Segfault matching copy of dereferenced owned string

[jdm]

fn main() {
    for os::args().each |arg| {
        match copy *arg {
            s => { }
        }
    }
}

#0  0x000000395a835935 in raise () from /lib64/libc.so.6
#1  0x000000395a8370e8 in abort () from /lib64/libc.so.6
#2  0x000000395a874e8b in __libc_message () from /lib64/libc.so.6
#3  0x000000395a87c00e in _int_free () from /lib64/libc.so.6
#4  0x00007ffff7dae669 in __morestack () from /tmp/../usr/local/lib/rustc/x86_64-unknown-linux-gnu/lib/librustrt.so
#5  0x00007ffff7d9dc68 in call_on_c_stack (fn_ptr=0x7ffff7d9cb70, args=0x7ffff0101650, this=0x407180) at /run/media/jdm/ssd/rust/src/rt/rust_task.h:494
#6  call_upcall_on_c_stack (fn_ptr=0x7ffff7d9cb70, args=0x7ffff0101650, task=<optimized out>) at /run/media/jdm/ssd/rust/src/rt/rust_upcall.cpp:45
#7  upcall_exchange_free (ptr=<optimized out>) at /run/media/jdm/ssd/rust/src/rt/rust_upcall.cpp:208
#8  0x0000000000401d7f in glue_free_1939 ()
#9  0x00000000004021c7 in main::anon::expr_fn_1955 ()
#10 0x000000000040175b in vec::__extensions__::each_1861::anon::expr_fn_1914 ()
#11 0x0000000000401472 in vec::each_1864::anon::expr_fn_1894 ()
#12 0x00000000004011a2 in vec::as_imm_buf_1867::_43d6e68cf6296159::_00 ()
#13 0x00000000004022b4 in __morestack ()
#14 0x0000000000401087 in vec::each_1864::_83f0fe0dda0fc51::_00 ()
#15 0x0000000000400fe5 in vec::__extensions__::each_1861::_a7cec32bfdc93c6::_00 ()
#16 0x0000000000401f83 in main::_fe18229bb2f275::_00 ()
#17 0x000000000040203e in _rust_main ()
#18 0x00007ffff7d9b4e4 in task_start_wrapper (a=0x407fd0) at /run/media/jdm/ssd/rust/src/rt/rust_task.cpp:164
#19 0x0000000000000000 in ?? ()

[catamorphism]

This simpler example pokes the same bug:

fn main() {
    let a = &~1;
    match copy *a {
        b => ()
    }
}

[catamorphism]

This also segfaults, and prints what appears to be garbage:

fn main() {
    let a = &~1;
    let mut c: ~int;
    match copy *a {
        b => c = b
    };
    log(error, *c);
}

I think borrowck should reject this, since the copy of *a copies a non-copyable value.

[nikomatsakis]

@catamorphism What is noncopyable about ~1? I think this is basically valid.

[catamorphism]

@nikomatsakis You're right; I'm coming up with another example.

[nikomatsakis]

@catamorphism ok. btw, I am working on the move code right now, which is close to this area. it's possible I could look into this bug afterward, I expect the problem is centered around how match handles freeing memory.

[catamorphism]

@nikomatsakis I reassigned this and #4541 to you (that one looks pretty similar to this) -- if you end up not wanting to look into it, feel free to assign back to me once you're done with the move code :-)

[catamorphism]

Reproduced with 373504a. Seems at least mildly important.

[nikomatsakis]

Not critical for 0.6; removing milestone.