mikros/rust
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
7,455 Commits 2 Branches 0 Tags 1.4 GiB
3ebf7b42bd
Commit Graph

2620 Commits

Author SHA1 Message Date
Marijn Haverbeke
8cf3ca3b55 Make sure by-copy arguments take last-use into account 2011-11-18 16:43:30 +01:00
Marijn Haverbeke
196b2b920f Make sure trans translates record fields in the order they appear in code 
This prevents surprising side-effect orders, and makes them easier for
the other passes to deal with.
 2011-11-18 15:59:47 +01:00
Marijn Haverbeke
68db68c4cc Make trans use last_use info to not actually generate copies 
Issue #925
 2011-11-18 15:56:53 +01:00
Marijn Haverbeke
f6491bb426 Update stdlib, compiler, and tests to new kind system 
This involved adding 'copy' to more generics than I hoped, but an
experiment with making it implicit showed that that way lies madness --
unless enforced, you will not remember to mark functions that don't
copy as not requiring copyable kind.

Issue #1177
 2011-11-18 12:49:01 +01:00
Marijn Haverbeke
8f8ebb550c Implement a last-use-of-local finding algorithm 
Issue #925
 2011-11-18 12:49:01 +01:00
Marijn Haverbeke
0c97fcbf66 Properly check kinds in instantiation of generics 
Issue #1177
 2011-11-18 12:49:01 +01:00
Marijn Haverbeke
7bef89f9b5 Prevent alias pass from inserting implicit copies for noncopyable types 
Issue #1177
 2011-11-18 12:49:01 +01:00
Marijn Haverbeke
17b78c4a7a Overhaul the kind-checking pass 
Not really useful yet because missing last-use-of-local optimization.

Also: instantiation of type parameters needs to be checked.

Issue #1177
 2011-11-18 12:49:01 +01:00
Marijn Haverbeke
cefff237bf Make tag, resource and object constructors take their arguments by copy 
Doing something like some([1, 2, 3]) will now no longer create a temporary
copy of the vector. It will also be easier for the kind checker to see that
putting a resource into a data-structure constructor is safe.
 2011-11-18 12:49:01 +01:00
Marijn Haverbeke
4e03112141 Add a pass-by-copy parameter passing convention 
This is intended to solve the problem of how to pass arguments to
constructor functions -- you want to move in rvalues, but not have to
explicitly copy stuff that is not an rvalue. The by-copy passing
convention will ensure the callee gets its own copy of the value. For
rvalues, it'll just pass off the value. For lvalues, it'll make a
copy.

Issue #1177
 2011-11-18 12:49:00 +01:00
Marijn Haverbeke
9cf48d3753 Preparation for kind system overhaul 
This goes before a snapshot, so that subsequenct patches can make the
transition without breaking the build. Disables kind checking pass, makes
parser accept both new and old-style kind annotation.

Issue #1177
 2011-11-18 10:52:28 +01:00
Niko Matsakis
b30f7af98b remove blank line 2011-11-17 15:11:52 -08:00
Niko Matsakis
8925875ccb add FIXME for issue #1184 2011-11-17 13:44:13 -08:00
Niko Matsakis
0d874476cf fix resource-generic 2011-11-17 13:44:13 -08:00
Brian Anderson
dd275cd749 Stop using temporary intrinsic_2 functions 2011-11-17 11:55:04 -08:00
Niko Matsakis
adc79e2f44 remove compile-command from local variable blocks 2011-11-17 11:46:13 -08:00
Haitao Li
388eed383f rustc: Add a flag '--warn-unused-imports' 
Followup of issue #889
 2011-11-17 09:28:30 +01:00
Haitao Li
fe6484d3e6 rustc: Fix warn on unused import bug 
rustc generates incorrect warning for cascaded import declarations like:

  use std;
  import std::io;
  import io::println;
  fn main() { println("hello"); }

  warning: unused import io

A followup of issue #889
 2011-11-17 09:27:17 +01:00
Niko Matsakis
809ca13bfa fix minor merge errors 2011-11-16 16:13:43 -08:00
Niko Matsakis
037a1de405 apply calling convention at call site too 2011-11-16 15:27:09 -08:00
Niko Matsakis
834b6879ea temp workaround for failure to pass ulonglong successfully 2011-11-16 15:27:09 -08:00
Niko Matsakis
cbcdeb80d9 remove wrappers from intrinsics 2011-11-16 15:27:08 -08:00
Niko Matsakis
e3699a2636 make stdcall use shim 2011-11-16 15:24:56 -08:00
Niko Matsakis
319f3b02f3 fix natives with user-specified link names, remove unused imports 2011-11-16 15:24:38 -08:00
Niko Matsakis
b27a88e99c fix bug in shape where s_int/s_uint were not customized to platform 2011-11-16 15:16:43 -08:00
Niko Matsakis
b3ae63c9d9 temporarily disable these tests b/c they crash rustc 2011-11-16 15:16:43 -08:00
Niko Matsakis
cb9675259f enable comments in generated asm, ll 2011-11-16 15:16:42 -08:00
Niko Matsakis
96cdfa11db correct translation of neg. numbers in 64-bit architectures 2011-11-16 15:16:42 -08:00
Niko Matsakis
5cabfb3a39 remove fixed FIXME 2011-11-16 15:16:42 -08:00
Niko Matsakis
4fc34e23fa wrap long line 2011-11-16 15:16:40 -08:00
Niko Matsakis
d77968dd7c finish up the shim approach 2011-11-16 15:16:40 -08:00
Niko Matsakis
4f28419d0c add back call to zero_and_revoke 2011-11-16 15:16:40 -08:00
Niko Matsakis
85083ec51b begin efforts to use shim functions for c-stack calls 2011-11-16 15:16:40 -08:00
Haitao Li
3b683f5205 rustc: Use link_name attribute for native function 
Fixes issue #906
 2011-11-16 23:45:07 +08:00
Brian Anderson
a7188a6577 rustc: Don't sign extend uints during translation 
Doing so produces incorrect results on a 32-bit compiler targeting 64-bits
 2011-11-15 17:53:44 -08:00
Haitao Li
eae478cd74 Use native fn's link name attribute if given 
Fixes issue #905
 2011-11-15 22:31:33 +08:00
Marijn Haverbeke
9ff6f816ba Fix handling of loops and conditionals in alias.rs 
It now threads information about invalidated aliases through the AST
properly. This makes it more permissive for conditionals (invalidating
an alias in one branch doesn't prevent you from using it in another),
and less permissive for loops (it now properly notices when a loop
invalidates an alias that it might still use in another iteration).

