ICE: wrong-structured lhs for follow check with macro

[Ms2ger]

Fiddling with some code from JeanSprouts on IRC,

fn main() {
    macro_rules! dispatch(

                    $($id:expr, $name:ident, $parser:expr, $conda:ident, $condb:expr),+
             => {
                    general_dispatch!(
                            id,
                            start,
                            size,

                            $(
                            ($id, $name, {
                                    header.$name = match $parser.0 {
                                            $conda if $condb => $conda,
                                            _ => {
                                                    return Err(ParserError::Incompatible);
                                            },
                                    }
                            }),
                            )+
                    )
            }
    );

    dispatch!(
            (0x4286, EBMLVersion,       try!(self.parse_uint(size)), n, n == 1),
            (0x42F7, EBMLReadVersion,   try!(self.parse_uint(size)), n, n == 1),
            (0x42F2, EBMLMaxIDLength,   try!(self.parse_uint(size)), n, n == 4),
            (0x42F3, EBMLMaxSizeLength, try!(self.parse_uint(size)), n, n == 8),
            (0x4282, DocType,              try!(self.parse_string(size)), s, s == "matroska"),
            (0x4287, DocTypeVersion,       try!(self.parse_uint(size)), n, n <= 2),
            (0x4285, DocTypeReadVersion,   try!(self.parse_uint(size)), n, n <= 2)
    );


}

failed on playpen with:

<anon>:2:5: 23:7 error: internal compiler error: wrong-structured lhs for follow check
<anon>:2     macro_rules! dispatch(
<anon>:3             
<anon>:4                     $($id:expr, $name:ident, $parser:expr, $conda:ident, $condb:expr),+
<anon>:5              => {
<anon>:6                     general_dispatch!(
<anon>:7                             id,
         ...
note: the compiler unexpectedly panicked. this is a bug.

[jdm]

Presumably we're getting a TtSequence here. @cmr is this just something we should have a better error for?

[emberian]

Nope, this isn't a legitimate error. I just didn't find a case where this could happen, so I didn't handle it. I'll fix it.

[emberian]

(A workaround is to wrap the lhs in parens)