jagan Teki, Lokesh Vutla, Matthias Brugger, Masahiro Yamada.
- Talked about more automated testing, ala kernelci. I noted that we have travis-ci setup, and with test.py that runs on on qemu on a number of platforms. With respect to real hardware, I run mine on a handful of TI platforms and RPi 3 (I've tried sunxi before, but had power vampire problems, and imx6, but my platform wasn't supported in SPL at the time). Michal noted he has a setup for some of his platforms. - A main next branch for testing / integration? I don't know if I have resources to maintain one, but can evaluate again with this longer release cycle. I do continue to encourage custodians to have a next branch if that helps them. - Communication. A lot of people didn't see the notice about this meeting until very late. A lot of people in general have missed a lot of things (see for example, the fallout from the DM deadlines). It has been suggested that we have a custodians and/or board maintainers only list to better help people filter emails and see very important things. - I've asked Wolfgang to set these up, and I'll be asking people to sign up. - As part of the "didn't see this until very late", I need to poll again about when to do these calls. - Some talk in general about the DM deadlines and what to do about missed deadlines. In general, and looking at immediately post-v2019.04, I am not in favor of just dropping every platform that hasn't converted. I think in some cases it will make sense to mark drivers as depending on BROKEN so that we can then remove the driver in question. I also agreed that if platforms don't see the value in migration for legacy platforms, it's OK for them to also say they don't see value in being upstream for that platform anymore. And removing a feature that wasn't really used (but was instead carried over from the "kitchen sink evm" config) is fine too. - One thing I do need to do from this is make a build (or several) where, after I apply all of the "migrate to ..." things in my queue I see what's still where with these warnings. We have things ranging from for example TI platforms that we've converted some of the family over to, but not all, and should be able to mechanically convert, to PowerPC platforms where I've seen nothing happen in a while. For the latter we'll end up with BROKEN and for the former after I can test one I might blind-convert the rest.
Thanks again all!

Hello Tom,
Am 19.02.2019 um 00:53 schrieb Tom Rini:
Hey all,
While I'm not the best at taking notes, and for next time I'll see how badly the record function slows down my connection, here's my notes / action items from the call:
Attendees: Myself, Philipp Tomsich, Michal Simek, Stefano Babic, Vignesh R, Jagan Teki, Lokesh Vutla, Matthias Brugger, Masahiro Yamada.
Talked about more automated testing, ala kernelci. I noted that we have travis-ci setup, and with test.py that runs on on qemu on a number of platforms. With respect to real hardware, I run mine on a handful of TI platforms and RPi 3 (I've tried sunxi before, but had power vampire problems, and imx6, but my platform wasn't supported in SPL at the time). Michal noted he has a setup for some of his platforms.
I also have/had a running testsetup with tbot, see: http://xeidos.ddns.net/tests/test_db_auslesen.php#987
But as you see, dead for newer tests, as I had no time to fixup the at91 boards for example, for which wdt is broken. Yeah, testing *is* a time consuming task, as you always find something, which does not work.
I just can vote for using tbot, added Harald Seiler to cc, as he rewrote tbot completly in python 3.6, see [1] and [2], feel free to take a look at it. I really like the interactive testcases for U-Boot and linux ... you start tbot and you are on the boards U-Boot or linux commandline ...
I start for example test/py with tbot, or on the U-Boot commandline the "ut all" command if enabled (or enable the ut command with tbot before compiling U-Boot). You do not need to have the boards you want to test near you, it could be around the world (in my case I run tbot and the above webpage on a raspberry pi in hungary, while the boards I test are in munich/germany)
The "problem" with tbot is, that it activly opens a ssh connection to a so called Lab Host see [3] and a lot of customers do not allow to use this for automated testing purposes ...
But I talked with Kevin Hillman some months ago, and he mentioned that it should be possible to run kernelci without LAVA, so may it is worth to look into kernelci and use it for "ubootci" purposes. Than we would have an API for reporting testresults, and can start for example tbot locally without the need of an ssh connection. It should be easy than to generate a testreport for kernelci with a tbot generator, see [4] for generator examples. I am sorry that I did not found time yet, for more investigations...
And customers are more willing to send testresults to us instead to allow ssh access for testing... hopefully.
bye, Heiko
[1] https://github.com/Rahix/tbot [2] https://rahix.de/tbot/getting-started.html [3] https://rahix.de/tbot/index.html [4] https://github.com/Rahix/tbot/tree/master/generators
participants (2)
-
Heiko Schocher
-
Tom Rini