2653 Commits

Author SHA1 Message Date
Marijn Haverbeke
54cf0e9c0d Use precise return type to allocate retslot in trans_args
Using type_of_or_i8 did, predictably, allocate an i8 for a type parameter,
which leads to memory corruption and general confusion.

Closes #1443
2012-01-05 22:45:02 +01:00
Niko Matsakis
f832edc369 require a non-semi expr acting as a stmt to have unit return type 2012-01-05 10:44:38 -08:00
Stefan Plantikow
c8467977c4 Merge branch 'master' into kmath
Conflicts:
	src/libcore/float.rs
2012-01-05 17:20:13 +01:00
Stefan Plantikow
1f10ee365c Moved generic float::min, max to core::math and cleaned up some imports 2012-01-05 17:11:34 +01:00
Marijn Haverbeke
60ae1590af Switch to new param kind bound syntax
And remove support for the old syntax
2012-01-05 15:50:02 +01:00
Marijn Haverbeke
1f71a0f48d Clean up some FIXMEs related to impls 2012-01-05 14:55:23 +01:00
Marijn Haverbeke
3a20dda7ea Get external interfaces / impls working
Issue #1227
2012-01-05 14:17:31 +01:00
Marijn Haverbeke
6559aa8bfb Write the iface type of an impl in the crate data
Also, move checking of ifaces into the collect phase of typeck
to give further passes some guarantees.

Issue #1227
2012-01-05 11:04:41 +01:00
Marijn Haverbeke
3eb35900c7 Remove ty::bind_params_in_type
It does nothing that substitute_type_params can't do better.
2012-01-05 10:07:33 +01:00
Josh Matthews
0e98e64bc2 Add more item types to the ast_map so the test suite can build with debug info. 2012-01-05 01:20:51 -05:00
Marijn Haverbeke
3971b520bc Use the right types for methods in trans_impl
This prevents misalignment between function and argument
types in corner cases.
2012-01-04 22:12:48 +01:00
Tim Chevalier
1e54684093 Reformat typestate error messages so as not to confuse emacs compilation mode 2012-01-04 10:53:56 -08:00
Marijn Haverbeke
94d40be746 Prevent typenames in param bounds from resolving to their own param
I.e. fn foo<T: seq<T>>(...). This leads to weird circularities that seem to
never make any sense, so it seems prudent to forbid it.

Issue #1227
2012-01-04 17:30:36 +01:00
Marijn Haverbeke
42f6608ffd Add visit_ty_params to visit.rs
And use it to make typechecking of bounds less error-prone.
2012-01-04 17:30:36 +01:00
Marijn Haverbeke
a7e1a35f88 Properly typecheck and compile invocations of generic methods.
Aligning the type parameters of the ifaces, impls, and methods
correctly in typeck is almost brain surgery. Seems to work now for
everything I threw at it, but might still break in other corner cases.

Issue #1227
2012-01-04 17:28:51 +01:00
Tim Chevalier
439e28b751 Add missing ty_constr cases to trans::type_of_inner and ty::fold_ty.
Closes #970
2012-01-03 15:50:05 -08:00
Niko Matsakis
72a3667eb3 Allow tail expressions even in no_value blocks. Type checker
will guarantee they have unit type.
2012-01-03 13:13:52 -08:00
Brian Anderson
513a263e81 Merge pull request #1392 from Lenny222/list
list: add "is_not_empty" requirement to "head" and "tail" (analogous to "vec")
2012-01-03 12:18:53 -08:00
Marijn Haverbeke
4e88d5ae92 Make resolution of dictionaries on bounded params work
Issue #1227
2012-01-03 16:37:41 +01:00
Marijn Haverbeke
5ea3c96938 Wire in resolution of param bounds for method calls
Issue #1227
2012-01-03 16:18:40 +01:00
Marijn Haverbeke
e1dc40b271 More work on translating dictionary-passing
Reached a point where simple uses of interfaces without bounds work.

Issue #1227
2012-01-03 15:36:57 +01:00
Marijn Haverbeke
8c14943dea Properly handle expression blocks in kind.rs
It was only noticing them in expr_block form, not as function
bodies.

Closes #1390
2012-01-02 15:23:11 +01:00
Marijn Haverbeke
1741ef75ac Write out vtables for interface implementations
Issue #1227
2012-01-02 15:15:28 +01:00
Marijn Haverbeke
371b61a23d Drop two useless sub-passes from trans
Neither collect_tag_ctors nor the second pass in collect_items needed
to be separate passes.

