Commit Graph

153 Commits

Author SHA1 Message Date
Tim Chevalier
664d71f1f0 Improve a typechecker error message (wrong type for main())
As per #1903, state what type is expected for main().

Closes #1903
2012-03-11 20:55:19 -07:00
Patrick Walton
46aec2c960 rustc: Record the parent function of each function 2012-03-11 17:02:16 -07:00
Patrick Walton
cb1efb0092 rustc: Fix long lines 2012-03-11 16:18:52 -07:00
Patrick Walton
8f9afec5a3 rustc: Turn on errors for region checking 2012-03-11 13:29:23 -07:00
Patrick Walton
3e8bb22af1 rustc: Unify regions in most cases. (Nested functions aren't properly handled yet.) 2012-03-11 13:28:43 -07:00
Patrick Walton
f7efc29b38 rustc: Determine regions for locals 2012-03-11 12:36:56 -07:00
Patrick Walton
e968672ac7 rustc: Map local variables to their containing blocks so we know their regions 2012-03-11 12:05:17 -07:00
Tim Chevalier
8ab075ee45 Clean up occurs check code and give non-breaking loop {..}s _|_ type
The latter change is so that code dominated by a loop{ } without a
break gets considered unreachable.

The former change is just cosmetic (occurs_check_fails was a predicate
when it should be a unit-typed function that can fail).
2012-03-10 20:35:41 -08:00
Tim Chevalier
205cefdc6e In typestate, consider infinite loops w/ conts to be infinite
If a loop { } contains a cont, that doesn't affect whether the entire
loop diverges. Only breaks affect that. Fix that in typestate.
2012-03-10 20:34:57 -08:00
Tim Chevalier
35400e13ad Use loop instead of while(true) in libraries and compiler itself
And remove spurious fails/unreachable() calls.
2012-03-10 20:34:17 -08:00
Tim Chevalier
98260a2a22 Handle infinite-loop poststate correctly in typestate
If control passes an infinite loop (that doesn't have non-local
exits), then everything is true.
2012-03-10 20:15:27 -08:00
Brian Anderson
91e5a1c8b3 core: Remove the nearly empty math module
This mod only had two functions, all of whose users have been changed
to use the uint module.
2012-03-10 17:43:19 -08:00
Brian Anderson
e4bb2d707f core: Rename vec::tail_n to vec::tailn to match other fns 2012-03-10 00:44:13 -08:00
Tim Chevalier
a0521971b1 Merge branch 'master' into floop-for-snapshot 2012-03-09 22:16:52 -08:00
Patrick Walton
e8c7b5347d rustc: Add a region checking pass
Errors aren't reported yet, because the regions aren't properly inferred from the & operator at the moment and that would break the tests.
2012-03-09 17:39:32 -08:00
Tim Chevalier
321fd80219 Add an infinite loop construct
Add a loop {} construct for infinite loops, and use it in test
cases. See #1906 for details.
2012-03-09 16:40:58 -08:00
Patrick Walton
4ffcb95974 rustc: Perform region inference 2012-03-09 16:40:37 -08:00
Patrick Walton
7b938d8903 rustc: Add region def ids to the AST 2012-03-09 15:55:13 -08:00
Patrick Walton
e2f6052c6d Revert "rustc: Create def IDs for region parameters"
This reverts commit 0d4cb75949.
2012-03-09 14:40:36 -08:00
Patrick Walton
0d4cb75949 rustc: Create def IDs for region parameters 2012-03-09 14:06:43 -08:00
Patrick Walton
8ce6b0dd21 rustc: Add appropriate type predicates for references. Simple uses of references/region pointers now work. 2012-03-08 17:22:43 -08:00
Patrick Walton
71799cff2b rustc: Actually write the type back 2012-03-08 16:44:33 -08:00
Patrick Walton
ddeaea2287 rustc: Change the address-of operator to an explicit production so that we can parse the mutability 2012-03-08 16:34:36 -08:00
Patrick Walton
0722786664 rustc: Add some missing cases in various folds for reference types and some broken trans code for the address-of operator 2012-03-08 15:55:28 -08:00
Tim Chevalier
0c5fdc8745 Rename last to last_opt, last_unsafe to last
As per discussion on IRC. I am about to file an RFC for further
discussion about the more general issue of whether to enforce
invariants through types, typestate, or dynamic checks, but for now,
removing the misleading name "last_unsafe".
2012-03-08 15:25:56 -08:00
Patrick Walton
8047c0cd68 rustc: Typecheck dereference operations on safe references.
They require unsafe blocks for now. This will be removed once they typecheck properly.
2012-03-08 15:08:26 -08:00
Patrick Walton
fcde161f4f rustc: Stub typechecking for the address-of operator 2012-03-08 14:55:51 -08:00
Tim Chevalier
d048a00cf3 Change util::unreachable to core::unreachable
Closes #1931
2012-03-08 14:30:01 -08:00
Patrick Walton
0e17cdb627 rustc: Add regions to the type system 2012-03-08 14:05:56 -08:00
Tim Chevalier
ebc1d3e704 Rename last_total to last_unsafe
See Issue 1943 for any discussion (reopen it if necessary).

