1231 Commits

Author SHA1 Message Date
Michael Sullivan
4618e802db Fix comparisons of the nil type to do something sensible.
Closes #576.
2011-07-11 14:45:21 -07:00
Eric Holk
a84310aa48 Started working on MapReduce.
Currently it's only sequential, but it can do word frequency
counting. In an ideal world it would all be polymorphic, but that
pushes the limits of our type system right now. We can generalize it
later.
2011-07-11 14:05:09 -07:00
Michael Sullivan
7340824cbc Fix autoderef of function calls when the function is not an lval.
As it turns out, the correct way to handle this is much simpler than what I
did originally.
Also add more tests.
2011-07-11 11:38:01 -07:00
Michael Sullivan
418aa52510 Add a test that exercises these cases in bind. 2011-07-11 11:37:56 -07:00
Marijn Haverbeke
7595fe5153 Implement record patterns
Closes #469.
2011-07-11 15:32:41 +02:00
Marijn Haverbeke
86ee3454a1 Implement or-patterns in case clauses
You can now say

    expr_move(?dst, ?src) | expr_assign(?dst, ?src) { ... }

to match both expr_move and expr_assign. The names, types, and number
of bound names have to match in all the patterns.

Closes #449.
2011-07-11 11:01:54 +02:00
Patrick Walton
d30f22eb65 rt: Remove the _2 prefix from intrinsics; all rust-intrinsics now use explicit return pointers 2011-07-10 01:20:18 -07:00
Patrick Walton
f42c94740a rustc: Make rust-intrinsics take an explicit return pointer 2011-07-09 21:13:25 -07:00
Brian Anderson
9af59f9d81 Add sha1 and int tests to stdtest. Issue #428 2011-07-09 12:35:30 -07:00
Brian Anderson
45b00935f3 Begin running the test for int::pow
Somehow this has been disabled forever
2011-07-09 12:35:30 -07:00
Tim Chevalier
9ec5e90608 Tests for constraint propagation 2011-07-08 22:25:57 -07:00
Brian Anderson
7498d03693 Improve the error message for import glob collisions. Closes #482
Instead of noting where the imported things were defined, note where they were
imported. This is more useful and avoids issue #482.
2011-07-08 18:55:01 -07:00
Brian Anderson
394b8fcd1c Add test case for issue #333. Closes #333
This has been fixed by a mystery hacker.
2011-07-08 17:43:44 -07:00
Eric Holk
79ca673af7 Adding more support for working with u64s. 2011-07-08 14:02:26 -07:00
Patrick Walton
7c66894a71 rustc: Do dynamic size calculations properly when concatenating interior vectors. Closes #640. 2011-07-08 11:29:29 -07:00
Marijn Haverbeke
c7bfef43c8 Ignore current scope when resolving self-shadowing imports
That is, for example, import x::y::x, which defines a local x,
and thus wouldn't be able to find x::y anymore.

Closes issue #624
2011-07-08 16:04:40 +02:00
Marijn Haverbeke
022363a674 Auto-bind generic functions when their value is taken in non-call context
trans::trans_lval will now autobind if the given expression was the
name of a generic functions. Those callees (trans_call and trans_bind)
that are interested in the generics information call trans_lval_gen
now.
2011-07-08 14:28:46 +02:00
Eric Holk
2d57b25f6b Added a stress test mode to pfib. 2011-07-07 18:22:27 -07:00
Lindsey Kuper
e7111fe147 Un-xfail a test. 2011-07-07 17:39:35 -07:00
Lindsey Kuper
3243144046 Fix a bug that was interfering with method overriding. Issue #543.
Previously, we were creating both a normal vtable entry and a
forwarding function for overriding methods, when they should have just
gotten a vtable entry.  This patch fixes that.
2011-07-07 17:29:15 -07:00
Marijn Haverbeke
3bdbf74d47 Make moving of temporaries do the right thing, use it to optimize
This adds support for dropping cleanups for temporary values when they
are moved somewhere else. It then adds wraps most copy operations
(return, put in data structure, box, etc) in a way that will fall back
to a move when it is safe.

This saves a lot of taking/dropping, shaving over a megabyte off the
stage2/rustc binary size.

In some cases, most notably function returns, we could detect that the
returned value is a local variable, and can thus be safely moved even
though it is not a temporary. This will require putting some more
information in lvals.

I did not yet handle function arguments, since the logic for passing
them looked too convoluted to touch. I'll probably try that in the
near future, since it's bound to be a big win.
2011-07-07 15:54:01 +02:00
Lindsey Kuper
130006cdda Tie the knot for self-calls inside extended objects. Closes #539. 2011-07-06 18:39:01 -07:00
Brian Anderson
2f7bc90514 Add a stdtest crate to hold the standard library tests
This will link to std and compile with the --test flag. Eventually the
run-pass/lib* tests will move here.

We could also put the std tests directly into the library and compile both a
library version and a test version, but I think this way will make for faster
builds.