Closes #1144
 2011-11-15 13:29:57 +01:00
Haitao Li
b8dd148444 rustllvm: Add a GetOrInsertFunction wrapper 
Fixes issue #1161

Test-case-by: Brian Anderson <banderson@mozilla.com>
Signed-off-by: Haitao Li <lihaitao@gmail.com>
 2011-11-15 00:33:29 +08:00
Brian Anderson
3cc2641fc3 Make various binops not typecheck for unsafe pointers. Closes #1173 2011-11-13 13:55:12 -08:00
Haitao Li
5a7249c935 Cleanup unused imports 2011-11-10 19:19:40 -08:00
Haitao Li
9bc2f1800d rustc: Add support of warning on unused imports 
Fixes issue #889
 2011-11-10 19:19:10 -08:00
Brian Anderson
1afc943c91 rustc: Rename native_abi_c_stack_cdecl to native_abi_cdecl 2011-11-09 16:47:44 -08:00
Brian Anderson
10c35959e7 rustc: Rename native_abi_c_stack_stdcall to native_abi_stdcall 2011-11-09 16:47:44 -08:00
Brian Anderson
c7f0ca6063 Remove native "x86stdcall" ABI 2011-11-09 16:47:43 -08:00
Brian Anderson
4a4d31cf0e Make native "c-stack-stdcall" ABI work 
The symbol name is based on the number of arguments, so we have to declare
it correctly.

For some reason GetLastError doesn't work now.
 2011-11-09 15:29:53 -08:00
Joshua Wise
c8fae5dc75 trans: Add binops (except for logic and comparators) and unops to trans_const_expr. Working towards issue #570. 2011-11-09 14:33:06 -08:00
Joshua Wise
e674a7e99f middle: Add a pass to reject bad const expressions earlier. Currently just rejects unimplemented const expressions, but will be needed later. 2011-11-09 14:33:05 -08:00
Brian Anderson
4f3f04643f Remove native "cdecl" ABI 2011-11-08 17:17:06 -08:00
Brian Anderson
fcd39b1191 Remove native "llvm" ABI 2011-11-07 15:29:05 -08:00
Brian Anderson
1103fe8ca0 Remove native "rust" ABI 2011-11-07 15:14:10 -08:00
Haitao Li
852e789e78 rustc: Set LLVM module identifier as crate name 2011-11-07 22:59:21 +08:00
Marijn Haverbeke
1a68a98824 Disallow writing to function arguments again 
Remove implicit copying hack.

Closes #1118
 2011-11-03 10:57:54 +01:00
Niko Matsakis
8385d77b32 fix line too long 2011-11-02 14:14:20 -07:00
Niko Matsakis
ec77144713 add -m64 to gcc args if appropriate 2011-11-02 14:14:20 -07:00
Niko Matsakis
e012fd03f3 correct more i32 vs int mismatches 2011-11-02 14:14:19 -07:00
Niko Matsakis
602ae9139c convert various ints to i32 and vice versa 2011-11-02 14:14:19 -07:00
Niko Matsakis
4bd05e5be2 convert GEP to i32 2011-11-02 14:14:19 -07:00
Niko Matsakis
f1bc9307d8 get things checking on ia32 2011-11-02 14:14:19 -07:00
Niko Matsakis
c0e9c42bd2 thread the context through so that int can be 64 bits on x86_64 2011-11-02 14:14:19 -07:00
Niko Matsakis
2521cda1ec work on making the size of ints depend on the target arch 2011-11-02 14:14:19 -07:00
Niko Matsakis
9146bb09a1 work on making the size of ints depend on the target arch 2011-11-02 14:14:19 -07:00
Niko Matsakis
5b3bddcd87 start to isolate target strings so that they can be configured for 
x64 or other targets
 2011-11-02 14:13:22 -07:00
Marijn Haverbeke
eef9a0bd39 Evaluate alt expressions in their own block context 
Closes #785
 2011-11-02 14:30:51 +01:00
Marijn Haverbeke
d8d35e7c40 Rename car/cdr to head/tail in std::list 
Closes #1086
 2011-11-02 14:23:49 +01:00
Marijn Haverbeke
5b0c103b39 Make it possible to cast unsafe pointers with the 'as' operator 2011-11-02 12:20:12 +01:00
Marijn Haverbeke
4e9b8a28f7 Make resolving of imports behave more sanely 
An import now ignores itself when resolving its target. This gets rid of
the previously existing (problematic) behaviour where the import would start
looking one scope up when its name was the same as its target's first
component.