Closes #1943
2012-03-08 13:07:31 -08:00
Patrick Walton
c9cf73f0a8 rustc: represend -> represent in a comment 2012-03-08 13:01:08 -08:00
Marijn Haverbeke
c71306b0db Explicitly store self_ids use for self locals in methods
This makes it possible to move them between crates without confusion,
and to instantiate them at a point where the monomorphizing
substitutions are known.

Issue #1944
2012-03-08 21:17:32 +01:00
Marijn Haverbeke
fd465f91a8 Drop collect_items pass, create decls on demand
This solves a problem with inlined functions that have inner functions.
2012-03-08 21:17:32 +01:00
Patrick Walton
ec8380f9b1 rustc: Add the safe address-of operator to the AST 2012-03-08 12:04:02 -08:00
Patrick Walton
6224fdcd08 rustc: Add region-annoted type parameters to the AST; stub associated patterns 2012-03-08 10:49:43 -08:00
Niko Matsakis
2bfed908e3 Fix #1941: inlining of items that themselves contain nested items
The fix is to drop nested items from the encoded AST.  Nested items may
themselves be inlined, but that is an independent question.
2012-03-07 18:06:29 -08:00
Patrick Walton
c9375fed8d stdlib: Stop incurring vtable dispatch costs when hashmaps are used
This required changing almost all users of hashmaps to import the hashmap interface first.

The `size` member in the hashmap structure was renamed to `count` to work around a name conflict.
2012-03-07 17:35:13 -08:00
Patrick Walton
c245d9e980 Revert "stdlib: Stop incurring vtable dispatch costs when hashmaps are used"
This reverts commit f0250a23d3.
2012-03-07 16:56:45 -08:00
Patrick Walton
f0250a23d3 stdlib: Stop incurring vtable dispatch costs when hashmaps are used
This required changing almost all users of hashmaps to import the hashmap interface first.

The `size` member in the hashmap structure was renamed to `count` to work around a name conflict.
2012-03-07 16:51:31 -08:00
Tim Chevalier
90ac69904f Improve a type error message (avoid the word 'non-dereferenceable') 2012-03-07 14:14:34 -08:00
Marijn Haverbeke
7b81cd8930 Fix #1503 in a saner way
Closes #1933
2012-03-07 09:19:29 +01:00
Niko Matsakis
712dd23541 make inline enabled by default 2012-03-06 09:01:40 -08:00
Marijn Haverbeke
4e4bd585cd Track all exports in exp_map
This is needed because the reachability checker needs to be able to
follow exports.

Issue #1934
2012-03-06 17:11:30 +01:00
Marijn Haverbeke
19508c7d53 Make trans pass responsible for pulling in inlined functions
This makes the logic for finding the inlinable items much easier --
they are simply pulled in lazily when encountered.
2012-03-06 14:00:21 +01:00
Marijn Haverbeke
9b88219723 Don't attach def ids to types when the type already contains the def id 2012-03-06 14:00:13 +01:00
Brian Anderson
77ed09728b rustc: Use std::util::unreachable 2012-03-05 18:16:25 -08:00
Brian Anderson
0ee96de4ee rustc: Lower case error messages 2012-03-05 17:05:20 -08:00
Brian Anderson
13a07ce1d4 rustc: Remove some redundant text from error messages 2012-03-05 17:05:20 -08:00
Niko Matsakis
0416a946b7 give user control of whether we use inline hint or inline always
supply inline hints for resource ctors/dtors
2012-03-05 16:47:51 -08:00
Niko Matsakis
cc2ebfc105 set LLVM inline hint for all #[inline] methods, whether CC or local 2012-03-05 16:47:51 -08:00
Brian Anderson
15df3950ad rustc: Remove remaining non-inbounds GEPs 2012-03-05 15:12:36 -08:00
Tim Chevalier
e2fa6f03f5 Translate simple classes
Programs using classes with fields only (no methods) compile and run,
    as long as nothing refers to a class in a different crate (todo).

    Also changed the AST representation of classes to have a separate
    record for constructor info (instead of inlining the fields in the
    item_class node), and fixed up spans and pretty-printing for
    classes.
2012-03-04 12:43:17 -08:00
Graydon Hoare
87c14f1e3d Move src/comp to src/rustc 2012-03-02 18:46:13 -08:00