Stream: wasmtime

Topic: broken build


view this post on Zulip Andrew Brown (Oct 28 2020 at 20:36):

Wasmtime looks like it was broken from the load_splat PR: https://github.com/bytecodealliance/wasmtime/commits

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

view this post on Zulip Andrew Brown (Oct 28 2020 at 20:38):

@Chris Fallin, @Anton Kirilov: @Johnnie Birch and I initially thought it was an FP conversion PR that did it but then noticed that there were a few more places in https://github.com/bytecodealliance/wasmtime/pull/2278 that need a SourceLoc

It corresponds to WebAssembly's load*_splat operations, which were previously represented as a combination of Load and Splat instructions. However, there are architectures such as Armv8-A that ...

view this post on Zulip Chris Fallin (Oct 28 2020 at 20:38):

Hmm, rebase issue it looks like

view this post on Zulip Andrew Brown (Oct 28 2020 at 20:39):

My current guess on what went wrong is that in the time between creating the load_splat PR and merging we added more instructions that use Inst::xmm_rm_r and the load_splat PR adds a parameter to that method

view this post on Zulip Chris Fallin (Oct 28 2020 at 20:40):

Yeah it looks like it. I can revert in the meantime then we can re-land

view this post on Zulip Andrew Brown (Oct 28 2020 at 20:40):

I wouldn't be opposed to a new PR that just fixes those places

view this post on Zulip Andrew Brown (Oct 28 2020 at 20:40):

shouldn't be too many: 14 at this point

view this post on Zulip Chris Fallin (Oct 28 2020 at 20:40):

That's fine too if it's a small fix (I haven't looked in detail) -- thanks!

view this post on Zulip Andrew Brown (Oct 28 2020 at 20:41):

Ok, I'll tag you guys in a PR in a few minutes

view this post on Zulip Anton Kirilov (Oct 28 2020 at 20:44):

@Andrew Brown @Chris Fallin If it is just that, can we push a fix instead of reverting?

view this post on Zulip Andrew Brown (Oct 28 2020 at 20:44):

Yeah, I'm building it right now

view this post on Zulip Andrew Brown (Oct 28 2020 at 20:47):

https://github.com/bytecodealliance/wasmtime/pull/2335

The changes in #2278 added SourceLocs to several x64 Inst variants; between when that PR was last run in CI and when it was merged, new instructions were added that require this new parameter. This...

view this post on Zulip Johnnie Birch (Oct 28 2020 at 20:47):

Yes .. it looks like incorrect number of parameters for some lowering functions on the x64 side. One of us .. but I'd be happy to take a stab now at a patch.

view this post on Zulip Chris Fallin (Oct 28 2020 at 20:48):

Thanks @Andrew Brown, that was quick!

view this post on Zulip Andrew Brown (Oct 28 2020 at 20:49):

Let's see how the CI reacts... :fingers_crossed:

view this post on Zulip Chris Fallin (Oct 28 2020 at 20:54):

Ah, looks like one was missed (x64/emit_tests.rs)

view this post on Zulip Andrew Brown (Oct 28 2020 at 20:56):

Bleh

view this post on Zulip Andrew Brown (Oct 28 2020 at 20:57):

Ok, it's building again..

view this post on Zulip Johnnie Birch (Oct 28 2020 at 21:13):

Although looks like a failure first occurs in the patch before mine, but looks like most the changes are in my code. An API change was merged as my tests were running. I just pushed a PR for a fix here: https://github.com/bytecodealliance/wasmtime/pull/2337

GitHub is home to over 50 million developers working together to host and review code, manage projects, and build software together.

view this post on Zulip Johnnie Birch (Oct 28 2020 at 21:16):

Oh .. lol. Looks like I was missing some of the conversation. I guess there are two patches. I can delete mine.

view this post on Zulip Chris Fallin (Oct 28 2020 at 21:17):

@Johnnie Birch no worries, was just looking to see if we were missing something... thank you though :-)

view this post on Zulip Anton Kirilov (Oct 28 2020 at 22:25):

I'd also like to thank @Andrew Brown for the quick fix!

view this post on Zulip Andrew Brown (Oct 29 2020 at 01:11):

So we fixed the broken CI, right? I just merged a pretty innocuous patch adding pmin/pmax support and the CI, which passed on the PR after a rebase on main, broke on main when I merged it in. The errors are:

Since none of these are obviously related to anything we've done recently, these tests have been passing consistently for ages, and the subsequent commits all pass, I propose we just observe and see if it happens again. But, yeah, please keep an eye out for this...

