RFC on long series to free libsyntax of vecs_implicitly_copyable

[erickt]

Good morning,

It's taken a long time, but I finally am almost done freeing libsyntax of vecs_implicitly_copyable in this pull request, but I'm running into some issues. I've confirmed that all but the last commit (which only disables vecs_implicitly_copyable pass the check tests. The last commit errors with this message, which makes no sense to me:

/Users/erickt/rust/rust/src/libcore/num/f32.rs:35:37: 35:43 error: expected `,` but found `=`
/Users/erickt/rust/rust/src/libcore/num/f32.rs:35         pub pure fn $name($( $arg : $arg_ty ),*) -> $rv {
                                                                                       ^~~~~~

and this stack trace:

#1  0x00000001000b059b in sys::begin_unwind_::_a923ca4ae164c::_06 ()
#2  0x00000001000b0542 in sys::begin_unwind::anon::anon::expr_fn_13876 ()
#3  0x00000001000048a1 in sys::begin_unwind::_8ec273289fc0adc0::_06 ()
#4  0x00000001005df999 in diagnostic::__extensions__::meth_7941::span_fatal::_efdf2d14612d79ec::_06 ()
#5  0x0000000100682d48 in parse::parser::__extensions__::meth_16938::fatal::_8aa3239426747a3::_06 ()
#6  0x00000001006850b8 in parse::common::__extensions__::meth_17005::expect::_d3604ec6c7698d5f::_06 ()
#7  0x00000001006b59f1 in parse::common::__extensions__::parse_seq_to_before_end_17860::_48c79835f9eb1011::_06 ()
#8  0x00000001006a50f7 in parse::parser::__extensions__::meth_17606::parse_fn_decl::_14f3785fe78967d::_06 ()
#9  0x00000001006b6f59 in parse::parser::__extensions__::meth_17987::parse_item_fn::_8a6be529cf7b2ca5::_06 ()
#10 0x00000001006ac839 in parse::parser::__extensions__::meth_17761::parse_item_or_view_item::_bfead947d6dd7d25::_06 ()
#11 0x00000001006c8b8f in parse::parser::__extensions__::meth_18364::parse_item::_96b54e33f65abe76::_06 ()
#12 0x000000010076179f in ext::tt::macro_rules::add_new_extension::generic_extension::anon::anon::expr_fn_23365 ()
#13 0x000000010072e793 in ext::expand::expand_item_mac::_a4f486c4465cfb1b::_06 ()
#14 0x00000001007b5ad3 in __morestack ()

There also a bunch of new warnings that I haven't cleaned up yet: https://gist.github.com/erickt/5048251.

@nikomatsakis thought there might be some scary bug in the parser caused by moving a vector in the parser instead of copying it, which is why I'm filing this pull request before it's ready. Thanks for any help!

[graydon]

I know this is unappealing to say, but have you tried bisecting the series to isolate that error?

[pcwalton]

It may also help to try running the test suite (make check-stage1), which often produces smaller test cases.

[erickt]

@graydon: unfortunately there isn't much to bisect. After I merge in with HEAD, all but the commit that disables vecs_implicitly_copyable should be landable after passing all tests. I can't bisect because I need these 31 commits before I can turn off vecs_implicitly_copyable :(

[erickt]

Err, to be clear, all tests pass already pass for all but the last commit.

[brson]

Since this is such a massive series, maybe we should go ahead and land all but the last commit.

[erickt]

@brson: Ok. I'll merge in HEAD and remove the last commit and file a bug.

[erickt]

This commit is mergable now, and all the tests passed on the try branch.

[erickt]

@bors: retry

[erickt]

@bors: retry p=50

[erickt]

r+