Devlog ⚡ Zig Programming Language

This page contains a curated list of recent changes to main branch Zig. This page contains entries for the year 2026. Other years are available in the Devlog archive page. June 30, 2026All Package Management Functionality Moved from Compiler to Build System Author: Andrew KelleyNow that there is a separate process for users’ build.zig scripts and the build system itself, it makes sense for that to be the place that package management logic lives.I moved these subcommands to the maker process:zig buildzig fetchzig initzig libcThis means that large parts of what used to be included in the compiler executable are now shipped in source form instead, including:package fetching logicHTTP client and networkingTLS (Transport Layer Security) and associated cryptoGit protocolxz, gzip, zstd, flate, zipparsing, validation, and otherwise dealing with build.zig.zon filesConsequently, this functionality can now be patched without rebuilding the compiler, making it easier for users and contributors to tinker.Furthermore, it means that package management in zig now has safety checks enabled when doing networking, since the maker executable is compiled in ReleaseSafe mode. Plus, all the crypto used for networking and file hashing can now take advantage of special CPU instructions available on the host, even the ones that are too rare to normally depend on when distributing software. We can have AOT cake and eat JIT, too!My original motivation for doing this was in relation to exposing a build server protocol in order to unblock ZLS after maker/configurer process separation made breaking changes to the --build-runner override flag.Originally, the process tree looked like this:zig build (the zig compiler + package manager) └─ builder (the user's build.zig logic + build system implementation) The process separation changeset made it look like this instead:zig build (the zig compiler + package manager) ├─ configurer (the user's build.zig logic) └─ maker (build system) At this point, consider a long-running zig build --watch process, watching files and rebuilding on source code changes. If any changes to build.zig are detected, or any files observed during execution of that logic, it means configurer needs to be rerun, meaning that maker process must exit to give zig build a chance to repeat the package management logic.

Now, after the changes described in this devlog entry, it looks like this:zig build (the zig compiler) └─ maker (build system + package manager) └─ configurer (the user's build.zig logic) Thus, when configuration needs to be rerun, maker process can continue to live because it is the parent process rather than a sibling. In terms of the upcoming build server, it means avoiding an awkward situation where the server has to exit and the client has to reconnect, rather than simply informing the client of a configuration change.This is almost entirely a non-breaking change, but there are some observable differences:Zig executable binary size: shrinks 4% from 14.1 to 13.5 MiB (no LLVM, ReleaseSmall)--maker-opt flag is replaced by ZIG_DEBUG_MAKER environment variable--zig-lib-dir flag is replaced by ZIG_LIB_DIR environment variableThe follow-up issues to this changeset are the main blockers until we tag Zig 0.17.0:build server protocol MVP (needed to unblock ZLS)introduce the concept of adding path dependencies of the build script itselfmake zig build --watch detect modifications to the build script and rerun itselfdifferent cwd causes build script cache missI have two conferences coming up in July and I need to work on my talks, so being realistic, I don’t think I will have time to wrap these up until early August. Contributions welcome, of course.Big thanks to Techatrix from the ZLS team for reaching out and working with me on the build server protocol! They are seeking sponsorship, by the way. June 26, 2026SPIR-V Backend Progress Author: Ali CheraghiThere’s quite a bit to cover. The SPIR-V backend had bitrotted in a number of places after the recent compiler changes, so I spent the past several weeks dragging it into a better state.@SpirvTypeSPIR-V has a handful of types that couldn’t be expressed in Zig’s type system. The new @SpirvType builtin has been introduced to address the longest-standing blocker for writing shaders. See #20550, #23326 and #35461 to trace the background.const Sampler = @SpirvType(.sampler); const Image = @SpirvType(.{

