From d5ed60cc91521cd1fbced79286358c2769f7f31e Mon Sep 17 00:00:00 2001 From: Hudson Ayers Date: Fri, 3 Jan 2025 09:49:33 -0800 Subject: [PATCH] add core wg notes from 2025-01-03 --- doc/wg/core/notes/core-notes-2025-01-03.md | 112 +++++++++++++++++++++ 1 file changed, 112 insertions(+) create mode 100644 doc/wg/core/notes/core-notes-2025-01-03.md diff --git a/doc/wg/core/notes/core-notes-2025-01-03.md b/doc/wg/core/notes/core-notes-2025-01-03.md new file mode 100644 index 0000000000..8cc9901e57 --- /dev/null +++ b/doc/wg/core/notes/core-notes-2025-01-03.md @@ -0,0 +1,112 @@ +# Tock Meeting Notes 2025-01-03 + +## Attendees + +- Hudson Ayers +- Amit Levy +- Leon Schuermann +- Johnathan van Why +- Brad Campbell +- Tyler Potyondy +- Kat Fox + +## Updates + * None + +## Tock 2.2 Release Plans +* Leon: We tagged 2.2-rc1 a couple weeks ago +* Leon: It is clear we are going to go forward with less testing of all the + boards we have in tree because it has been discouraging us from making + releases, and we want more frequent releases +* Leon: Instead, we are mostly relying on the automated testing available via + the hardware CI infrastructure. The supported boards at least cover all of + the chips that are used across all of our their 1 boards. I know folks felt + uneasy about this not including any RISC-V testing, so I have also been + trying to run a full suite of Litex testing. It seems that we already have + pretty complete testing in CI for the LiteX emulated platform at least. +* Leon: One major issue is the status of the 15.4 subsystem — I have not really + gotten it to work, so would like some info from Tyler before stamping the + current state of things as ok for release. +* Leon: We are almost 2 years from our last release +* Amit: I recall that 15.4 required some magic incantations for stuff to work, + has anyone done anything else? +* Leon: I know Tyler submitted https://github.com/tock/libtock-c/pull/485 which + fixes one bug we have seen +* Amit: I think the main issue is we have multiple 15.4 stacks now, one with + Thread in user space and one with ipv6 in the kernel, and I do not know what + is enabled by default and how that affects the example applications. +* Leon: I did test openThread and it worked on the nrf boards. +* Brad: There is a shared subset of 15.4 that is exposed to user space, and + then the OT driver exposes the same set of things plus some additional stuff +* Leon: Tyler, do you know if UDP works with your libtock-c PR? +* Tyler: It should, but I have not touched this since Spring of 2023. The + problem is the standard 15.4 driver that includes MAC/framing has issues + because the new 15.4 driver for the radio defaults to being off, but we do + not have a way to expose turning on the radio to the MAC layer driver used by + the UDP stack. Yesterday I was working on a fix to add that back and writing + some treadmill tests for it. The main problem was 15.4 could not transmit on + the base stack so the UDP tests were failing. +* Leon: What I am hearing is we should probably wait for end of today and see + if we can get it working, and then release regardless with a note in the + release notes if needed. +* Tyler: Sounds good to me +* Tyler: Where should I document the stuff that is working now that I have + tested? +* Leon: The only tests are the UDP ones right? +* Hudson: There are also some raw 15.4 user space tests +* Leon: Can we summarize all this in a comment? +* Tyler: Sure, we can sync up offline on this +* Leon: I just sent a PR that allows us to trigger a treadmill workflow on an + arbitrary revision. We can’t really test this until it’s merged but it + already is approved, I think we should just merge and then test it. +* Leon: It is very easy to make mistakes in these GitHub actions PR, but if it + does not work I will send a fixup PR +* Leon: We also have https://github.com/tock/tock/pull/4275 to specify a + minimum stable rustc version. +* Brad: This PR originated because I failed to build hail when wanting to test + it for this release. The issue was my local stable toolchain was too old. And + it seems there is no way to just specify a minimum version in cargo without + forcing a specific version, and the error messages are confusing. +* Brad: This PR adds this infrastructure to our make build system instead, it + automatically updates your toolchain after giving you a warning. +* Leon: The alternative suggested by Johnathan is to specify an exact version + in rust-toolchain.toml so that we always have users build with the exact + version. This means more downloaded versions but has some upsides described + in the PR. +* Leon: To me, it feels like a step backwards to force users to use an older + compiler version when newer ones work. +* Johnathan: Clippy will disable lints if your MSRV does not support certain + features, it would be nice for us to take advantage of that. We might + eventually run into something similar with compiler warnings, since we treat + warnings as errors. +* Hudson: We only treat warnings as errors in CI, so this would not actually + break user builds who are using a newer compiler. +* Johnathan: Lints affect everything in dependency trees but this seems like + not an issue since we only treat warnings as errors in CI. +* Brad: I think it is important that we accept any stable version, I agree with + Leon. That is kind of the point of Hail using stable, as an example. +* Leon: Next PR that needs to be merged is + https://github.com/tock/tock/pull/4280, this targets the release branch, I + will also add one to backport it to master. It adds the version bump and + release notes. I have another version of this in a hackMD that has more + detail (https://hackmd.io/TNqklrfYQlqFSJLI5VSYYg ), but I will just use that + in the release notes and keep the compressed version in the changelog in the + repo. If there are issues with UDP, they will get added to this PR before + Leon merges it (as well as to the long version). +* Leon: Last thing before release. Our risc-v tests are mostly just the LiteX + CI tests, though I will run them on a board as well. I have a dummy PR that + bumps the libtock-c version pinned for the litex-ci so I can get the CI tests + to run against the latest libtock-c: https://github.com/tock/tock/pull/4285 . +* Hudson: Tyler do you have PRs up for the 15.4 changes? +* Tyler: Not yet, I am trying to do some testing because I do not have the + physical boards with me. They should be up in the next few hours. +* Leon: I will wait until tomorrow so those should have a chance to go in. +* Hudson: Tyler can you post on slack when those go up? +* Tyler: Sure, and this is the fix we talked about in the network wg in early + December, to add a syscall for turning on the radio to that stack. +* Brad: So Hail does not work, but I think that is because of user space. Amit, + did you get a chance to look at that? +* Amit: Let me do that right now +* Brad: Do you think that it is solvable? +* Amit: yes +