SIGSEGV when compiling regex_automata wasm32-unknown-emscripten

[lopopolo]

Code

$ rustc --crate-name regex_automata /Users/lopopolo/.cargo/registry/src/github.com-1ecc6299db9ec823/regex-automata-0.1.8/src/lib.rs --error-format=json --json=diagnostic-rendered-ansi,artifacts --crate-type lib --emit=dep-info,metadata,link -C debuginfo=2 -C metadata=53e0301393a8dd99 -C extra-filename=-53e0301393a8dd99 --out-dir /Users/lopopolo/dev/artichoke/playground/target/wasm32-unknown-emscripten/debug/deps --target wasm32-unknown-emscripten -L dependency=/Users/lopopolo/dev/artichoke/playground/target/wasm32-unknown-emscripten/debug/deps -L dependency=/Users/lopopolo/dev/artichoke/playground/target/debug/deps --extern byteorder=/Users/lopopolo/dev/artichoke/playground/target/wasm32-unknown-emscripten/debug/deps/libbyteorder-72d386ad6d326ca8.rmeta --cap-lints allow
{"artifact":"/Users/lopopolo/dev/artichoke/playground/target/wasm32-unknown-emscripten/debug/deps/regex_automata-53e0301393a8dd99.d","emit":"dep-info"}
{"artifact":"/Users/lopopolo/dev/artichoke/playground/target/wasm32-unknown-emscripten/debug/deps/libregex_automata-53e0301393a8dd99.rmeta","emit":"metadata"}
Called function must be a pointer!
  call addrspace(16046453) void <badref>(i8* %20), !dbg !53
in function _ZN4core6option15Option$LT$T$GT$5ok_or17h536d2927739d3b4fE
LLVM ERROR: Broken function found, compilation aborted!

Meta

rustc --version --verbose:

rustc 1.41.0 (5e1a79984 2020-01-27)
binary: rustc
commit-hash: 5e1a799842ba6ed4a57e91f7ab9435947482f7d8
commit-date: 2020-01-27
host: x86_64-apple-darwin
release: 1.41.0
LLVM version: 9.0

emcc --version:

emcc --version
emcc (Emscripten gcc/clang-like replacement) 1.39.8 (commit 1458145cf4f3db0fb548343e6acab267eef8e4ef)
Copyright (C) 2014 the Emscripten authors (see AUTHORS.txt)
This is free and open source software under the MIT license.
There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Release builds complete successfully.

Error output

Called function must be a pointer!
  call addrspace(16046453) void <badref>(i8* %20), !dbg !53
in function _ZN4core6option15Option$LT$T$GT$5ok_or17h536d2927739d3b4fE
LLVM ERROR: Broken function found, compilation aborted!

Backtrace

{"artifact":"/Users/lopopolo/dev/artichoke/playground/target/wasm32-unknown-emscripten/debug/deps/regex_automata-53e0301393a8dd99.d","emit":"dep-info"}
{"artifact":"/Users/lopopolo/dev/artichoke/playground/target/wasm32-unknown-emscripten/debug/deps/libregex_automata-53e0301393a8dd99.rmeta","emit":"metadata"}
Called function must be a pointer!
  call addrspace(6475917) void Segmentation fault: 11

[lopopolo]

I created this as an internal compiler error. I'm not sure if that's right since this is a segfault and not a panic.

[pnkfelix]

triage: P-high to determine underlying issue and actual severity, and the regression point. Removing nomination. Assigning to self.

[lopopolo]

Put together a smaller reproducer:

Cargo.toml

[package]
name = "bisect"
version = "0.1.0"
authors = ["Ryan Lopopolo <rjl@hyperbo.la>"]
edition = "2018"

# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html

[dependencies]
ansi_term = "0.11"

src/main.rs:

fn main() {
    println!("Hello, world!");
}

This project segfaults rustc when running:

• cargo +1.41.1 build --target wasm32-unknown-emscripten
• cargo +1.41.0 build --target wasm32-unknown-emscripten
• cargo +1.40.0 build --target wasm32-unknown-emscripten

With this emcc (1.39.8-upstream):

$ emcc --version
emcc (Emscripten gcc/clang-like replacement) 1.39.8 (commit 1458145cf4f3db0fb548343e6acab267eef8e4ef)
Copyright (C) 2014 the Emscripten authors (see AUTHORS.txt)
This is free and open source software under the MIT license.
There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Rust 1.39 and older use the fastcomp backend which is incompatible with the default "upstream" (LLVM) backend. Downgrading emcc to 1.38.48-fastcomp enables builds to succeed on 1.37.0, 1.38.0, and 1.39.0.

The segfault does not appear on this nightly.

rustc +nightly --version --verbose
rustc 1.43.0-nightly (6d69caba1 2020-02-27)
binary: rustc
commit-hash: 6d69caba110c0c2fb90180df1cbc8be5033b91d4
commit-date: 2020-02-27
host: x86_64-apple-darwin
release: 1.43.0-nightly
LLVM version: 9.0

[pnkfelix]

Determined via bisection over nightlies that the problem was "fixed" between +nightly-2020-01-25 and +nightly-2020-01-26

Update: More info on those nightlies:

13-28-06 issue69364/repro (git:) % rustc +nightly-2020-01-25 --version
rustc 1.42.0-nightly (c2d141df5 2020-01-24)
13-28-14 issue69364/repro (git:) % rustc +nightly-2020-01-26 --version
rustc 1.42.0-nightly (6d3f4e0aa 2020-01-25)
13-28-16 issue69364/repro (git:) % ( cd ../../rust.git/ && git log c2d141df5..6d3f4e0aa --author=bors --format=oneline )
6d3f4e0aab3e36ceb8b83d1e9467514685f6b751 Auto merge of #68530 - estebank:abolish-ice, r=petrochenkov
8ad83afe5bcfa983a24b6f720c9ef389350f414b Auto merge of #68525 - tlively:emcc-codegen-sigsegv-66308, r=alexcrichton
3bf71b3d5c07afc0908881a0868080ed399ca6ca Auto merge of #68516 - oli-obk:spaces, r=eddyb
80a65bcaf2f2b8a5c659b21b32b42bc300338a0e Auto merge of #68448 - maurer:dyn-cdylib, r=alexcrichton
8bf17584e01d85c631bcb19b7cd0f95e84c9a9b1 Auto merge of #68269 - csmoe:temp, r=estebank
8647aa1a2ce279f8ec7cc5252d10b8cb9ea504eb Auto merge of #68526 - JohnTitor:rollup-3mmljof, r=JohnTitor
13-28-19 issue69364/repro (git:) % 

[pnkfelix]

Looks like this was deliberately fixed via an LLVM upgrade in PR #68525.

[pnkfelix]

And it looks like the above fix is part of the beta channel, which is imminently scheduled for release.

Closing as duplicate of #66308 (which is fixed).