Zig-recover - Regain control when a function panics

Showcase
1
9 Likes
2

Proper use of recover is only for testing panic and undefined behavior runtime safety checks.

Do you have an example of how you do this?

2 Likes
3

Let’s say that foo(0) is designed to panic.

To test that the function panics you can:

test "foo(0) panics" {
     const err = recover.call(foo, .{0}, error.Panic);
     std.testing.expectError(error.Panic, err).
}

For this to work, you need a custom test_runner with a panic handler:

pub fn panic(msg: []const u8, error_return_trace: ?*std.builtin.StackTrace, ret_addr: ?usize) noreturn {
    recover.panicked();
    std.builtin.default_panic(msg, error_return_trace, ret_addr);
}
4 Likes
4

I see. Thanks, that’s a nice use case.

5

zig-recover version 1.0.0 released for zig 0.12.0 or later

To fetch as a dependency in your build.zig.zon:

zig fetch --save https://github.com/dimdin/zig-recover/archive/refs/tags/1.0.0.tar.gz

Usage is simplified:

const recover = @import("recover");
pub const panic = recover.panic;

fn function() void {
    unreachable;
}

pub fn main() !void {
    recover.call(function, .{}) catch {};
    // flow continues here because recover.call recovers
    // from unreachable and returns error.Panic
}

See the project README.md for more information.

5 Likes