Closes #1114
 2011-11-02 12:20:11 +01:00
Brian Anderson
61b604580b Fix alignment of interior pointers of dynamic-size types. Closes #1112 
GEP_tup_like finds interior pointers by creating a tuple of all the types
preceding the element it wants a pointer to, then asks for the size of that
tuple. This results in incorrect pointers when the alignment of that tuple
is not the alignment of the interior type you're getting a pointer to.
 2011-11-01 12:13:00 -07:00
Matt Brubeck
b510a74a36 Cleanup: Remove uneccesary generic_cmp function. 2011-10-28 14:44:39 -07:00
Matt Brubeck
000b2fe9a6 Use IEEE 754 semantics for NaN (Issue #1084) 2011-10-28 14:44:39 -07:00
Brian Anderson
f96ad30dfc stdlib: Make reinterpret_cast and leak unsafe 2011-10-28 14:19:17 -07:00
Marijn Haverbeke
7dacccde94 Make shared kind the default only for generic functions 
You almost never want a function with pinned type params. For
types, objects, resources, and tags, pinned types are actually often
more sane. For most of these, shared rarely makes sense. Only tricky
case is objs -- you'll have to think about the kinds you want there.

Issue #1076
 2011-10-28 17:00:14 +02:00
Marijn Haverbeke
7a0aee74bf Move to short type parameter keywords 
Issue #1076
 2011-10-28 15:25:33 +02:00
Matt Brubeck
5d6fe1a533 Remove uint::max/min in favor if math::max/min 2011-10-27 13:37:53 -07:00
Marijn Haverbeke
013107a25c Properly take mutable object fields into account during alias analysis 
Closes #1055
 2011-10-25 17:57:26 +02:00
Marijn Haverbeke
cfdf193c46 Update our code to new type parameter kind syntax 
Closes #1067
 2011-10-25 15:56:55 +02:00
Niko Matsakis
c78b1639b4 fix c-stack-cdecl when used w/ i64 2011-10-24 17:03:18 -07:00
Niko Matsakis
d41385a69b kill errant log_err 2011-10-24 16:06:18 -07:00
Niko Matsakis
8f2d75d53c switch over sqrt from llvm to c-stack-cdecl, exposing a bug in 
the supported return types of upcall_c_stack
 2011-10-24 16:06:18 -07:00
Niko Matsakis
020726c981 update trans_c_stack_native_call() to use type_of_explicit_args() 
currently trans_c_stack_native_call() had some ad-hoc code for
determining the type of the arguments. this code was not in
agreement with the rest of trans. now it uses the same code path.
 2011-10-21 16:19:28 -07:00
Niko Matsakis
80102c9d24 avoid extra load for by_mutable_ref parameters 2011-10-21 16:19:27 -07:00
Marijn Haverbeke
7114702496 Change the way block calls are parsed, mark them as block-calls. 
This makes it possible to omit the semicolon after the block, and will
cause the pretty-printer to properly print such calls (if
pretty-printing of blocks wasn't so broken). Block calls (with the
block outside of the parentheses) can now only occur at statement
level, and their value can not be used. When calling a block-style
function that returns a useful value, the block must be put insde the
parentheses.

Issue #1054
 2011-10-21 14:24:14 +02:00
Marijn Haverbeke
0ce40f60e7 Drop support for iter, put, and for-each 
Closes #1056
 2011-10-21 13:47:24 +02:00
Marijn Haverbeke
050170d2af Remove remaining uses of iter and for-each 
Issue #1056
 2011-10-21 13:34:35 +02:00
Marijn Haverbeke
fd5a08384d Move ast_util::pat_bindings over to new iter system. 
Issue #1056
 2011-10-21 12:41:42 +02:00
Marijn Haverbeke
6bcb4a426c Remove last uses of iterators from stdlib 
Issue #1056
 2011-10-21 12:31:48 +02:00
Marijn Haverbeke
4ebbbe597e Move hash table iteration over to block-taking functions 
Issue #1056
 2011-10-21 12:21:27 +02:00
Brian Anderson
4ef2c37059 Give native functions proto_bare 2011-10-20 20:53:52 -07:00
Brian Anderson
6fb9d4f83b Remove typeck::proto_to_ty_proto 2011-10-20 20:30:20 -07:00
Brian Anderson
9efdd0f326 Merge ast::proto_shared and ast::proto_closure 
Now they are both just proto_shared and proto_shared takes an
argument indicating that it is sugared as 'lambda'
 2011-10-20 18:23:48 -07:00
Brian Anderson
4b30a06abe Rename ast::proto_fn to ast::proto_shared 2011-10-20 18:23:48 -07:00
Brian Anderson
29ad3bdb10 Make fn denote a bare function. Convert fn to fn@ as needed 2011-10-20 18:23:48 -07:00
Brian Anderson
391e12124b Tweak typecheck to enforce covariance on higher-order function arguments 2011-10-20 18:23:48 -07:00
Brian Anderson
342a400e79 Remove type coercion from fn to block 
fn is a proper subtype of block now
 2011-10-20 18:23:48 -07:00
Brian Anderson
8136b92ee8 Give subtype relationships to function types 2011-10-20 18:23:48 -07:00
Brian Anderson
ed985b61d5 Bare functions can coerce to blocks 2011-10-20 18:23:48 -07:00
Brian Anderson
354bfc8292 Bare functions can coerce to shared closures 2011-10-20 18:23:48 -07:00
Brian Anderson
3bb020aaf8 Revert "Remove the environment argument from bare functions" 
This reverts commit 1b0f1f0b79.
 2011-10-20 18:23:48 -07:00
Brian Anderson
0714bb5f7d Fix the problem with check-fast randomly failing 
Trans has been assuming that tag node id's are unique across crates and they
are not so, depending on which way the wind is blowing, it would choose to use
a crate-local tag variant instead of the correct one from std.

