Strip component-type* custom sections during componentization (#397)

This commit leverages a new convenience API in `wasmparser` to strip the
`component-type*` custom sections emitted by `wit-bindgen` guests during
the componentization process for a core wasm module. These custom
sections don't have any place in the core wasm module except as part of
the build process and will end up just duplicating bits and pieces of
the component itself, so there's no need for them to stick around once a
component has been created.

Author: alexcrichton

Cargo.lock

@@ -27,10 +27,10 @@ env_logger = "0.9.1"
 
 wasmtime = { git = "https://github.com/bytecodealliance/wasmtime", features = ["component-model"] }
 wasmtime-environ = { git = "https://github.com/bytecodealliance/wasmtime" }
-wasmprinter = "0.2.41"
-wasmparser = "0.92.0"
-wasm-encoder = "0.18.0"
-wat = "1.0.49"
+wasmprinter = "0.2.42"
+wasmparser = "0.93.0"
+wasm-encoder = "0.19.0"
+wat = "1.0.50"
 
 wit-bindgen-core = { path = 'crates/bindgen-core', version = '0.3.0' }
 wit-bindgen-gen-guest-c = { path = 'crates/gen-guest-c', version = '0.3.0' }

Cargo.toml

@@ -12,7 +12,6 @@ wit-parser = { workspace = true }
 anyhow = { workspace = true }
 wit-component = { workspace = true }
 wasmtime-environ = { workspace = true, features = ['component-model'], optional = true }
-wasmparser = { workspace = true, optional = true }
 
 [features]
-component-generator = ['dep:wasmtime-environ', 'dep:wasmparser']
+component-generator = ['dep:wasmtime-environ']

crates/bindgen-core/Cargo.toml

@@ -9,10 +9,10 @@
 
 use crate::{Files, WorldGenerator};
 use anyhow::{Context, Result};
-use wasmparser::{Validator, WasmFeatures};
 use wasmtime_environ::component::{
     Component, ComponentTypesBuilder, StaticModuleIndex, Translator,
 };
+use wasmtime_environ::wasmparser::{Validator, WasmFeatures};
 use wasmtime_environ::{ModuleTranslation, PrimaryMap, ScopeVec, Tunables};
 use wit_component::ComponentInterfaces;
 

crates/bindgen-core/src/component.rs

@@ -23,23 +23,27 @@ pub struct ModuleInterfaces {
 /// later on.
 pub fn extract_module_interfaces(wasm: &[u8]) -> Result<ModuleInterfaces> {
     let mut ret = ModuleInterfaces::default();
+    let mut new_module = wasm_encoder::Module::new();
 
     for payload in wasmparser::Parser::new(0).parse_all(wasm) {
-        if let wasmparser::Payload::CustomSection(cs) =
-            payload.context("decoding item in module")?
-        {
-            if !cs.name().starts_with("component-type") {
-                continue;
+        let payload = payload.context("decoding item in module")?;
+        match payload {
+            wasmparser::Payload::CustomSection(cs) if cs.name().starts_with("component-type") => {
+                ret.decode(cs.data())
+                    .with_context(|| format!("decoding custom section {}", cs.name()))?;
+            }
+            _ => {
+                if let Some((id, range)) = payload.as_section() {
+                    new_module.section(&wasm_encoder::RawSection {
+                        id,
+                        data: &wasm[range],
+                    });
+                }
             }
-            ret.decode(cs.data())
-                .with_context(|| format!("decoding custom section {}", cs.name()))?;
         }
     }
 
-    // TODO: should remove the custom sections decoded above from the wasm binary
-    // created here, and bytecodealliance/wasmparser#792 should help with that
-    // to make the loop above pretty small.
-    ret.wasm = wasm.to_vec();
+    ret.wasm = new_module.finish();
 
     Ok(ret)
 }

crates/wit-component/src/extract.rs

@@ -26,7 +26,7 @@
   )
   (core instance (;0;) (instantiate 2))
   (alias core export 0 "0" (core func (;0;)))
-  (core instance (;1;) 
+  (core instance (;1;)
     (export "thunk" (func 0))
   )
   (core instance (;2;) (instantiate 1
@@ -36,7 +36,7 @@
   (core instance (;3;) (instantiate 0))
   (alias core export 0 "$imports" (core table (;0;)))
   (alias core export 3 "thunk" (core func (;1;)))
-  (core instance (;4;) 
+  (core instance (;4;)
     (export "$imports" (table 0))
     (export "0" (func 1))
   )

