Test Pulley on 32-bit platforms in CI#9745
Conversation
github-actions
Bot
added
cranelift
Label Messager: wasmtime:configIt looks like you are changing Wasmtime's configuration options. Make sure to
DetailsTo modify this label's message, edit the To add new label messages or remove existing label messages, edit the |
alexcrichton
force-pushed
the
pulley-runtime
branch
from
a318d49 to
7108888
Compare
| sharded.push(Object.assign( | ||
| {}, | ||
| config, | ||
| { bucket: '--test wast' }, | ||
| )); |
There was a problem hiding this comment.
Is this not going to break for the crates which support 32-bit but don't have a tests/wast.rs? I.e. all of them other than wasmtime? But even the wasmtime crate doesn't have that test, only wasmtime-cli does... I guess I am pretty confused around the logic here...
There was a problem hiding this comment.
This is abusing the fact that cargo test --workspace --test wast only actually executes tests named wast and everything else is ignored. So this effectively runs cargo test --test wast and nothing else is named wast.
In the long-run I want to get more 32-bit tests running, just needs more support in Pulley for that.
| // FIXME: at least on the `gcc-arm-linux-gnueabihf` toolchain on Ubuntu | ||
| // these symbols are not provided by default like they are on other targets. | ||
| // I'm not ARM expert so I don't know why. For now though consider this an | ||
| // optional integration feature with the platform and stub out the functions | ||
| // to do nothing which won't break any tests it just means that | ||
| // runtime-generated backtraces won't have the same level of fidelity they | ||
| // do on other targets. | ||
| if #[cfg(target_arch = "arm")] { | ||
| unsafe extern "C" fn __register_frame(_: *const u8) {} | ||
| unsafe extern "C" fn __deregister_frame(_: *const u8) {} | ||
| unsafe extern "C" fn wasmtime_using_libunwind() -> bool { | ||
| false | ||
| } | ||
| } else { | ||
| extern "C" { | ||
| // libunwind import | ||
| fn __register_frame(fde: *const u8); | ||
| fn __deregister_frame(fde: *const u8); | ||
| #[wasmtime_versioned_export_macros::versioned_link] | ||
| fn wasmtime_using_libunwind() -> bool; | ||
| } | ||
| } |
There was a problem hiding this comment.
@Kmeakin, happen to know how we can properly get access to these JIT code registration functions so that profiling and debugging and all that with the system tools is a better experience?
This commit extends our CI for i686 and armv7 to test the Pulley backend, namely the full `*.wast` test suite as well as the `wasmtime` crate itself. Note that many `*.wast` tests are still expected to fail at this time. This involved fixing a number of 32-vs-64 bit issues throughout the test suite in various location in this commit.
alexcrichton
force-pushed
the
pulley-runtime
branch
from
7108888 to
e0a36a6
Compare
2 participants
This PR is based on #9743 and #9744 and those should be reviewed first.
Opening this to get a full run of CI to see what happens on these platforms...