Replace the non-portable PATH_MAX constant with a new WREN_PATH_MAX define set to 4096 in os.h, and correct the void function processCwd to use bare return instead of return wrenAbortFiber(vm, 0) which was causing a broken build on platforms without PATH_MAX defined.
- Fix premake invocation by changing `--os=` flag syntax from positional argument to named parameter
- Remove dependency on `command -v` for binary detection, making script portable across platforms
- Add platform-aware default binary name (appending `.exe` on Windows)
- Implement multi-step binary resolution: check `WREN_PREMAKE` env var, then local `projects/premake` directory, then system PATH
- Add graceful error handling with user-friendly guidance when premake binary cannot be located
The build status badge was previously placed at the bottom of the file, after the build instructions section. This commit relocates it to the top of the README, directly below the contribution link, for better visibility and standard project convention. The badge URL is updated to target the master branch explicitly.
- Inserted a new changelog line under the CLI section documenting the addition of prebuilt binaries alongside existing project generation and vendoring notes.
- Fix broken markdown link format for libuv reference by removing square brackets
- Add trailing space after wren.io/cli link for consistency
- Rephrase contribution invitation from "Like with Wren" to "Like with Wren itself" for clarity
The .travis.sh script was referencing "projects/gmake" and "projects/gmake.mac" as working directories, but the actual build system uses "projects/make" and "projects/make.mac". This fix updates both the default and macOS-specific paths to match the correct directory naming convention used by the project's make-based build system.
Add `// skip:` annotation to read_byte_eof.wren, read_line.wren, and read_line_eof.wren
to prevent test failures on non-linux platforms until the underlying issue is fixed.
Replace the broken `64bit-no-nan-tagging_64bit` and `64bit-no-nan-tagging_32bit` configurations with properly named `release_64bit-no-nan-tagging` and `debug_64bit-no-nan-tagging` across all gmake Makefiles and wren_cli.make files. Add the missing `release_64bit-no-nan-tagging` build target definitions for linux, mac, and bsd platforms. Introduce `.travis.sh` script that runs both nan-tagged and non-nan-tagged debug and release builds, and simplify `.travis.yml` to use the new script with per-os env variables instead of matrix options.
Introduce new build configuration "64bit-no-nan-tagging" for both 64-bit and 32-bit platforms, defining WREN_NAN_TAGGING=0 to disable NaN tagging in the Wren CLI. Update Makefiles, premake5.lua, and Visual Studio solution/project files to support the new config with appropriate compiler flags, linker settings, and platform-specific defines (e.g., _GNU_SOURCE on Linux, _DARWIN_* on macOS).
Add premake filter for system:macosx to generate Makefiles into projects/gmake.mac directory, and create the initial gmake.mac Makefile and wren_cli.make with support for release/debug 64/32-bit configurations using clang on Darwin.
On Windows, `stat` on a directory reports size as 0, causing the previous
`stat.size > 0` assertion to fail. Change the test to only verify that
`stat.size` is a numeric value, accommodating platform-specific behavior
where directory sizes may be zero.
Add new GNU Make workspace and project makefiles under projects/gmake.bsd/ directory, supporting release_64bit, release_32bit, debug_64bit, and debug_32bit configurations for the wren_cli target. Update premake5.lua to route BSD system builds to the new gmake.bsd location instead of the default gmake directory.
The diff updates the wren_cli.vcxproj file by replacing the PlatformToolset from v142 to v141 across all four active build configurations (Release 64bit, Release 32bit, Debug 64bit, Debug 32bit). Additionally, it removes four orphaned PropertyGroup entries that referenced mismatched platform/architecture combinations (e.g., Debug 32bit|x64, Debug 64bit|Win32, Release 64bit|Win32, Release 32bit|x64) which were left over from a previous toolset migration and no longer correspond to valid build targets.
Add new projects/vs2017 directory containing wren-cli.sln solution file,
wren_cli.vcxproj project file with Debug/Release 32/64-bit configurations,
and wren_cli.vcxproj.filters file organizing source files into deps/libuv,
deps/wren, and src filter categories for the VS2017 build system.