No test case since I can't come up with a reliable one that compiles in a
reasonable amount of time.
 2011-10-20 15:38:51 -07:00
Marijn Haverbeke
ef63f09dcd Don't rely on main.ll anymore 
I'll remove the actual file after I register a snapshot.

Issue #992
 2011-10-20 17:18:38 +02:00
Marijn Haverbeke
8124846b2c Get rid of taskpointer-passing throughout the compiler 
Only intrinsics still take a dummy taskptr. We'll have to do some
makefile stunts to snapshot a version without taskptrs-in-intrinsics.

Issue #466
 2011-10-20 14:22:17 +02:00
Marijn Haverbeke
457683c7fe Do not pass taskpointers to "rust" native functions 
Issue #466
 2011-10-20 14:22:17 +02:00
Marijn Haverbeke
6323a012bd Remove taskptr argument from upcalls 
Issue #466
 2011-10-20 13:15:21 +02:00
Marijn Haverbeke
e927df17f7 Remove spawn_wrap and main_wrap kludges 
This isn't needed now that our functions are cdecl (and was apparently
only still working by accident).

Issue #992
 2011-10-20 13:15:09 +02:00
Brian Anderson
314c011d71 Immutable and mutable? are covariant on their inner types 
Whereas [mutable T] is invariant with respect to T, [T] and [mutable? T]
are covariant with respect to T.
 2011-10-18 10:37:04 -07:00
Marijn Haverbeke
71a4a66135 Add a math module to the standard lib 
I need some rudimentary stdlib stuff for the tutorial.

Closes #1042
 2011-10-18 13:54:35 +02:00
Marijn Haverbeke
d9d0782b5f Don't try to build an LLVM switch on floats 
Closes #1046
 2011-10-18 13:54:35 +02:00
Brian Anderson
0f0fa26c72 Enforce variance rules for mutable types 2011-10-17 20:51:34 -07:00
Brian Anderson
48d351b21e Add a variance transform for calculation of recursive variance 2011-10-17 19:37:34 -07:00
Brian Anderson
d6613384fd Introduce covariance, contravariance and invariance to the type unifier 
This will be used to resolve some problems with mutable? covariance and also
to implement function subtyping
 2011-10-17 19:34:12 -07:00
Brian Anderson
b61578e1cd Don't unify from mutable? to immutable 
Immutable is supposed to be a subtype of mutable-huh.
 2011-10-17 16:36:49 -07:00
Brian Anderson
f7a825c614 Split record_var_binding into methods for expected and actual 
This allows unify to maintain the same subtype relationship between
expected and actual throughout unify, which we are going to need
for mutable? and for function types.
 2011-10-17 16:36:49 -07:00
Brian Anderson
1a64153b24 Correct two more analyses of mutable? as mutable 
I can't come up with test cases but this seems correct.
 2011-10-17 16:35:33 -07:00
Brian Anderson
4d9d889dbf Don't allow assignment to mutable-wha? 2011-10-17 16:35:33 -07:00
Marijn Haverbeke
0a01144797 Use trans_expr_save_in instead of trans_expr when appropriate 
This guards against passing a save_in destination to nil or bot-typed things.
 2011-10-14 11:27:16 +02:00
Marijn Haverbeke
06257d66e7 Don't crash when reporting an unresolved import in the nil scope 
Closes #1027
 2011-10-14 11:25:14 +02:00
Brian Anderson
3808a498cd Remove an unnecessary visitor from rustc::middle::fn_usage 2011-10-13 14:55:50 -07:00
Brian Anderson
acf9b83e71 Fix a use of visit in rustc::middle::fn_usage 
This was working by just skipping an AST node instead of doing the advertised
check. I can't find a test case for it, but this is more correct.
 2011-10-13 14:55:47 -07:00
Brian Anderson
bc4c3df1bf Don't allow the value of generic bare functions to be taken 
Issue #1038
 2011-10-13 14:39:30 -07:00
Brian Anderson
89466ffd74 Rename rustc::middle::unsafeck to fn_usage 
I'm going to add further checks unrelated to unsafe.

Issue #1038
 2011-10-13 12:25:57 -07:00
Marijn Haverbeke
6a7800508b Make trans_anon_obj properly thread its bcx 
Closes #893

