34.0.0
Released 2025-06-20.
Added
-
Support for SIMD in the Pulley interpreter can now be disabled at compile-time to shrink the size of the final binary. #10727
-
The C API now has
wasmtime_trap_new_code
to create awasm_trap_t
from its code. #10765 -
Winch's support for x86_64 is now classified with tier 1 support in Wasmtime. #10755
-
Winch's support for aarch64 now implements stack checks to pass many more spec tests. #10763
-
Cranelift's s390x backend now has full support for the
f128
type. #10774 -
Wasmtime's C API for the component model has initial support for calling functions. #10697 #10841 #10858 #10864 #10877
-
The
wasmtime wast
command now has a--generate-dwarf
flag to show filename/line number information for backtraces. #10780
Changed
-
The shape of
bindgen!
-generatedadd_to_linker
functions has changed with the removal ofGetHost
and replacement of aHasData
trait. For more information see the associated PR. #10770 -
Wasmtime's
Store<T>
now requires thatT: 'static
. This is done in preparation for merging WASIp3 work to the main repository with some more information on the associated PR. #10760 -
The
wasmtime::component::Instance::instance_pre
method is now public. #10761 -
Wasmtime and Cranelift's minimnum supported version of Rust (MSRV) is now 1.85.0. #10785
-
Cranelift's
debugtrap
on aarch64 now generatesbrk #0xf000
for debuggers to recognize it. #10813 -
The wasi-http implementation no longer generates a trap if the handle to receive the response on the host is dropped early. #10833
-
The
wasmtime serve
command will now send some boilerplate descriptive HTML on a 500 server error instead of nothing. #10851 -
A significant amount of work has gone into the new assembler for the x64 backend. Too many PRs to list here but progress continues apace at defining all machine instructions in a standalone crate.
-
Cranelift will now reject unimplemented big-endian loads/stores on backends that do not implement this functionality. #10863
-
The
wasmtime explore
generated HTML handles large modules better now. #10892 -
Wasmtime's internal representation of
wasmtime::Func
has changed and a previous optimization ofFunc::call
has been lost. If affected it'd recommended to useFunc::call_unchecked
instead or to open an issue. #10897