Also remove obsolete obj_methods table kludge.
2012-01-02 15:15:28 +01:00
Marijn Haverbeke
a4694ce146 Create a trans_impl module 2012-01-02 15:15:27 +01:00
Marijn Haverbeke
3a1710d21e Key tcx.ty_param_bounds on node_ids, not def_ids
This makes it clearer that it's only valid for local nodes.
2012-01-02 15:15:27 +01:00
Marijn Haverbeke
9f2369dc9f Box arrays of parameter bounds 2012-01-02 15:15:27 +01:00
Marijn Haverbeke
270b4273e7 Pass bounds to trans::type_of_fn 2012-01-02 15:15:27 +01:00
Marijn Haverbeke
664a0443ad More resolving and typechecking of bounded type parameters.
Extern interfaces still don't get recognized.

Issue #1227
2012-01-02 15:15:27 +01:00
Marijn Haverbeke
40d5f288c3 Check that type parameter bounds are interface types
Issue #1227
2012-01-02 15:15:27 +01:00
Marijn Haverbeke
bd6646e698 Make last-use pass properly handle closed-over variables
Closes #1399
2012-01-02 10:20:58 +01:00
User Jyyou
a59c4b1b47 freebsd support 2012-01-01 20:18:55 -08:00
Niko Matsakis
aa5382bb13 split proto from fn_decl, as not all fn_decls know the proto.
this will address the (crashing) new test added.
2011-12-29 20:29:28 -08:00
Lenny222
d07c6e8a0e list: use predicate to enforce non-empty requirement 2011-12-29 21:24:03 +01:00
Marijn Haverbeke
816b0ac8ae Box ty_param_bounds_and_ty
It contains a vector, which shouldn't be copied all the time.
2011-12-29 09:29:08 +01:00
Graydon Hoare
2ed1005bd7 On second thought, re-land pull request #1385 with backquotes; easier to grep, can change in the future. 2011-12-28 13:41:31 -08:00
Graydon Hoare
f2c88c1b80 Merge pull request #1385 from Lenny222/quotes
Use singlequotes in the typechecker error messages too, to distinguish code and English
2011-12-28 13:08:28 -08:00
Marijn Haverbeke
9236fdf39f Move the kind datatype to middle::ty
The AST no longer references it.
2011-12-28 21:18:16 +01:00
Marijn Haverbeke
1ed6a27548 Change representation of type params to handle interface bounds
Issue #1227
2011-12-28 21:18:16 +01:00
Stefan Plantikow
7fd62bb2b5 Made matching machine types equal to float, int, uint (fixes #1376) 2011-12-28 18:02:49 +01:00
Lenny222
0b9751b141 Use singlequotes in the typechecker too, to distinguish code and English 2011-12-25 20:27:09 +01:00
Marijn Haverbeke
46cddffb8f Go back to a single visit_fn function in visit.rs 2011-12-23 18:45:02 +01:00
Marijn Haverbeke
711ff657e2 Get rid of visit_fn_block in visit.rs
It seems redundant and error-prone.
2011-12-23 18:24:49 +01:00
Marijn Haverbeke
25e65239ad Check impls methods against the type of their iface. 2011-12-23 18:11:36 +01:00
Marijn Haverbeke
057617c665 Parse iface items and interface references in impl items.
The (temporary) syntax is

    iface seq<T> {
        fn len() -> uint;
        fn iter(f: block(T));
    }

    // The 'blah<T>' can be left of to default the name of the
    // impl to seq<T>. The 'of seq<T>' can be left off when
    // not implementing a named interface.
    impl blah<T> of seq<T> for [T] {
        fn len() -> uint { vec::len(self) }
        fn iter(f: block(T)) { for x in self { f(x); } }
    }
2011-12-23 18:09:25 +01:00
Marijn Haverbeke
9292744959 Use the same type of record in ty::ty_fn and ty::method
Removes some more code duplication.
2011-12-23 16:11:35 +01:00
Marijn Haverbeke
970f5cc0e4 Make ast::ty_method hold a fn_decl, rather than duplicating its fields 2011-12-23 13:32:17 +01:00
Graydon Hoare
f0dfbe7b1b Register new snapshots, purge log_err and log_full in favour of log(...). 2011-12-22 17:53:53 -08:00
Graydon Hoare
389329ef1e Merge all 3 log syntaxes, tidy up residual misuses. 2011-12-22 16:14:00 -08:00
Graydon Hoare
dcd771cd48 Merge branch 'master' of github.com:graydon/rust 2011-12-22 14:50:52 -08:00