assertion failures in quasi-quoting

[nikomatsakis]

The following test fails with assertion failures about the "gather" ordering not being properly setup:

use rustc;
import rustc::driver::diagnostic;
import rustc::syntax::ast;
import rustc::syntax::codemap;
import rustc::syntax::print::pprust;

fn new_parse_sess() -> parser::parse_sess {
    let cm = codemap::new_codemap();
    let handler = diagnostic::mk_handler(option::none);
    let sess = @{
        cm: cm,
        mutable next_id: 1,
        span_diagnostic: diagnostic::mk_span_handler(handler, cm),
        mutable chpos: 0u,
        mutable byte_pos: 0u
    };
    ret sess;
}

iface fake_ext_ctxt {
    fn session() -> fake_session;
}

type fake_options = {cfg: ast::crate_cfg};

type fake_session = {opts: @fake_options,
                     parse_sess: parser::parse_sess};

impl of fake_ext_ctxt for fake_session {
    fn session() -> fake_session {self}
}

fn mk_ctxt() -> fake_ext_ctxt {
    let opts : fake_options = {cfg: []};
    {opts: @opts, parse_sess: new_parse_sess()} as fake_ext_ctxt
}


fn main() {
    let ext_cx = mk_ctxt();
    let s = #ast(expr){__s};
    let e = #ast(expr){__e};
    let f = #ast(expr){$(s).foo {|__e| $(e)}}; // THIS IS THE PROBLEM EXPR
    pprust::print_expr(f);
}

[nikomatsakis]

@kevina for some reason I cannot assign you, but I think you'd be interested.

[kevina]

I try to look into this in the next few days. It might be as easy as reordering the gather list, but at this point I'm not sure.

The string slicing approach is inherently fragile, in the long term I think I want to switch a serialized AST in some fashion now that we has a working serializer.

[nikomatsakis]

My workaround was to construct the expr call myself, so:

let f = #ast(expr){$(s).foo {|__e| $(e)}}; // THIS IS THE PROBLEM EXPR

becomes

            cx.expr(
                ty.span,
                ast::expr_call(
                    #ast(expr){$(s).emit_from_vec},
                    [#ast(expr){{|__e| $(ser_e)}}],
                    false)))]

where cx.expr is a helper method I wrote that generates an @ast::expr given a span and a expr_

[kevina]

@nikomatsakis the above commit should fix it, please try it out and let me know if it causes you any problems, if it works for you fell free to use the commit.

[nikomatsakis]

That seems to fix it. I've cherrypicked the commit into a branch and will push it hopefully soon.

[kevina]

Thanks, fell free to close once pushed.

[catamorphism]

I had to modify this code somewhat to get it to compile, but the mentioned assertion failure doesn't arise, so am closing this.