GitHub is where people build software. More than 50 million people use GitHub to discover, fork, and contribute to over 100 million projects.
GitHub is where people build software. More than 50 million people use GitHub to discover, fork, and contribute to over 100 million projects.
GitHub is where people build software. More than 50 million people use GitHub to discover, fork, and contribute to over 100 million projects.

view this post on Zulip Andrew Brown (Mar 02 2021 at 20:27):

@Alex Crichton, more breakages: curl: (6) Could not resolve host: git.io in https://github.com/bytecodealliance/wasmtime/runs/2015963896?check_suite_focus=true

GitHub is where people build software. More than 56 million people use GitHub to discover, fork, and contribute to over 100 million projects.

view this post on Zulip Andrew Brown (Mar 02 2021 at 20:30):

I merged https://github.com/bytecodealliance/wasmtime/pull/2699 so let's see if the DNS resolution is fixed

This instruction has a single instruction lowering in AVX512F/VL and a three instruction lowering in AVX but neither is currently supported in the x64 backend. To implement this, we instead subtrac...

view this post on Zulip Andrew Brown (Mar 02 2021 at 21:57):

And another weird failure: thread 'main' panicked at 'success: Os { code: 2, kind: NotFound, message: "No such file or directory" }', tests/all/debug/gdb.rs:30:31 in https://github.com/bytecodealliance/wasmtime/runs/2016650452?check_suite_focus=true#step:13:119

GitHub is where people build software. More than 56 million people use GitHub to discover, fork, and contribute to over 100 million projects.

view this post on Zulip Andrew Brown (May 17 2021 at 18:33):

Anyone know what is going on when the "publish gh-pages" step fails? See https://github.com/abrown/wasmtime/runs/2603472822?check_suite_focus=true#step:13:12.

GitHub is where people build software. More than 65 million people use GitHub to discover, fork, and contribute to over 200 million projects.

view this post on Zulip Chris Fallin (May 17 2021 at 18:35):

Related to this being a fork maybe? It feels like a "key is only available in canonical repo" sort of issue...

view this post on Zulip Chris Fallin (May 17 2021 at 18:36):

weirdly if I take the last main-repo CI run it looks like that step is skipped ("/" icon): https://github.com/bytecodealliance/wasmtime/runs/2603390153?check_suite_focus=true

GitHub is where people build software. More than 65 million people use GitHub to discover, fork, and contribute to over 200 million projects.

view this post on Zulip Stephan Renatus (May 17 2021 at 20:13):

What you've linked there (the disabled step) is disabled because it's on a branch != main (a PR) -- after it was merged, the gh-push in "Publish" happened: <https://github.com/bytecodealliance/wasmtime/runs/2603915443?check_suite_focus=true>. the config controlling this is <https://github.com/bytecodealliance/wasmtime/blob/main/.github/workflows/main.yml#L591>

GitHub is where people build software. More than 65 million people use GitHub to discover, fork, and contribute to over 200 million projects.
Standalone JIT-style runtime for WebAssembly, using Cranelift - bytecodealliance/wasmtime

view this post on Zulip Stephan Renatus (May 17 2021 at 20:14):

The step failed for @Andrew Brown 's fork because it was run -- there was push to that fork's main. It's not supposed to be run for forks, which usually use a different branch -- for the reason @Chris Fallin mentioned, secrets.

view this post on Zulip Andrew Brown (May 17 2021 at 21:23):

Ah, that sort of makes sense. I did push main to my fork so that I could see better which of my branches I should delete. I guess I didn't know that the CI would build my fork...

view this post on Zulip Chris Fallin (May 17 2021 at 21:28):

FWIW I think you can disable GitHub Actions explicitly in your fork (in the repo settings) -- this should prevent the problem in the future :-)

view this post on Zulip Stephan Renatus (May 18 2021 at 07:40):

just for clarity, looks like me second reference already changed underneath -- the stable link to the config part is s https://github.com/bytecodealliance/wasmtime/blob/18c61cdfa4bf4944dd09a8d383bedd85f6b97b46/.github/workflows/main.yml#L604

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

view this post on Zulip Stephan Renatus (May 18 2021 at 07:41):

Adding something like github.repository == 'bytecodealliance/wasmtime' would be another option -- any forks will not run that specific step then.

view this post on Zulip Andrew Brown (Jul 22 2021 at 18:00):

For those interested in CI issues, I've been noticing more failures than normal recently: https://github.com/bytecodealliance/wasmtime/commits/main. Anyone know what's going on?