Unfortunately, anon objs are still so broken as to be useless
(fields don't work).
 2011-10-13 18:04:09 +02:00
Marijn Haverbeke
e4c91fdf65 Properly fill single-variant zero-arg tag values. 
Closes #1034. Closes #1035.
 2011-10-13 15:14:09 +02:00
Niko Matsakis
e8a0e592da reimplement some of the unsafe stuff which got lost 
- blocks inherit unsafety
  - remove the --check-unsafe flag
  - add unsafe annotations where needed to get things to compile
 2011-10-12 16:33:07 -07:00
Niko Matsakis
215b1ab000 make compiler emit more than 1 error in the case of unsafe 2011-10-12 16:33:07 -07:00
Niko Matsakis
3b66806624 prohibit ptr deref unless in unsafe code 2011-10-12 16:33:07 -07:00
Niko Matsakis
a56698575c fix error msg 2011-10-12 16:33:07 -07:00
Niko Matsakis
81533ff737 add pass to check that unsafe fns cannot be used as values 2011-10-12 16:33:06 -07:00
Niko Matsakis
f2cd33df72 continue to annotate trans functions as unsafe where neccessary 2011-10-12 16:33:06 -07:00
Niko Matsakis
2094668949 continue to annotate trans functions as unsafe where neccessary 2011-10-12 16:33:06 -07:00
Niko Matsakis
d4d7eb069b add unsafe tags into various points in the translation chains 
and so forth
 2011-10-12 16:33:06 -07:00
Niko Matsakis
c1cefa52d0 Add unsafe tags to usage of LLVM funcs 2011-10-12 16:33:06 -07:00
Niko Matsakis
212707ce84 make native functions markable as unsafe and incorporate that 
into the type check
 2011-10-12 14:39:34 -07:00
Niko Matsakis
24b201fa48 skip test, remove whitespace 2011-10-12 14:39:34 -07:00
Niko Matsakis
b9a9559c91 correct lines over 78 chars 2011-10-12 14:39:34 -07:00
Niko Matsakis
f3c68e7238 enable unsafe checking but only with a flag --check-unsafe 2011-10-12 14:39:34 -07:00
Niko Matsakis
cbe8da0655 make treatment of unchecked/unsafe blocks more uniform 
also repair various errors in the parser related to such blocks.
rename checked_blk to default_blk to reflect the fact that it
inherits its purity from the surrounding context.
 2011-10-12 14:39:24 -07:00
Niko Matsakis
046ca827dd Add unsafe blocks, unsafe functions, and two rudimentary tests 
related to them
 2011-10-12 14:26:47 -07:00
Niko Matsakis
29584cc5ac Extend the unchecked block stuff to allow unsafe blocks as well. 2011-10-12 14:26:47 -07:00
Niko Matsakis
58b8e88356 expand purity to include unsafe 2011-10-12 14:26:47 -07:00
Brian Anderson
99f876e3a8 Make build_environment and trans_bind_thunk GEP bound arguments the same 
These functions both use GEP_tup_like to get at the arguments bound to the
environment, but they were starting from a different 'level' of the
environment-box structure. Frighteningly, this was leading to them having
different opinions of how the bound arguments were aligned in some cases.
 2011-10-12 12:11:07 -07:00
Marijn Haverbeke
306f7fb25f Simplify copying of uniques 
We can call take glue, rather than duplicating it inline.
 2011-10-12 11:05:20 +02:00
David Rajchenbach-Teller
8c9dd54ded [Renaming] str_to_float is now float::from_str, float_to_str is now float::to_str 2011-10-12 10:56:05 +02:00
Brian Anderson
1b0f1f0b79 Remove the environment argument from bare functions 
Issue #1022
 2011-10-11 10:51:10 -07:00
Brian Anderson
5b0f79b75a Don't allow bind to produce bare functions 
Issue #1022
 2011-10-11 10:51:10 -07:00
Brian Anderson
b277039325 Introduce ast::proto_bare 
Issue #1022
 2011-10-11 10:51:10 -07:00
David Rajchenbach-Teller
2ec85135e2 [Move] Moved str_to_float, float_to_str from compiler to lib 2011-10-11 16:41:34 +02:00
Marijn Haverbeke
33167f7dec Adjust function signatures to allow for vecs being immediate 
Some code was relying on vectors being implicitly by-reference (as
non-immediate value). This adds the necessary &&-sigils.

Closes #1021
 2011-10-10 16:01:51 +02:00
Marijn Haverbeke
b4bae8fea5 Make vectors and strings immediates again 
There's no good reason to force them to be spilled anymore. Some
pieces of trans become more elegant this way, and less stack allocs
and load/stores are needed.

Issue #1021
 2011-10-10 15:59:16 +02:00
Marijn Haverbeke
5c53959f20 Make bcx_tcx-style accessors return by reference 2011-10-10 12:25:58 +02:00
Marijn Haverbeke
00daeb485c Move type_is_immediate into ty.rs 2011-10-10 12:23:58 +02:00
Marijn Haverbeke
a35dbf3fd5 Don't spill immediates in order to drop them 
Issue #1012
 2011-10-10 12:20:28 +02:00
Marijn Haverbeke
fa1295343f Make free glue take a pointer to the heap part (box) 
This way, it can be used to drop values without first spilling them.

Issue #1012
 2011-10-10 10:25:32 +02:00
Marijn Haverbeke
aff536ec0e Fix bad float-type detection in trans 
Closes #1017
 2011-10-09 10:07:10 +02:00
Marijn Haverbeke
c7eee8fb05 Stop registering cleanups for types that don't need them 2011-10-07 14:46:23 +02:00
Marijn Haverbeke
31586339b1 Make it possible to have locals that don't live on the stack 
Local values that are not mutated, don't need to be cleaned up, and
are immediate, don't need to be spilled. (All immediate args, and
non-pointer immediate let locals.)
 2011-10-07 14:43:19 +02:00
Marijn Haverbeke
8db71530f5 Forbid passing dynamically-sized types by value 
Issue #1008
 2011-10-07 10:41:40 +02:00
Marijn Haverbeke
2ff89469d4 Actually pass parameters specified as by-val by value 
Issue #1008
 2011-10-07 10:41:40 +02:00
Marijn Haverbeke
fe916fb9f0 Give up on providing a by-value version of map, convert fold over to 
passing pointers by ref

Issue #1008
 2011-10-07 09:56:12 +02:00
Marijn Haverbeke
f9fbd86f52 Parse and typecheck by-value and by-ref arg specs 
Add sprinkle && throughout the compiler to make it typecheck again.

