Stream: wasmtime

Topic: towards wasmtime 0.23.0


view this post on Zulip Dan Gohman (Feb 16 2021 at 20:45):

Hello all, I'm hoping to release wasmtime 0.23.0 today; if anyone has any special mentions for the release notes, let me know!

view this post on Zulip Dan Gohman (Feb 16 2021 at 21:00):

Notes are up: https://github.com/bytecodealliance/wasmtime/pull/2656

This switches from term to termcolor, switches from memmap to memmap2, and updates to rand_core 0.6.2.

view this post on Zulip Dan Gohman (Feb 16 2021 at 21:00):

Lots of cool stuff in this release :smile:

view this post on Zulip Dan Gohman (Feb 16 2021 at 22:52):

@Andrew Brown Do you know what this error is?

  --- stderr
  CMake Error: The source directory "/home/sunfish/.cargo/registry/src/github.com-1ecc6299db9ec823/openvino-sys-0.1.8/upstream" does not appear to contain CMakeLists.txt.

view this post on Zulip Andrew Brown (Feb 16 2021 at 23:10):

yeah, that means that someone is trying to compile the openvino-sys crate from the OpenVINO sources (upstream) but the sources aren't there

view this post on Zulip Andrew Brown (Feb 16 2021 at 23:10):

why would we build that crate?

view this post on Zulip Andrew Brown (Feb 16 2021 at 23:11):

does it happen with a cargo build --all-features or something like that?

view this post on Zulip Andrew Brown (Feb 16 2021 at 23:13):

in the CI wasi-nn is usually a disabled feature but if you want to build it with that feature and avoid the error you should be able to use OPENVINO_SKIP_LINKING=1

Standalone JIT-style runtime for WebAssembly, using Cranelift - bytecodealliance/wasmtime
Standalone JIT-style runtime for WebAssembly, using Cranelift - bytecodealliance/wasmtime

view this post on Zulip Andrew Brown (Feb 16 2021 at 23:33):

But that does not seem to be the main issue at least for CI... glancing at the CI for the 0.23.0 PR, something is wrong when attempting to compile wasi-nn from the updated version of the wasi-nn spec: https://github.com/bytecodealliance/wasmtime/pull/2657/checks?check_run_id=1914400426#step:6:339. Do we have to update the spec commit in this PR? (It's good to have figured out there's an issue there but to not block the release we should just be able to use the original commit, right?)

This updates a few .gitmodules entries and depedency versions in preparation for releasing 0.23.0.

view this post on Zulip Dan Gohman (Feb 17 2021 at 00:10):

I'm getting that error when I just do cargo check on wasi-nn

view this post on Zulip Dan Gohman (Feb 17 2021 at 00:10):

And yeah, #2657 is getting different errors, and I'm trying to work on them, but the openvino-sys error is preventing me from getting there

view this post on Zulip Dan Gohman (Feb 17 2021 at 00:39):

OPENVINO_SKIP_LINKING seems to work.

view this post on Zulip Andrew Brown (Feb 17 2021 at 00:39):

(more details here if you care)

Rust bindings for OpenVINO™ . Contribute to intel/openvino-rs development by creating an account on GitHub.

view this post on Zulip Dan Gohman (Feb 17 2021 at 00:44):

I care in the sense that right now, git clone wasmtime + cargo build doesn't seem to work.

view this post on Zulip Dan Gohman (Feb 17 2021 at 00:48):

Or rather, with --workspace it doesn't work.

view this post on Zulip Dan Gohman (Feb 17 2021 at 00:49):

Ok, I'm trying to approach this in "I'll just fix this one thing" mode to get 0.23.0 out, but it's clear that this is going to need more dedicated problem-solving attention mode.

view this post on Zulip Dan Gohman (Feb 17 2021 at 00:52):

There are some witx API changes, and it appears the publish script is unhappy if the witx version in the tree differs from the witx dependencies of other code in the tree, but it'

view this post on Zulip Dan Gohman (Feb 17 2021 at 00:52):

s not trivial to update because of the API changes

view this post on Zulip Chris Fallin (Feb 17 2021 at 00:57):

I poked at this for a few minutes as I was curious -- it seems openvino-sys just doesn't include the submodule. Is the crate upload missing it, or is there a flag somewhere to tell cargo to check it out?

view this post on Zulip Alex Crichton (Feb 17 2021 at 01:02):

@Dan Gohman if you back out the submodule updates does CI pass?

view this post on Zulip Alex Crichton (Feb 17 2021 at 01:02):

those should probably be done in separate PRs rather than during a version bump

view this post on Zulip Dan Gohman (Feb 17 2021 at 01:02):

CI passes, but the publish bump fails

view this post on Zulip Alex Crichton (Feb 17 2021 at 01:03):

oh dear

view this post on Zulip Alex Crichton (Feb 17 2021 at 01:03):

are there logs for that?

view this post on Zulip Dan Gohman (Feb 17 2021 at 01:03):

verify-publish passes on trunk

view this post on Zulip Dan Gohman (Feb 17 2021 at 01:04):

when I run publish bump manually, it's unhappy that the witx in-tree is 0.8.8, while wiggle depends on 0.8.7

view this post on Zulip Andrew Brown (Feb 17 2021 at 01:05):

I poked at this for a few minutes as I was curious -- it seems openvino-sys just doesn't include the submodule. Is the crate upload missing it, or is there a flag somewhere to tell cargo to check it out?

The crate cannot include the submodule code because it exceeds the 10MB crate limit on crates.io; plus, some people may want to link to an installed OpenVINO anyways. OPENVINO_SKIP_LINKING is the escape hatch to avoid linking altogether.

view this post on Zulip Alex Crichton (Feb 17 2021 at 01:05):

@Dan Gohman hm when you say unhappy how so?

view this post on Zulip Dan Gohman (Feb 17 2021 at 01:06):

thread 'main' panicked at '"crates/wiggle/generate/Cargo.toml" has a dep on witx but doesn't list version 0.8.6', scripts/publish.rs:223:17

view this post on Zulip Dan Gohman (Feb 17 2021 at 01:06):

0.8.6 rather than 0.8.7

view this post on Zulip Alex Crichton (Feb 17 2021 at 01:07):

this is probably just a bug in the script

view this post on Zulip Alex Crichton (Feb 17 2021 at 01:07):

it shouldn't be trying to bump deps on witx

view this post on Zulip Alex Crichton (Feb 17 2021 at 01:08):

it's a pretty simple script but it should be ok to edit it to help it limp along

view this post on Zulip Dan Gohman (Feb 17 2021 at 01:10):

Ok, I myself need to log off for the day, so I'll pick this up tomorrow

view this post on Zulip Dan Gohman (Feb 17 2021 at 22:41):

Ok, I edited the script to help it limp along

view this post on Zulip Dan Gohman (Feb 17 2021 at 22:42):

This turned out to be much simpler than the path I took yesterday


Last updated: Nov 22 2024 at 17:03 UTC