The commit adds two trailing spaces before line breaks in the Linux and FreeBSD sections of the README.md file, specifically after the sentences describing the `projects/gmake` folder contents. This ensures proper Markdown line-break rendering for the subsequent `From that folder` instructions.
Add the Xcode 4 project configuration under `projects/xcode4/`, including the workspace file `wren-cli.xcworkspace` and the full `wren_cli.xcodeproj` with 641 lines of build settings. The project integrates all wren source files (core.c, wren_vm.c, wren_compiler.c, wren_value.c, wren_primitive.c, wren_opt_random.c) and platform-specific modules (darwin.c, bsd-ifaddrs.c, dl.c, io.c, os.c, process.c) into a single CLI target, enabling native builds and testing on macOS via Xcode.
- Inserted a new top-level heading and introductory paragraph describing the Wren CLI as a REPL/CLI tool backed by libuv
- Moved the original "Wren is a small, fast..." description below the new CLI section
- Updated the readme structure to clarify the CLI project's purpose and dependencies
The entire deps/libuv directory is deleted, including its .gitignore,
.mailmap, AUTHORS, CONTRIBUTING.md, ChangeLog, LICENSE, and
MAINTAINERS.md files, as part of upgrading to a separately managed
libuv dependency.
Add the complete Wren 0.3.0 source tree under deps/wren/, including the public C/C++ headers, the core VM implementation, and optional modules for meta-programming and random number generation. This provides an embedded scripting language with WELL512a PRNG, runtime compilation, and module introspection capabilities.
Add initial build system configuration for the wren-cli command-line tool,
including a premake5.lua script defining workspace, project, and dependency
paths for libuv and wren, plus generated Visual Studio 2019 solution and
project files with Debug/Release and 32bit/64bit platform support.
Delete all VM-related test files including benchmark, call, error, foreign_class, and other API tests, along with their C source, headers, and Wren scripts. Rewrite test/README.md to describe only CLI unit tests and module resolution tests, removing references to VM, compiler, core library, and language semantics tests.
The util/benchmark.py runner and all Visual Studio 2017 solution/project files under util/vs2017/ are deleted, cleaning up legacy build and benchmarking infrastructure no longer needed for the 0.3.0 release.
When a baseline file contains "None" as the best time (e.g. when no interpreter
was available during baseline generation), the previous code attempted to parse
it as float and failed. This change checks if the value starts with "None" and
assigns 0.0 instead, allowing benchmarks without a valid baseline to proceed
without crashing.
The small navigation layout was using a single table row with two cells, which prevented proper vertical stacking and omitted the blog link. Changed the row to three div elements, each containing a link, and added the missing "Blog" entry to match the full navigation.
- Update 0.2.0-and-beyond.markdown to use <s>...</s> HTML elements instead of ~~...~~ markdown syntax for completed checklist items
- Refine phrasing of two sentences in the issue management section for improved clarity and flow
Create new blog entry detailing the 0.2.0 release, including the closure of 70+ issues, consolidation of the issue tracker, and plans for the upcoming 0.3.0 release. Update the blog index page with a link to the new post and add the corresponding RSS feed entry with publication date and description.
Update WREN_VERSION_MINOR from 1 to 2 and WREN_VERSION_STRING from "0.1.0" to "0.2.0" in wren.h header. Add comprehensive 0.2.0 changelog entry documenting new language features (relative imports, Fiber value passing, String/Sequence/List/Num methods, Map iteration), C API additions (wren.hpp, user data, wrenAbortFiber), and CLI improvements (path module, --version flag, REPL, Platform class, process→os rename).
Add a new error code (-3) from wrenDefineVariable to detect when a lowercase (local-style) module variable is referenced before its definition. The compiler now reports the line of the first use, improving diagnostics for forward-declared top-level variables.
Fix two documentation errors: update incorrect function name from `greet` to `callMe` in the functions example, and fix the expected count value from 4 to 5 in the lists example after appending an item.