Issue #1008
 2011-10-07 09:09:50 +02:00
Marijn Haverbeke
39b468e17a Fix problem with polymorphic return value being instantiated as nil 
Closes #1005
 2011-10-06 09:40:56 +02:00
Marijn Haverbeke
142f63fe78 Tie up the last ends in destination-passing-conversion 
Closes #667
 2011-10-05 12:23:18 +02:00
Marijn Haverbeke
3bc3f96138 Make trans_lval lval-only 
Call trans_temp_lval if you want the old fallback-to-temporary behaviour.

Issue #667
 2011-10-05 11:51:41 +02:00
Marijn Haverbeke
2cc0008ec0 Add trans_temp_expr for what used to be trans_expr, rename t_e_dps to trans_expr 
Issue #667
 2011-10-05 11:26:27 +02:00
Marijn Haverbeke
975ac55b3d Lose the by_ref destination style 
It's not needed.

Issue #667
 2011-10-05 10:41:30 +02:00
Marijn Haverbeke
babd1ff3c7 Revert "Revert "Reorganize translation of expr_field"" 
This reverts commit 11e407aaa9.
 2011-10-05 10:21:57 +02:00
Marijn Haverbeke
c33e61e75c Revert "Revert "Get rid of 'overwrite' destination kind"" 
This reverts commit ce9e0fc94f.
 2011-10-05 10:21:48 +02:00
Marijn Haverbeke
062aa3272c Revert "Revert "Move trans_call to destination_passing style"" 
This reverts commit 276dfc6133.
 2011-10-05 10:21:40 +02:00
Patrick Walton
990233eb3d rustc: Allow calling native functions on the C stack that take generic arguments. They get turned into i8 pointers. 2011-10-04 17:41:57 -07:00
Patrick Walton
276dfc6133 Revert "Move trans_call to destination_passing style" 
This reverts commit 6e56ec0066.
 2011-10-04 17:05:35 -07:00
Patrick Walton
ce9e0fc94f Revert "Get rid of 'overwrite' destination kind" 
This reverts commit 6e652588bf.
 2011-10-04 17:05:27 -07:00
Patrick Walton
11e407aaa9 Revert "Reorganize translation of expr_field" 
This reverts commit 007ec666ad.
 2011-10-04 17:05:16 -07:00
Marijn Haverbeke
007ec666ad Reorganize translation of expr_field 
Issue #667
 2011-10-04 17:52:23 +02:00
Marijn Haverbeke
6e652588bf Get rid of 'overwrite' destination kind 
It wasn't safe (computing the rval might invalidate the lval addr), and needlessly
complicating things (code was already building up intermediary results to work
around other unsafeties).

Issue #667
 2011-10-04 17:52:23 +02:00
Marijn Haverbeke
6e56ec0066 Move trans_call to destination_passing style 
Issue #667

The retptr passed to a function will now often be the actual
destination of the returned value (as in `{field1: func1()}`).
 2011-10-04 17:41:34 +02:00
Patrick Walton
968b66ad40 rustc: Implement C stack stdcall 2011-10-03 13:59:38 -07:00
Patrick Walton
e9287e55cc rustc: Stub a --stack-growth option; it's behind a flag for now because it requires patches to LLVM. 2011-09-30 18:21:39 -07:00
Patrick Walton
c065280a0b rustc: Remove some debug code 2011-09-30 14:26:06 -07:00
Patrick Walton
dcf0e807fb trans: Use inttoptr as appropriate when casting return values of C stack functions 2011-09-30 12:07:50 -07:00
Patrick Walton
79e4961c7f rustc: Remove some debug code 2011-09-29 18:53:00 -07:00
Patrick Walton
f9e3a604a6 rustc: Make the generic C stack function signature include a usable return value 2011-09-29 18:52:23 -07:00
Patrick Walton
a5b422fe6b rustc: Translate calls on the C stack 2011-09-29 17:34:49 -07:00
Patrick Walton
41ab324539 rustc: Generate types in trans for the C-stack native ABI 2011-09-29 17:34:49 -07:00
Marijn Haverbeke
e4436c3a5d Don't build up results for ignored expressions. 
Issue #667
 2011-09-29 12:02:37 +02:00
Marijn Haverbeke
d74a7a83d7 Fail with an informative error when trying to bind a method 
Issue #435
 2011-09-29 11:55:29 +02:00
Marijn Haverbeke
d243ea57f5 Move last remaining expression types out of trans_expr 
Issue #667
 2011-09-29 11:41:45 +02:00
Marijn Haverbeke
f089f533e8 Move expr_cast and expr_copy to DPS style 
Issue #667
 2011-09-29 10:46:49 +02:00
Marijn Haverbeke
d946e09a72 Move closure construction over to DPS style 
Issue #667
 2011-09-29 10:36:40 +02:00
Marijn Haverbeke
508c48ce10 Reduce move ops to a construct-in-place when possible 2011-09-29 10:36:40 +02:00
Brian Anderson
15e34deba3 Eliminate glue.o. Closes #990 2011-09-28 15:35:21 -07:00
Brian Anderson
cb4e99b688 Don't allow vectors of pinned kinds 
Vectors of pinned kinds can't be safe because most interesting uses of vector
perform copies
 2011-09-28 15:35:21 -07:00
Patrick Walton
f7d0c1cec3 rustc: Add a new "C stack cdecl" native ABI 2011-09-28 12:58:33 -07:00
Brian Anderson
381205b747 Revert "Revert "Implement pattern ranges for all numeric types."" 
This reverts commit a034f87146.