.image = .{ .usage = .{ .sampled = u32 }, .format = .unknown, .dim = .@"2d", .depth = .unknown, .arrayed = false, .multisampled = false, .access = .unknown, } }); const SampledImage = @SpirvType(.{ .sampled_image = Image }); const RuntimeArray = @SpirvType(.{ .runtime_array = u32 }); const sampled_image = @extern(*addrspace(.constant) const SampledImage, .{ .name = "sampled_image", .decoration = .{ .descriptor = .{ .set = 0, .binding = 1 } }, }); Execution Mode on the Calling ConventionExecution mode info (workgroup size, fragment origin, etc.) is now carried by the calling convention instead of being emitted via inline assembly OpExecutionMode. The old std.gpu.executionMode() helper is gone, and the SPIR-V assembler now rejects manual OpExecutionMode instructions. Two new calling conventions, spirv_task and spirv_mesh, were also added for mesh shading pipelines.export fn vert() callconv(.spirv_vertex) void {} export fn frag() callconv(.{ .spirv_fragment = .{ .depth_assumption = .greater } }) void {} export fn comp() callconv(.{ .spirv_kernel = .{ .x = 8, .y = 8, .z = 1 } }) void {} export fn task() callconv(.{ .spirv_task = .{ .x = 1, .y = 1, .z = 1 } }) void {} export fn mesh() callconv(.{ .spirv_mesh = .{ .stage_output = .output_lines, .max_primitives = 1, .max_vertices = 2 } }) void {} Capabilities and Extensions from CPU FeaturesCapabilities and extensions used to be emitted ad hoc by codegen or via inline assembly. They’re now driven entirely by the CPU feature set like other targets, with dependency chains extracted from SPIRV-Headers (excluding external vendors for now), and the assembler now rejects any attempt to emit OpCapability or OpExtension directly.Multi-Threaded CodegenFrom day one, the SPIR-V backend ran codegen single-threaded inside the linker thread.

Each codegen job now produces an Mir value just like every other self-hosted backend, and gets scheduled on the compiler’s thread pool.The same change brought back two ISel passes that had been removed during earlier refactors: dedup_types (which merges equivalent type instructions) and prune_unused (which strips dead code from the final module). These had originally been deleted back when codegen was single-threaded.Object File Linking.spv files are now recognised as object files. You can compile multiple .zig files (or external .spv objects) and have the SPIR-V linker stitch them into a single module.Tens of bugs have also been fixed along the way with a nearly 10% increase in total passing behavior tests (49% now) on the spirv64-vulkan target, std.gpu was renamed to std.spirv and the SPIR-V backend is meaningfully more useful than it was a month ago, but there’s still a long way to go. Plenty of behavior tests remain skipped on SPIR-V. That said, if you’ve been on the fence about trying Zig for shaders or compute kernels, this is a good time to give it a shot. Bug reports are very welcome on Codeberg. Happy hacking! June 25, 2026New @bitCast Semantics and LLVM Backend Improvements Author: Matthew Lugg(Quite long devlog coming up, apologies—I got a little carried away with this one!)A few weeks ago, I began working on a branch implementing an improvement to the LLVM backend which had been planned for a long time. This ended up snowballing into a bigger change which implemented a few language proposals you might be interested to hear about.LLVM Backend Integer LoweringZig has always lowered arbitrary bit-width integer types (e.g. u4, i13, u40) directly to LLVM IR’s bit-int types (i4, i13, i40). However, we’ve known for a long time that this lowering is not optimal, because LLVM’s documented semantics for representing these types in memory are unnecessarily restrictive to the optimizer.

Perhaps more importantly, because Clang never emits LLVM IR like this, these code paths in LLVM have never been properly tested, and so are poorly supported in practice—over the past few years, we have observed many instances of trivial optimizations being missed and even straight-up miscompilations.So, the original goal of the PR was to only use these bit-int types when manipulating values in SSA form, and to zero- or sign-extend them to ABI-sized types (i8, i16, i32, etc) when storing them in memory. This should be well-supported, not least because it matches how Clang lowers C’s _BitInt(N)!That change was actually fairly straightforward, but I hit one issue which led me down a bit of a rabbit-hole.The Problem with @bitCast@bitCast is an interesting builtin. In the past, it was defined as being equivalent to the following sequence of operations:Take a pointer to the operand valueCast it to a pointer to the destination typeLoad from that pointerIn other words, it was essentially syntax sugar for reinterpreting bytes of memory. However, over time, we diverged from this definition—for instance, it became allowed to use @bitCast to reinterpret a [3]u8 as a u24, even though on most targets @sizeOf(u24) is greater than @sizeOf([3]u8) so the above definition would invoke Illegal Behavior.Up to now, the LLVM backend had implemented these underspecified semantics for the @bitCast builtin. However, because that definition involved reinterpreting memory, changing how we store integer types in memory ended up impacting the implementation of @bitCast, and introducing Illegal Behavior which led to crashes in the compiler test suite.The easiest solution to this would probably have been to implement logic in the LLVM backend to approximately match the old behavior. I instead opted for a better solution—implement a new definition of @bitCast.Redefining @bitCastIn 2024, Jacob Young wrote up language proposal #19755 which aimed to solve the problems with @bitCast by precisely specifying a new set of semantics for it. This proposal was accepted shortly after it was submitted, and in fact, the semantics it details are already implemented by the self-hosted x86_64 backend!