Unexpected panic/error. Am I doing something wrong?

[scripturial]

I am working through the example code trying to make my own test zig file to create/update a table. The create database works, and update works, but this "insert" function causes a bus error. I am at a loss to work out why. It certainly could be something I am doing wrong, but I am beginning to suspect it's not me. Is this the right way to do an insert?

pub fn add_user(name: []const u8, admin: bool) !u64 {
    const uid: u64 = rand.int(u64);
    //_ = admin;
    _ = name;
    var i = try db.prepare(
        \\insert into user (uid, name, admin) 
        \\values (?,?,?)
    );
    defer i.deinit();
    try i.exec(.{}, .{
        .uid = uid,
        .name = "bob",
        .admin = admin,
    });
    return uid;
}

If I comment out the i.exec() request it doesnt panic.

zig build run
run
└─ run mytest failure
error: the following command terminated unexpectedly:
/Volumes/mytest/zig/mytest/zig-out/bin/mytest
Build Summary: 7/9 steps succeeded; 1 failed (disable with --summary none)
run transitive failure
└─ run mytest failure
error: the following build command failed with exit code 1:

If I pass in the name parameter it panics like this:

Bus error at address 0x97a23f1ee39fa8e9
Panicked during a panic. Aborting.

[scripturial]

I am poking around example code and found this other code that calls insert slightly differently, this is different to what is in the main readme for the repo:

    try db.exec("INSERT INTO user(id, age, name) VALUES($id{usize}, $age{u32}, $name{[]const u8})", .{}, .{ @as(usize, 10), @as(u32, 34), user_name });

Is this the old way to do it? Or is the README the old way to do it?

[vrischmann]

Both are fine. This is a special "dialect" of SQL that is parsed at compile time by zig-sqlite that lets you embed type information for the bind parameters.

In your example, it says that:

• the $id bind parameter must be a usize
• the $age bind parameter must be a u32
• the $name bind parameter must be a []const u8

and if the types don't match it fails with an error like this:

/Users/vincent/.cache/zig/p/1220940ae067451e7e6824e9b92baceac93b0bd6fa9ffd315179cc9a7ce5430a46ac/sqlite.zig:2101:17: error: value type u64 is not the bind marker type usize
                @compileError("value type " ++ @typeName(Actual) ++ " is not the bind marker type " ++ @typeName(Expected));
                ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/Users/vincent/.cache/zig/p/1220940ae067451e7e6824e9b92baceac93b0bd6fa9ffd315179cc9a7ce5430a46ac/sqlite.zig:2088:58: note: called from here
                        else => comptime assertMarkerType(struct_field.type, typ),
                                         ~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~
``` 

But you don't have to use this, your first query above works fine too.

[vrischmann]

Can you share a reproducer ? Usually a bus error means you're accessing illegal memory, and in my experience this usually means you're accessing a stack variable after it has been freed (for example if you're taking a pointer to a stack variable and returning it from a function).

Without the code it's hard to say though.

I can also recommend running your program with gdb or lldb and inspect name to see if it is indeed valid when you're trying to use it.