cgen: array assertion fails on MacBook Pro M2

[gchumillas]

Describe the bug

The following code does not terminate properly on an MacBook Pro M2 (and M1 too):

// main_test.v
fn foo(i int) !int {
    m := [3, 4, 5]

    // This works
    // r := m[i] or {
    //     return error('bad boy!')
    // }

    // this does not work
    r := m[i]!

    return r
}

fn test_foo() {
    g := foo(3) or { 12 }
    assert g == 12
}

Test the file:

v test main_test.v

See the output:

fn test_foo
   > assert g == 12
     Left value (len: 7): `5119952` (this is a random value)
    Right value (len: 2): `12`

Reproduction Steps

Buy a Macbook Pro or ask a friend who already has one for a favor and run the code above.

Expected Behavior

It should complain.

Current Behavior

It doesn't complain.

Possible Solution

No response

Additional Information/Context

No response

V version

V 0.4.10 6be298d

Environment details (OS name and version, etc.)

V full version	V 0.4.10 6be298d
OS	macos, macOS, 15.5, 24F74
Processor	10 cpus, 64bit, little endian, Apple M4
Memory	0.14GB/16GB
V executable	/Users/gonzalo/Documents/Projects/Personal/v/v
V last modified time	2025-05-20 11:21:19
V home dir	OK, value: /Users/gonzalo/Documents/Projects/Personal/v
VMODULES	OK, value: /Users/gonzalo/.vmodules
VTMP	OK, value: /tmp/v_501
Current working dir	OK, value: /Users/gonzalo/Documents/Projects/Personal/v
Git version	git version 2.49.0
V git status	6be298d
.git/config present	true
cc version	Apple clang version 17.0.0 (clang-1700.0.13.3)
gcc version	Apple clang version 17.0.0 (clang-1700.0.13.3)
clang version	Apple clang version 17.0.0 (clang-1700.0.13.3)
tcc version	tcc version 0.9.28rc 2024-02-05 HEAD@105d70f7 (AArch64 Darwin)
tcc git status	thirdparty-macos-arm64 e447816c
emcc version	N/A
glibc version	N/A

Note

You can use the 👍 reaction to increase the issue's priority for developers.

Please note that only the 👍 reaction to the issue itself counts as a vote.
Other reactions and those to comments will not be taken into account.

[huly-for-github]

Connected to Huly®: V_0.6-22897

[gchumillas]

It only fails on test files. For example:

// v run app.v OK!
fn foo(i int) !int {
    m := [3, 4, 5]
    return m[i]!
}

fn main() {
    g := foo(3) or { 12 }
    assert g == 12
}

// v app_test.v
//
// Output
// app_test.v:9: fn test_main
//    > assert g == 12
//      Left value (len: 8): `14917584` (random value)
//     Right value (len: 2): `12`
fn foo(i int) !int {
    m := [3, 4, 5]
    return m[i]!
}

fn test_main() {
    g := foo(3) or { 12 }
    assert g == 12
}

[gchumillas]

Probably the problem is here: https://github.com/vlang/v/blob/master/vlib/v/gen/c/cgen.v#L7090

// maybe sizeof is incorrect
g.writeln2('\t${styp} ${err_obj};', '\tmemcpy(&${err_obj}, &${cvar_name}, sizeof(${result_name}));')

According to ChatGPT:

The problem is that `sizeof(_result)` (or `sizeof(_option)`) is **not sufficient** to copy the entire structure of a specific type like `_result_int` or `_option_int`.
Both types (`_result_int` and `_option_int`) have additional fields (such as the `data` array) whose size depends on the specific type (`int`, `string`, etc.).
Copying only the base size results in the `data` field of the destination structure being garbage or zeroed out, which is why the returned value is not what was expected.

[felipensp]

Can you try PR #24564 ?