crates/wit-component/tests/components/adapt-empty-interface/component.wat

@@ -32,7 +32,7 @@
       local.tee 0
       i32.const -1
       i32.eq
-      if  ;; label = @1
+      if ;; label = @1
         unreachable
       end
       local.get 0
@@ -48,7 +48,7 @@
     (start $initialize_stack_pointer)
   )
   (type (;0;) (func (result "a" u32) (result "b" u32)))
-  (type (;1;) 
+  (type (;1;)
     (instance
       (alias outer 1 0 (type (;0;)))
       (export "get-two" (func (type 0)))
@@ -88,19 +88,19 @@
   )
   (core instance (;0;) (instantiate 2))
   (alias core export 0 "1" (core func (;0;)))
-  (core instance (;1;) 
+  (core instance (;1;)
     (export "get_sum" (func 0))
   )
   (core instance (;2;) (instantiate 1
       (with "old" (instance 1))
     )
   )
   (alias core export 2 "memory" (core memory (;0;)))
-  (core instance (;3;) 
+  (core instance (;3;)
     (export "memory" (memory 0))
   )
   (alias core export 0 "0" (core func (;1;)))
-  (core instance (;4;) 
+  (core instance (;4;)
     (export "get-two" (func 1))
   )
   (core instance (;5;) (instantiate 0
@@ -112,7 +112,7 @@
   (alias export 0 "get-two" (func (;0;)))
   (core func (;2;) (canon lower (func 0) (memory 0)))
   (alias core export 5 "get_sum" (core func (;3;)))
-  (core instance (;6;) 
+  (core instance (;6;)
     (export "$imports" (table 0))
     (export "0" (func 2))
     (export "1" (func 3))

crates/wit-component/tests/components/adapt-inject-stack/component.wat

@@ -17,7 +17,7 @@
   )
   (type (;0;) (list u8))
   (type (;1;) (func (result 0)))
-  (type (;2;) 
+  (type (;2;)
     (instance
       (alias outer 1 1 (type (;0;)))
       (export "read" (func (type 0)))
@@ -59,7 +59,7 @@
   )
   (core instance (;0;) (instantiate 2))
   (alias core export 0 "1" (core func (;0;)))
-  (core instance (;1;) 
+  (core instance (;1;)
     (export "read" (func 0))
   )
   (core instance (;2;) (instantiate 1
@@ -68,7 +68,7 @@
   )
   (alias core export 2 "memory" (core memory (;0;)))
   (alias core export 0 "0" (core func (;1;)))
-  (core instance (;3;) 
+  (core instance (;3;)
     (export "read" (func 1))
   )
   (core instance (;4;) (instantiate 0
@@ -80,7 +80,7 @@
   (alias export 0 "read" (func (;0;)))
   (core func (;3;) (canon lower (func 0) (memory 0) (realloc 2)))
   (alias core export 4 "read" (core func (;4;)))
-  (core instance (;5;) 
+  (core instance (;5;)
     (export "$imports" (table 0))
     (export "0" (func 3))
     (export "1" (func 4))

crates/wit-component/tests/components/adapt-list-return/component.wat

@@ -5,7 +5,7 @@
     (export "log" (func $log))
   )
   (type (;0;) (func (param "s" string)))
-  (type (;1;) 
+  (type (;1;)
     (instance
       (alias outer 1 0 (type (;0;)))
       (export "log" (func (type 0)))
@@ -46,7 +46,7 @@
   )
   (core instance (;0;) (instantiate 2))
   (alias core export 0 "1" (core func (;0;)))
-  (core instance (;1;) 
+  (core instance (;1;)
     (export "log" (func 0))
   )
   (core instance (;2;) (instantiate 1
@@ -55,7 +55,7 @@
   )
   (alias core export 2 "memory" (core memory (;0;)))
   (alias core export 0 "0" (core func (;1;)))
-  (core instance (;3;) 
+  (core instance (;3;)
     (export "log" (func 1))
   )
   (core instance (;4;) (instantiate 0
@@ -66,7 +66,7 @@
   (alias export 0 "log" (func (;0;)))
   (core func (;2;) (canon lower (func 0) (memory 0) string-encoding=utf8))
   (alias core export 4 "log" (core func (;3;)))
-  (core instance (;5;) 
+  (core instance (;5;)
     (export "$imports" (table 0))
     (export "0" (func 2))
     (export "1" (func 3))