generate precise traps in component adapters (#12215)

* generate precise traps in component adapters

Previously, we generated a generic `unreachable` instruction for each trap in a
fused adapter, p

generate precise traps in component adapters (#12215)

* generate precise traps in component adapters

Previously, we generated a generic `unreachable` instruction for each trap in a
fused adapter, plus some metadata to recover the specific kind of trap.
However, that metadata was never hooked up to anything, so we only got a generic
`unreachable` message at runtime.

This commit removes the metadata tracking and instead simply calls a host
intrinsic, passing the trap code as a parameter. This is somewhate pessimal
compared with what we had before, but improves ergonomics and allows us to avoid
forking as many tests from the component-model repo.

If performance becomes an issue, we can easily add an option to skip the host
call and only emit an `unreachable` instruction as before.

Note that I've removed forked versions of three tests in favor of their upstream
equivalents.

Fixes #11683

* update `strings.rs` tests

* bless disas tests

show more ...

Make Wasmtime's `FuncKey` one-to-one with Cranelift's `ir::UserExternalName` (#11415)

* Make Wasmtime's `FuncKey` one-to-one with Cranelift's `ir::UserExternalName`

`FuncKey`, which used to be call

Make Wasmtime's `FuncKey` one-to-one with Cranelift's `ir::UserExternalName` (#11415)

* Make Wasmtime's `FuncKey` one-to-one with Cranelift's `ir::UserExternalName`

`FuncKey`, which used to be called `CompileKey`, is now one-to-one with
`cranelift_codegen::ir::UserExternalName`, and is used for not just identifying
compilation objects but also relocations and call-graph edges. This allows us to
determine the `StaticModuleIndex` and `DefinedFuncIndex` pair for any
`cranelift_codegen::ir::FuncRef`, regardless of inlining depth, which fixes some
fuzz bugs on OSS-Fuzz.

This continues pushing on the idea that Wasmtime's compilation orchestration and
linking should be relatively agnostic to the kinds of things it is actually
compiling and linking, allowing us to tweak, add, and remove new kinds of
`FuncKey`s more easily. Adding a new `FuncKey` should not require modifying
relocation resolution, for example, just a little bit of code to run the
associated compilation and optionally some code to extract metadata into our
final artifacts for querying at runtime. Everything in between should Just
Continue Working. We still aren't all the way there yet, but this does bring us
a little bit closer.

Finally, in Cranelift's inlining pass, this adds a check that a block is
inserted in the layout before attempting to remove it from the layout, which
would otherwise cause panics. This was triggered by multi-level inlining and
now-unreachable blocks in the inner callees.

I'll note that this does update basically all of the disas tests, or at least
nearly all of them that make function calls. This is because the namespace/index
numbering pair changed slightly to align with `FuncKey`, but that should pretty
much be the only changes.

* remove debug info from panic message, it is only available in some `cfg`s

* fill out module doc comment

* Fix compilation without `component-model` feature

* Fix some more cfg compilations

* cargo fmt

* fix a wrong `&dyn Any` auto coercion; add helpful debug logging and assertions for this kind of thing

show more ...

Make direct calls to statically-known imported functions (#11228)

* Make direct calls to statically-known imported functions

This will allow for their eventual inlining.

* address review feedback

Make direct calls to statically-known imported functions (#11228)

* Make direct calls to statically-known imported functions

This will allow for their eventual inlining.

* address review feedback and fix up some disas tests

* Add native code disassembly test

* disas: sort `read_dir` entries for deterministic order

Because we always use namespace 0 for wasm functions when creating CLIF
functions, sorting the parsed CLIF functions by (namespace, index) does not
differentiate between the `index`th Wasm function in two different modules in
the same component, meaning we would pass through whatever ordering reading from
the filesystem gave us, which can be non-deterministic across
platforms. Instead, sort the directory entries based on file paths.

* fix clippy

show more ...

Handle `uextend` in conditional trap lowerings on x64 (#11044)

* Handle uextend in conditional trap lowerings on x64

* Logging and rebuilding for test programs and debug tests

* Forward srcloc deb

Handle `uextend` in conditional trap lowerings on x64 (#11044)

* Handle uextend in conditional trap lowerings on x64

* Logging and rebuilding for test programs and debug tests

* Forward srcloc debug info to conditional trap instructions from conditional branches to unconditional traps

show more ...

Remove dead `callee` field of vmctx (#11014)

* Remove dead `callee` field of vmctx

This was used long ago but refactorings since then have removed the need
for its existence, so remove it outright.

Remove dead `callee` field of vmctx (#11014)

* Remove dead `callee` field of vmctx

This was used long ago but refactorings since then have removed the need
for its existence, so remove it outright.

* Update test expectations

* Shuffle around some imports

show more ...

Cranelift: Rewrite conditional branches to unconditional traps into conditional traps during legalization (#10988)

* Cranelift: Legalize via a backwards walk, rather than forwards

Note: the test ex

Cranelift: Rewrite conditional branches to unconditional traps into conditional traps during legalization (#10988)

* Cranelift: Legalize via a backwards walk, rather than forwards

Note: the test expectation change in `filetests/egraph/misc.clif` is simply
because we happen to change the order in which we created the legalized
`stack_addr` instructions that get GVN'd together, and therefore also changed
their relative value numbering. We dedupe to the value that occurs first in the
function, which is the one inserted into the DFG *after* the other now because
of the backwards traversal, and it therefore has a different value number from
before. The resulting program is identical, modulo value numbering.

* Cranelift: Rewrite conditional branches to unconditional traps into conditional traps during legalization

Given this instruction:

```clif
brif v0, block1, block2
```

If we know that `block1` does nothing but immediately trap then we can rewrite
that `brif` into the following:

```clif
trapz v0, <trapcode>
jump block2
```

(And we can do the equivalent with `trapz` if `block2` immediately traps).

This transformation allows for the conditional trap instructions to be GVN'd and
for our egraphs mid-end to generally better optimize the program. We
additionally have better codegen in our backends for `trapz` than branches to
unconditional traps.

Fixes https://github.com/bytecodealliance/wasmtime/issues/10941

* Update CLIF filetests and Wasmtime disas tests

show more ...

Align symbol names with `--emit-clif` (#10947)

* Align symbol names with `--emit-clif`

This commit fixes two issues: (1) is that `--emit-clif` generates file
names that are different from symbol na

Align symbol names with `--emit-clif` (#10947)

* Align symbol names with `--emit-clif`

This commit fixes two issues: (1) is that `--emit-clif` generates file
names that are different from symbol names and (2) is that `--emit-clif`
for components today has collisions on filenames which means that output
files can get corrupted. Currently (2) is causing spurious failures in
CI due to the test added in #10939 being the first test with a component
and functions in modules where they clash to emit the same file.

* Fix typo

* Fix tests

* Fix tests on Windows

Can't use `:` in filenames on Windows

prtest:full

show more ...

Simplify and optimize the way that `enum`s are passed in fused adapters (#10939)

* Simplify and optimize the way that `enum`s are passed in fused adapters

instead of treating them like variants tha

Simplify and optimize the way that `enum`s are passed in fused adapters (#10939)

* Simplify and optimize the way that `enum`s are passed in fused adapters

instead of treating them like variants that don't have payloads, just check that
the discriminant is in bounds.

* cargo +stable fmt

show more ...