Issue #428
2011-07-06 14:39:40 -07:00
Brian Anderson
0eac640fdd Swap the expected/actual for typechecking fields of updated records
In 'rec(a = b with c)', if a exists in c then the expected type for b is the
type of c.
2011-07-06 14:04:52 -07:00
Lindsey Kuper
6bd80d0121 Add a compile-fail test for attempts to extend non-objects. 2011-07-06 10:47:36 -07:00
Marijn Haverbeke
8e585e7008 Support paren-free and lightweight-case syntax
(The old syntax is still supported as well, for now.)

It is now possible to leave out the parens around if, while, and
do/while conditions, and around alt expressions. Cases in an alt block
can now leave off the case keyword and parens around the pattern.

After the next snapshot, we can start migrating our code to use the
new alt syntax, probably with a pretty-printer pass. The paren-free
syntax will remain optional (you may always parenthesize expressions),
but the old case syntax will no longer be supported in the future.
2011-07-06 17:50:52 +02:00
Tim Chevalier
f279e0d76e Handle pattern-bound upvars
If a closure inside a case alternative (for example, a for each loop)
referenced a pattern-bound variable, this would cause an assertion
failure in trans. Changed trans::collect_upvars to handle pattern-bound
vars correctly.

Incidentally, eliminated all direct uses of option::get in trans.
2011-07-05 19:58:49 -07:00
Patrick Walton
1bba585044 stdlib: Implement ivec::foldl 2011-07-05 18:29:47 -07:00
Brian Anderson
2cb129355b Change ast::meta_name_value to accept any literal, not just string
This isn't useful for much of anything yet, since metadata::encoder doesn't
know how to handle the non-string variants.

Issue #611
2011-07-05 17:25:18 -07:00
Patrick Walton
8261d2e6fb stdlib: Add filter_map to ivec 2011-07-05 17:22:02 -07:00
Brian Anderson
375f0b2211 Require that the meta items in a use statement have unique names 2011-07-05 16:12:24 -07:00
Tim Chevalier
a793b85fbd Handle fail inside a for-each loop properly 2011-07-05 16:05:38 -07:00
Brian Anderson
b79ea489e3 Support conditional compilation of native items. Closes #610 2011-07-05 13:30:00 -07:00
Brian Anderson
8499c77cee Parse attributes for native items. Closes #609 2011-07-05 12:43:44 -07:00
Brian Anderson
6f8bc404f8 Error if the link attribute has duplicate items. Issue #614 2011-07-05 11:46:02 -07:00
Marijn Haverbeke
c59ebf0f01 Add a test file for distinct ty_native types
This should have gone into patch
77f5d14f14630da997f13de28ea86e8320957180, had I not forgotten to add
the file.
2011-07-05 11:50:41 +02:00
Patrick Walton
001397da3c stdlib: Add ivec::any() and ivec::all(); put out burning tinderbox 2011-07-04 23:52:47 -07:00
Patrick Walton
f71c8dd918 stdlib: Implement interior vector map 2011-07-04 22:48:42 -07:00
Patrick Walton
d3a4102bc1 stdlib: Add an inefficient implementation of ivec::pop 2011-07-04 21:20:18 -07:00
Josh Matthews
a2775a5b72 Make non-str fail expression a type checking failure instead of a translation one. 2011-07-03 21:46:17 -07:00
Tim Chevalier
b110bbf886 Un-XFAIL fail-arg 2011-07-03 11:56:23 -07:00
Tim Chevalier
046c25dbc0 Handle fail as an argument; parse fail expressions unambiguously
An expression like:

foo(1, fail, 2)

was failing to parse, because the parser was interpreting the comma
as the start of an expression that was an argument to fail, rather
than recognizing that the fail here has no arguments

Fixed this by using can_begin_expr to determine whether the next
token after a fail token suggests that this is a nullary fail or a
unary fail.

In addition, when translating calls, check before translating each
argument that the block still isn't terminated. This has the effect
that if an argument list includes fail, the back-end won't keep trying
to generate code for successive arguments and trip the !*terminated
assertion.
2011-07-03 11:51:47 -07:00
Tim Chevalier
c425a422ed Manipulate contexts correctly in trans_fail_expr
This fixes Issue #617
2011-07-03 02:38:11 -07:00
Tim Chevalier
9349826909 XFAILing the new test case...
...because the parsing of fail got changed out from under me.
Will fix shortly.
2011-07-03 01:15:41 -07:00
Tim Chevalier
7426252549 Generate code properly for calls with _|_ - typed arguments
The code for translating a fail (for example) would call
Unreachable(), which terminates the block; if a fail appeared as an
argument, this would cause an LLVM assertion failure. Changed
trans_call to handle this situation correctly.
2011-07-02 22:02:59 -07:00
Josh Matthews
a4421cc4e8 Fix assertion failure when syntax extension name is missing. 2011-07-02 00:06:58 -07:00
Brian Anderson
7c8ee712be Fix the error-pattern on compile-fail/fail-simple
Recent changes to the fail statement changed the way the parser errors in this
scenario.
2011-07-02 00:06:58 -07:00
Josh Matthews
d485e0d15a Allow any string expression to be used with fail. 2011-07-02 00:06:58 -07:00
Brian Anderson
22be98264e Add a test for mismatched crate metadata in use statements 2011-07-01 12:26:57 -07:00
Marijn Haverbeke
d863cdb98f Add tests for newtype-style tags 2011-07-01 17:51:49 +02:00