[BUG]: Mojo variadics don't work with non-trivial register-only types.

[elhigu]

Bug description

Application crashes as soon as I try to do a.values()[0]

I would expect either working code or compile time error for using brackets in this case. Actually now that I think I might be confused with bracket getitem and bracket compile time parameters... so feel free to close if was complete user error.

Steps to reproduce

Pretty minimal example code included:

        (venv) mlepisto@stevetop:~/projects/learn-machine$ cat test.mojo 
        from python import Python 
        
        def main():
          var json = Python.import_module("json")
          a = json.loads('{"a": 1}')
          print(a.values())
          print(a.values()[0]) 
        (venv) mlepisto@stevetop:~/projects/learn-machine$ mojo run test.mojo 
        dict_values([1])
        [15228:15228:20240313,030428.656443:ERROR elf_dynamic_array_reader.h:64] tag not found
        [15228:15228:20240313,030428.658258:ERROR file_io_posix.cc:144] open /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq: No such file or directory (2)
        [15228:15228:20240313,030428.658285:ERROR file_io_posix.cc:144] open /sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq: No such file or directory (2)
        Please submit a bug report to https://github.com/modularml/mojo/issues and include the crash backtrace along with all the relevant source codes.
        Stack dump:
        0.      Program arguments: mojo run test.mojo
        Stack dump without symbol names (ensure you have llvm-symbolizer in your PATH or set the environment var `LLVM_SYMBOLIZER_PATH` to point to it):
        0  mojo                0x000055b953a9fe47
        1  mojo                0x000055b953a9da1e
        2  mojo                0x000055b953aa051f
        3  libc.so.6           0x00007fafbf6f5520
        4  libpython3.9.so.1.0 0x00007fafa0333e7a _PyObject_Call + 26
        5  libpython3.9.so.1.0 0x00007faf2004b34a _PyObject_Call + 18446744071559017706
        Segmentation fault
        (venv) mlepisto@stevetop:~/projects/learn-machine$ mojo --version
        mojo 24.1.0 (55ec12d6)
        (venv) mlepisto@stevetop:~/projects/learn-machine$

System information

Running in WSL2 (ubuntu22)
mojo 24.1.0 (55ec12d6)
modular 0.5.1 (1b608e3d)

[elhigu]

Tried also iterating over that, which fails to iterator not being mutable or something like that.

(venv) mlepisto@stevetop:~/projects/learn-machine$ cat test.mojo 
from python import Python

fn main() raises:
    var json = Python.import_module("json")
    var a = json.loads('{"a": 1, "b": 2}')
    print(a.values())
    for v in a.values():
        print(v)
(venv) mlepisto@stevetop:~/projects/learn-machine$ mojo run test.mojo 
/home/mlepisto/projects/learn-machine/test.mojo:8:22: error: invalid call to '__iter__': argument #0 must be mutable in order to pass as a by-ref argument
    for v in a.values():
             ~~~~~~~~^~
/home/mlepisto/projects/learn-machine/test.mojo:1:1: note: function declared here
from python import Python
^
mojo: error: failed to parse the provided Mojo

[JoeLoser]

Your second example works fine on top-of-tree and will work in 24.2:

from python import Python

fn main() raises:
    var json = Python.import_module("json")
    var a = json.loads('{"a": 1, "b": 2}')
    print(a.values())
    for v in a.values():
        print(v)

It prints:

dict_values([1, 2])
1
2

Your first example:

from python import Python 
        
def main():
    var json = Python.import_module("json")
    a = json.loads('{"a": 1}')
    print(a.values())
    print(a.values()[0]) 

crashes on the subscript operator as you mentioned. We should fix that to give a better error message. The print(a.values()) prints:

dict_values([1])

as expected

[lattner]

@akirchhoff-modular helped track this down - this is caused by register-only types being handled by VariadicList instead of VariadicListMem, so their copy constructors don't get run correctly. There is a big migration to move the variadic list representation over to this new form, which can hopefully help, but I'm not exactly sure when it will converge

[lattner]

This is fixed in the next update, thank you for reporting this!