The `printError` function in `wren_compiler.c` previously dereferenced
`parser->module->name->value` without checking if the module name is null,
causing a crash when compiling modules without assigned names. Now safely
falls back to the string `"<unknown>"` when the module name pointer is null.
Add bounds check in forStatement to ensure compiler->numLocals + 2 does not exceed MAX_LOCALS before calling addLocal for the hidden 'seq ' and iterator variables. Includes regression test with deeply nested for-loops that previously caused an array overflow.
When parsing hex numeric literals, the return value of strtoll (__int64) is
implicitly converted to double via the NUM_VAL macro. This adds an explicit
cast to double to silence the MSVC warning C4244 about potential loss of data
from a 64-bit integer to a double.
The floating-point remainder operator % now explicitly documents that the
returned value retains the sign of the dividend (this), matching the
behavior of C's fmod function. This addresses issue #695 by removing
ambiguity about sign handling in the modulus documentation.
The documentation for class encapsulation is updated to remove the comparison to C++ and Java from the initial description, add bold formatting to key getter/setter explanations, and include a new code example demonstrating that fields are private to instances and inaccessible from subclasses or other instances of the same class.
Add runtime error check after foreign object allocation in the interpreter loop to properly propagate fiber aborts triggered inside foreign class allocators. Previously, calling wrenAbortFiber from an allocate callback would set the fiber error state but the interpreter would continue execution, ignoring the abort. The fix adds a wrenHasError check after createForeign call, jumping to the runtime error handler if an error was raised.
Also adds a test case with a BadClass foreign class whose allocator calls wrenAbortFiber, verifying the fiber correctly catches and returns the error message "Something went wrong".
- Remove i386 from ARCHS in build_libuv.py to stop building 32-bit universal binary
- Drop ci_32 from CI_ARCHS in .travis.yml to skip 32-bit CI runs
- Add XCBuildData cache directory to .gitignore for macOS Xcode build artifacts
Correct "in a local projects" to "in local projects", add missing comma after "2.5+ years", and adjust parenthetical phrasing for clarity in the second paragraph of the Wren introduction blog.
The markdown.markdown() call was passing extensions as a positional argument,
which is deprecated in newer versions of the Python-Markdown library. This
change wraps the extensions list in the explicit 'extensions' keyword argument
to ensure compatibility with the current API and suppress deprecation warnings
during documentation generation.
The link target for "runtime error" reference was pointing to a flat file path
`error-handling.html` which would fail when the page is served from a subdirectory.
Changed to `../error-handling.html` to properly resolve relative to the current
`embedding/` directory path.
Replace all outdated reference links in README.md that pointed to the deprecated
munificent.github.io/wren domain with their current equivalents on wren.io,
covering syntax, performance, classes, concurrency, embedding, getting-started,
and contributing pages.
Implement the `log` method on Num by adding a DEF_NUM_FN(log, log) macro invocation in wren_core.c, registering the corresponding primitive `num_log` on the numClass, and creating a new test file test/core/number/log.wren that validates positive and negative input behavior.
Implement the `num_pow` primitive that wraps C's `pow()` function, enabling
exponentiation on Num instances via the `pow(_)` method. Register the new
primitive in `wrenInitializeCore` and add a test file verifying basic cases:
2^4=16, 2^10=1024, and 1^0=1.
Add a new primitive `string_indexOf_with_startIndex` that accepts a second integer argument to specify the starting position for the search. Modify the existing `wrenStringFind` function to accept a `startIndex` parameter, updating its Boyer-Moore-Horspool algorithm to begin scanning from the given offset. Update the `string_contains` primitive to pass `0` as the start index. Register the new overloaded method as `indexOf(_,_)` on the string class and add test cases covering boundary conditions and out-of-range start indices.
The previous link pointed to a non-existent ../range.html page. The corrected link now points to the proper anchor #ranges within the values documentation page, ensuring readers can access the friendly introduction as intended.
underscorediscovery