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

view this post on Zulip Andrew Brown (Jul 22 2021 at 18:01):

A few might have been from network errors but I'm seeing some like "4 successful and 19 cancelled checks"?

view this post on Zulip Alex Crichton (Jul 22 2021 at 18:30):

if a commit is pushed to the main branch while CI is running on a previous commit, the previous commit's CI is cancelled

view this post on Zulip Alex Crichton (Jul 22 2021 at 18:30):

to reduce the amount of resources we take in CI

view this post on Zulip Andrew Brown (Aug 06 2021 at 16:48):

Looks like the latest CI build on main is failing because of an API limit issue: https://github.com/bytecodealliance/wasmtime/runs/3263984846?check_suite_focus=true#step:12:3507.

GitHub is where people build software. More than 65 million people use GitHub to discover, fork, and contribute to over 200 million projects.

view this post on Zulip Andrew Brown (Aug 10 2021 at 17:00):

@Alex Crichton, @Pat Hickey: looks like wasi_tokio::poll_oneoff_files is creating issues again. Should I still merge that PR?

@radu-matei has recently been working on an alternate backend for wasi-nn using ONNX and has written about it on his blog. This PR refactors the wasi-nn crate to support multiple backends. The wasi...

view this post on Zulip Alex Crichton (Aug 10 2021 at 17:04):

yeah that should be fine to merge

view this post on Zulip Andrew Brown (Aug 10 2021 at 20:51):

Are builds now hanging? https://github.com/bytecodealliance/wasmtime/pull/3177/checks?check_run_id=3294242349? I ran ci/run-tests.sh on main locally to see if I could reproduce the hang-up but it finished successfully in 2m 30s.

Previously, the WAT was printed as a log message. This change standardizes all of the oracles to use log_wasm, which emits .wasm and .wat file for each case if log::debug is enabled and prints a me...

view this post on Zulip Andrew Brown (Aug 10 2021 at 20:52):

...or maybe GitHub actions is having problems: https://www.githubstatus.com.

view this post on Zulip Andrew Brown (Aug 11 2022 at 17:19):

I think main is broken due to some spurious error: https://github.com/bytecodealliance/wasmtime/runs/7791537726?check_suite_focus=true#step:4:106

A standalone runtime for WebAssembly. Contribute to bytecodealliance/wasmtime development by creating an account on GitHub.

view this post on Zulip Andrew Brown (Aug 11 2022 at 17:20):

cc: @Alex Crichton, @Chris Fallin, @fitzgen (he/him), @Dan Gohman

view this post on Zulip Andrew Brown (Aug 11 2022 at 17:20):

I was going to remove that mut but it is actually necessary... some tool seems to be broken

view this post on Zulip Andrew Brown (Aug 11 2022 at 17:22):

I guess it's the latest update of Rust to 1.63.0: https://github.com/bytecodealliance/wasmtime/runs/7791537726?check_suite_focus=true#step:3:7

A standalone runtime for WebAssembly. Contribute to bytecodealliance/wasmtime development by creating an account on GitHub.

view this post on Zulip Alex Crichton (Aug 11 2022 at 17:28):

sorry just noticed this, I posted https://github.com/bytecodealliance/wasmtime/pull/4691 with what I think is the fix

This commit fixes a build warning on Rust 1.63 when the memory-init-cow feature is disabled in the wasmtime-runtime crate. Some "tricks" were used prior to have the MemoryImage type be an empty enu...

view this post on Zulip Andrew Brown (Aug 11 2022 at 17:30):

oh... jinx: https://github.com/bytecodealliance/wasmtime/pull/4692

This change temporarily pins the Rust toolchain used to 1.62.1. The switch to 1.63.0 on August 11th seems to cause issues when checking the mutability of certain variables: https://github.com/bytec...

view this post on Zulip Andrew Brown (Aug 11 2022 at 17:31):

(and yours is the actual fix... closing)


Last updated: Dec 23 2024 at 14:03 UTC