Conflicts:

	src/comp/middle/check_alt.rs
	src/comp/middle/trans_alt.rs
	src/comp/syntax/ast.rs
	src/comp/syntax/ast_util.rs
	src/comp/syntax/fold.rs
	src/comp/syntax/print/pprust.rs

Conflicts:

	src/comp/middle/trans_alt.rs
 2011-09-28 12:46:29 -07:00
Patrick Walton
dbf472b0df rustc: Switch to cdecl for all calls. This is needed to make stack growth efficient, as I need to use eax and ecx in the function prologue and can't afford to stomp on incoming arguments. 2011-09-28 10:22:02 -07:00
Marijn Haverbeke
512cfb4b83 Remove a few kludges that worked around issue #913 2011-09-28 12:59:44 +02:00
Marijn Haverbeke
a3a27374bd Don't descend into functions from relax_precond_block 
Closes #913
 2011-09-28 12:37:28 +02:00
Marijn Haverbeke
759fc101fb Get rid of might_not_init kludge in init_local. 
Initializing something is now safe wrt to cleanups (so the cleanup for the
local is only registered after the initialization.)
 2011-09-28 11:12:35 +02:00
Marijn Haverbeke
16d5aaa962 Wait with registering cleanups for locals until after the init code ran 
Fixes a double-cleanup that was breaking unwind-tup.rs under valgrind.
 2011-09-28 10:41:19 +02:00
Marijn Haverbeke
e3626c9405 Use DPS for assignment and local initialization 
Issue #667
 2011-09-28 09:42:04 +02:00
Brian Anderson
a96b16e8c3 Make it again possible to initialize resource locals via assignment 
Some special cases allow both 'let a <- my_resource(x)' and
'let a = my_resource(x)' to work as expected despite ostensibly being
copies and moves.
 2011-09-27 23:20:27 -07:00
Brian Anderson
459353e107 Prevent copies of resources into various things 2011-09-27 21:50:07 -07:00
Brian Anderson
888e22aacc Add more comments about kind checking on local move initializers 2011-09-27 16:03:10 -07:00
Brian Anderson
b8bb663df7 Don't ever raise unique kinds of pinned kinds to shared (again) 
So *resource, ~resource, [resource] are all pinned. This is counter to the
design of the kind system, but this way is a much clearer path to type safety.
Once we've established a good baseline with lots of tests, then we can try to
make raising pinned kinds work.
 2011-09-27 16:03:10 -07:00
Brian Anderson
e50580aa66 Revert "Use DPS for assignment and local initialization" 
This reverts commit b1b202d302.
 2011-09-27 16:03:09 -07:00
Marijn Haverbeke
04f062f9d4 Fix lurking bug in matching of nil 
Which was somehow made visible by the DPS changes.
 2011-09-27 22:49:33 +02:00
Marijn Haverbeke
b1b202d302 Use DPS for assignment and local initialization 
Issue #667
 2011-09-27 20:33:09 +02:00
Marijn Haverbeke
e42f3b8481 Safeguard against self-assignment for unique types 2011-09-27 20:33:09 +02:00
Marijn Haverbeke
94db38a530 Move expr_unary to DPS style 
Issue #667
 2011-09-27 17:31:15 +02:00
Marijn Haverbeke
92d63ffa65 Make trans_rec and trans_tup understand overwrite dests 
Issue #667
 2011-09-27 17:31:15 +02:00
Marijn Haverbeke
8282f7fe80 Move expr_binary into trans_expr_dps 
Issue #667
 2011-09-27 17:30:54 +02:00
Marijn Haverbeke
806e74fbf5 Move expr_lit and expr_vec into trans_expr_dps 
Issue #667
 2011-09-27 17:30:45 +02:00
Marijn Haverbeke
8ea72fae99 Move more non-value-returning things to trans_expr_dps. 
Issue #667
 2011-09-27 17:30:45 +02:00
Brian Anderson
6b42ad5ea0 Enforce copy restrictions on let initializers 2011-09-26 22:00:15 -07:00
Jesse Ruderman
20c2d2a999 Distinguish #924 2011-09-26 20:44:57 -07:00
Brian Anderson
3778b6c6a8 Prevent copying of uncopyable things via the copy op 2011-09-26 20:03:30 -07:00
Brian Anderson
755001725a Prevent copying of uncopyable things via compound assignment ops 2011-09-26 20:03:24 -07:00
Brian Anderson
99cbea51a1 Don't allow vectors of pinned kinds to be copied 2011-09-26 19:57:00 -07:00
Brian Anderson
4bee452221 Don't allow copying of unique boxes of pinned kinds 
Issue #409
 2011-09-26 18:43:49 -07:00
Brian Anderson
389852b5c0 Fix tags of unique boxes 
Issue #409
 2011-09-26 14:44:08 -07:00
Brian Anderson
223f5be166 Unique vectors and boxes of pinned are unique kinds. Closes #977 2011-09-26 13:38:46 -07:00
Marijn Haverbeke
8c0918ac18 Construct records and tuples in-place 
Issue #667

Now, {a: {b: 10, c: 20}, d: @30} will simply write the values in the right
places, rather than creating intermediary records and then memmoving them.
Cuts about a megabyte off the unoptimized compiler size.
 2011-09-26 22:13:08 +02:00
Marijn Haverbeke
5837975f95 Convert more expression types to use DPS 
expr_rec, expr_tup, most of the exprs that don't return anything. Make
trans_ret almost trivial by using destination adaptors (trans_save_in,
trans_by_ref).

