feat: production-ready WASI + VDP chain fixes + filter chain tests (v3.1.0)

Production-ready WASI implementations:
- fd_write: Full iovec parsing, bounds-checked, stderr output for journal
- clock_time_get: Real nanosecond precision (REALTIME + MONOTONIC)
- random_get: Cryptographically secure (getrandom/arc4random_buf/urandom)

VDP chain API fixes:
- Updated to Varnish 7.6+ void **priv callback signatures
- Added vwasm_engine_get_pool() accessor for opaque pool access
- Removed duplicate *priv = NULL in vdp_wasm_fini

Test infrastructure:
- Added passthrough.wasm and transform.wasm filter chain test modules
- Fixed VTC server start deadlocks (pool_stats, filter_chain)
- Fixed filter_chain repeat count (4 clients = -repeat 4)
- Removed non-linkable unit tests; 19 VTC integration tests, all passing

All 19/19 VTC tests pass on Varnish 7.6 / wasmtime 28.

Author: RamazanKara

README.md

@@ -16,7 +16,7 @@ Includes a full [Proxy-Wasm ABI v0.2.1](https://github.com/proxy-wasm/spec) impl
 - Load `.wasm` modules at VCL init time
 - Call exported Wasm functions from VCL
 - Full Proxy-Wasm ABI v0.2.1 (header maps, trailers, buffers, HTTP callouts, properties, shared data, metrics, tick timer, stream control)
-- WASI stub support (modules compiled with `wasm32-wasi` target work out of the box)
+- WASI support (`fd_write`, `clock_time_get`, `random_get` with real implementations)
 - Epoch-based execution time limits (low-overhead, no per-instruction cost)
 - Memory limits (default 16 MiB)
 - Store pooling for fast per-request instantiation

docs/PRODUCTION.md

@@ -144,11 +144,11 @@ sub vcl_deliver {
 }
 ```
 
-- The `wasm_body` VDP buffers response body (up to 1 MiB) and calls
-  `proxy_on_response_body` on stream end
-- Body is passed through to the client immediately (no buffering delay)
-- The callback is for inspection only — body modification is not supported
-- If the response body exceeds 1 MiB, the callback receives truncated data
+- The `wasm_body` VDP streams response body chunks directly to
+  `proxy_on_response_body` as they arrive — no buffering
+- Each chunk is forwarded to the client immediately after inspection
+- `end_of_stream=1` is set on the final chunk
+- Memory usage is O(chunk_size), not O(body_size)
 
 ## Upgrading Modules
 

docs/SECURITY.md

@@ -58,7 +58,8 @@ following security considerations:
 - `proxy_get_header_map_value`: Read-only access to request/response headers
 - `proxy_add_header_map_value`: Can only add to the current request/response
 - `proxy_get_header_map_pairs`: Returns all headers including pseudo-headers
-- Pseudo-headers (`:method`, `:path`, `:authority`, `:status`) are read-only
+- Pseudo-headers (`:method`, `:path`, `:authority`) are writable via `proxy_set_property`
+- `:status` is read-only
 
 ### Body Access
 - `proxy_get_buffer_bytes`: Read-only access to request/response body

examples/proxy-wasm-filter/.cargo/config.toml

@@ -1,8 +1,2 @@
 [build]
 target = "wasm32-unknown-unknown"
-
-[source.crates-io]
-replace-with = "vendored-sources"
-
-[source.vendored-sources]
-directory = "vendor"

examples/rust/src/lib.rs

@@ -1,7 +1,7 @@
-/// Test Wasm module for vmod-wasm.
-///
-/// Phase 1: Simple exported functions (no host calls)
-/// Phase 2: Functions that call host functions to inspect requests
+// Test Wasm module for vmod-wasm.
+//
+// Phase 1: Simple exported functions (no host calls)
+// Phase 2: Functions that call host functions to inspect requests
 
 // --- Host function imports (provided by vmod-wasm) ---
 
@@ -284,7 +284,8 @@ extern "C" {
 }
 
 // Simple bump allocator for proxy_on_memory_allocate
-static mut ALLOC_BUF: [u8; 65536] = [0u8; 65536];
+const ALLOC_BUF_SIZE: usize = 65536;
+static mut ALLOC_BUF: [u8; ALLOC_BUF_SIZE] = [0u8; ALLOC_BUF_SIZE];
 static mut ALLOC_OFFSET: usize = 0;
 
 /// Proxy-Wasm memory allocator — host calls this to allocate space
@@ -293,12 +294,13 @@ static mut ALLOC_OFFSET: usize = 0;
 pub extern "C" fn proxy_on_memory_allocate(size: i32) -> i32 {
     unsafe {
         // Reset if we're running low on space
-        if ALLOC_OFFSET + (size as usize) > ALLOC_BUF.len() {
+        if ALLOC_OFFSET + (size as usize) > ALLOC_BUF_SIZE {
             ALLOC_OFFSET = 0;
         }
-        let ptr = ALLOC_BUF.as_mut_ptr().add(ALLOC_OFFSET) as i32;
+        let ptr = core::ptr::addr_of_mut!(ALLOC_BUF) as *mut u8;
+        let result = ptr.add(ALLOC_OFFSET) as i32;
         ALLOC_OFFSET += size as usize;
-        ptr
+        result
     }
 }