crossword-composer VS toolchains_llvm

A few wasm projects I've worked on:

- An in-browser crossword puzzle generator: https://crossword.paulbutler.org/ (source: https://github.com/paulgb/crossword-composer)

- A multi-player word game: https://redwords.paulbutler.org/

- A library for synchronizing state between clients, used for that word game: https://aper.dev/ (source: https://github.com/aper-dev/aper very WIP right now)

In my experience, the single biggest perk of using WebAssembly is that I can use a language I'm very productive in (Rust) compared to JavaScript. Everything else is secondary. That said, I think these projects have specific advantages by virtue of being WebAssembly:

- The backtracking search used for the crossword puzzle generator is carefully implemented to not allocate extra memory. This would be tough to do in JavaScript, and I believe it's partly responsible for its performance.

- The word game uses a compression algorithm that benefits very noticeably from wasm-opt, to the point that I can't run it without it. Given that wasm-opt takes a non-trivial amount of time at compile time, I suspect the JavaScript JIT would be slow at doing something similar at runtime. This is just conjecture, I haven't checked.

- What Aper does just wouldn't be possible without Rust features like Serde and macros.

load("@bazel_tools//tools/build_defs/repo:http.bzl", "http_archive") BAZEL_TOOLCHAIN_TAG = "0.7.2" BAZEL_TOOLCHAIN_SHA = "f7aa8e59c9d3cafde6edb372d9bd25fb4ee7293ab20b916d867cd0baaa642529" http_archive( name = "com_grail_bazel_toolchain", sha256 = BAZEL_TOOLCHAIN_SHA, strip_prefix = "bazel-toolchain-{tag}".format(tag = BAZEL_TOOLCHAIN_TAG), canonical_id = BAZEL_TOOLCHAIN_TAG, url = "https://github.com/grailbio/bazel-toolchain/archive/{tag}.tar.gz".format(tag = BAZEL_TOOLCHAIN_TAG), ) load("@com_grail_bazel_toolchain//toolchain:deps.bzl", "bazel_toolchain_dependencies") bazel_toolchain_dependencies() load("@com_grail_bazel_toolchain//toolchain:rules.bzl", "llvm_toolchain") llvm_toolchain( name = "llvm_toolchain", llvm_version = "15.0.5", ) load("@llvm_toolchain//:toolchains.bzl", "llvm_register_toolchains") llvm_register_toolchains() http_archive( name = "com_google_googletest", urls = ["https://github.com/google/googletest/archive/609281088cfefc76f9d0ce82e1ff6c30cc3591e5.zip"], strip_prefix = "googletest-609281088cfefc76f9d0ce82e1ff6c30cc3591e5", )

Yeah the cross-compilation thing is definitely a rough spot. I have one project that's able to work around it via extensive hacks with macros, but at some point I'll need to do it "the right way."

Honestly if the docs had a canonical example of e.g. using unix_cc_toolchain_config (example: [0]) + Bootlin to compile for aarch64, it'd probably go a long way to making things understandable. Because say what you will about the old CROSSTOOL approach, at least there was a nice tutorial for it.

[0] https://github.com/grailbio/bazel-toolchain/blob/f14a8a5de8f...

I'm really not familiar with this issue or Go nor C++ overall, but if all you need is to set up a C++ toolchain, this should be quite simple and solve your issue.

The trick is that to provide Bazel with a custom toolchain involves way more than just setting an environment variable, because Bazel wants to control installing and making available the compiler reliably (e.g., what if `emcc` is not present on the system where Bazel was invoked? Bazel solves that problem by fetching it and building it for that system)

There are projects that provide drop-in support for custom toolchains (e.g., we use this project[0] in Sorbet to fetch and build a custom LLVM/Clang toolchain for every host we build on (rather than relying on the system toolchain). But I'm not aware of a project that has done that for Emscripten. Maybe it would be as easy as plucking out what we've done in our project into a project that others could depend on, but to quote a colleague:

> Setting up a cc toolchain in Bazel is a unique sort of pain.

[0] https://github.com/grailbio/bazel-toolchain/