I have, predictably, forgotten to look at the agenda like I meant to earlier this week. Our agenda for tomorrow is empty so consider this a bit-too-late "last call" for agenda items.
If there's nothing on the agenda in the morning (or no PRs) I'll cancel this iteration of the meeting to meet again in 2 weeks.
We've got an agenda now so see you all tomorrow!
I have, predictably, forgotten to remind here more than >1 day in advance. I have, however, remembered more than 10 minutes in advance!
As a reminder if there's still an empty agenda tomorrow I'll cancel the meeting
Oh dear and now here I am forgetting to cancel the meeting as well, apologies for not being a great meeting admin :(
its fine we shared a couple of real brief updates and went on our way
Gonna try to do better this week!
Wasm I/O is this week as well so I imagine lots of folks are traveling. Currently there's an empty agenda and if it's still empty tomorrow night I'll cancel Wednesday's edition.
It's not last night, but better late than never, I've cancelled the iteration for today, see y'all in two weeks!
No agenda items for tomorrow so far, so I'll cancel the meeting tomorrow morning if there are no objections (and if I remember to)
I just added one: https://github.com/bytecodealliance/meetings/pull/259
:+1: see y'all soon!
We have no agenda items today so I'm going to cancel today's iteration
No objection to canceling the meeting, but I wanted to give a quick update here on async progress: I've got a simple end-to-end demo working using temporary forks of wit-bindgen
, wasm-tools
, and wasmtime
: https://github.com/dicej/component-async-demo. Planning to work on composition next.
already run, already works, wanna screw around with it MOAR
ubuntu 22.04 on wsl
Joel Dice said:
This worked for me too, as far as you expected. Walking through the code is very informative (no surprise).
One question. When the guest is composed with itself in your example (let me show this as host(guest1(guest2))), when guest2 calls to its imported foo
, is it meant to be reaching directly out to the host's implementation or is it mean to reach out to the guest1's implementation? That part of composition I'm not clear on yet.
I hope I haven't misunderstood what's going on to the point of muddling up your very nice demonstration! :pray:
wasm-compose
connects guest2's import to guest1's export, so the host's implementation isn't called until guest1 calls its import. So at runtime it looks like: host -> guest2 -> guest1 -> host. In the synchronous case, the host would then return a value and the stack would unwind. In the async case, things get more interesting if the host await
s, in which case tasks get suspended and control moves between the guests and hosts as tasks complete. See Luke's recent WasmIO presentation for a visual representation of that process.
Was late this week to see if folks had last-minute items, but there's currently no agenda for the meeting in a few hours so I'm going to cancel today's iteration as well
My status update: I've implemented fused adapters (i.e. the things that bind composed components together in Wasmtime) for all possible combinations: async->async, async->sync, and sync->async (in addition to the already-supported sync->sync). Next week, I'm planning to work on stream
and future
support in wit-bindgen
, wasm-tools
, and wasmtime
, with the goal of porting over the wasi:http@0.3.0-draft
tests from isyswasfa
. This week, I'm adding resource support to the C# binding generator for wit-bindgen
.
Quick update on my component model async implementation work: future
s and stream
s are now supported (see here and here for examples I've ported over from isyswasfa
). That rounds out the major features of CM async (at least for the callback-based ABI).
Now that all the major pieces are working (modulo bugs), I'll start opening PRs on the respective repos. We'll keep everything behind feature flags while we add tests, reconcile with the still-evolving specification, etc.
As a heads up I've moved tomorrow's meeting to next week on June 26 to accommodate how many US-based folks will have the day off tomorrow
Looks like we have nothing on the agenda for today and I forgot to check earlier in the week. Unless there are any last-minute items folks would want to add I'll cancel today's iteration of the meeting
Ok I'm going to go ahead and cancel, see y'all in 2 weeks!
I was too late to speak up for today, but I would like to receive some feed-back on my PoC for the symmetric ABI (variant).
Sure yeah! Want to add it to the next agenda?
As a reminder for folks we've got two agenda items for tomorrow and if anyone else would like to add something feel free to send a PR and/or bring it up during the meeting. See y'all tomorrow!
I've cancelled tomorrow's meeting due to an empty agenda
I've also cancelled today's meeting for no agenda
Just a note for those who might be watching this thread -- today's meeting was also cancelled since there was no agenda
I'm cancelling tomorrow's meeting in favor of the BA summit on Thurs/Fri
I'm cancelling tomorrow's meeting due to an empty agenda
Cancelling tomorrow's meeting due to an empty agenda too
Last updated: Nov 22 2024 at 16:03 UTC