Issue #667
 2011-09-26 21:34:10 +02:00
Marijn Haverbeke
345b5a47bd Add a workaround hack for issue #913 
I can't figure out what the real cause of this bug is, but I want
to be able to use blocks inside loops again.
 2011-09-26 21:34:10 +02:00
Graydon Hoare
93f0e9bac3 Refine notes in kind.rs some more. Add a fixme to ty.rs. 
Kinds are still pretty wobbly. See thread starting at
https://mail.mozilla.org/pipermail/rust-dev/2011-September/000807.html
 2011-09-25 21:20:17 -07:00
Marijn Haverbeke
1e988f228d Don't call trans_bind_thunk in an unreachable context 
Closes #973
 2011-09-26 04:08:45 +02:00
Marijn Haverbeke
4c5e42a139 Make 'fail fail;' compile. 
Closes #972
 2011-09-26 03:57:08 +02:00
Marijn Haverbeke
6164b72402 Fix bug in mutability-checking pass 
It wasn't properly handling function arguments.
 2011-09-26 03:56:39 +02:00
Brian Anderson
98db2febd3 Handle ~fail. Closes #968 2011-09-24 23:52:22 -07:00
Brian Anderson
4f7ecce7d1 Support non-immediates in trans_uniq::copy_val 
Issue #409
 2011-09-24 21:28:58 -07:00
Lindsey Kuper
822ed2c6d3 Small refactors. 
Factor out some shared code from the start of process_fwding_mthd and
process_bkwding_mthd; get rid of unneeded temp variable.
 2011-09-24 23:38:47 -04:00
Jesse Ruderman
3b5b29c7ec Reduce and clarify abuse of 'pure' in interner 2011-09-24 16:33:26 -07:00
Jesse Ruderman
064f52fa96 Mark some fns as pure so type_is_unique_box doesn't need to be unchecked 2011-09-24 15:36:09 -07:00
Brian Anderson
c5877450e6 Tweak the type allocated for unique pointers 
Seems to be more correctish...

Issue #409
 2011-09-24 12:36:51 -07:00
Brian Anderson
777a536313 Support non-immediates in unique-box type glue 
Issue #409
 2011-09-24 12:36:51 -07:00
Brian Anderson
856acbf66d Vectors containing pinned kinds become pinned 
Otherwise they could be copied
 2011-09-24 12:36:51 -07:00
Brian Anderson
e5d5682065 Unique pointers containing pinned kinds become pinned 
Issue #409
 2011-09-24 12:36:51 -07:00
Brian Anderson
d8f6e9f237 Add let destructuring for unique boxes 
Issue #409
 2011-09-23 16:12:04 -07:00
Brian Anderson
2082f67765 Begin to support pattern matching on unique boxes 
Issue #409
 2011-09-23 15:15:30 -07:00
Marijn Haverbeke
2c7f21c524 Properly mark unreachable alt bodies as unreachable 2011-09-23 23:37:15 +02:00
Marijn Haverbeke
572a4b71d5 Make AddIncomingToPhi take single values rather than arrays 2011-09-23 23:20:19 +02:00
Marijn Haverbeke
64c69aa7b8 Start on a piecemeal conversion to DPS 
Issue #667

Wires in a basic framework for destination-passing style, with
backwards-compatibility to the old approach, so that expression types
can be moved over to it one at a time (by moving them from trans_expr
to trans_expr_dps).
 2011-09-23 22:49:37 +02:00
Marijn Haverbeke
d114dedf9a Don't return unused values from zero_alloca and trans_stmt 2011-09-23 22:28:05 +02:00
Marijn Haverbeke
abf002ad60 Fix bug where the type of a function's top-level block wasn't fixed up 2011-09-23 22:28:05 +02:00
Marijn Haverbeke
348cd86d87 Properly check types of do-while condition expr 
Closes #956
 2011-09-23 11:19:15 +02:00
Marijn Haverbeke
0bbde048a4 Make sure no dynamic allocas are used before they are allocated 
Closes #965
 2011-09-23 11:11:37 +02:00
Marijn Haverbeke
49813cffb1 Use a slightly nicer hack to get zero-length strings in trans_build 
Wrapping calls in str::by_ref(, ...) heap-allocated an empty string
every time (and looked really bad).
 2011-09-23 11:11:37 +02:00
Marijn Haverbeke
420484579d Better handling of unreachable code in trans 
The builder functions in trans_build now look at an 'unreachable' flag
in the block context and don't generate code (returning undefined
placeholder values) when this flag is set. Threading the unreachable
flag through context still requires some care, but this seems a more
sane approach than re-checking for terminated blocks throughout the
compiler.

When creating a block, if you use its closest dominator as parent, the
flag will be automatically passed through. If you can't do that,
because the dominator is a scope block that you're trying to get out
of, you'll have to do something like this to explicitly pass on the
flag:

    if bcx.unreachable { Unreachable(next_cx); }

Closes #949. Closes #946. Closes #942. Closes #895. Closes #894.
Closes #892. Closes #957. Closes #958.
 2011-09-23 11:09:57 +02:00
Brian Anderson
dbd2ee015a Remove is_stateful flag from tydesc. Closes #136 2011-09-22 22:15:25 -07:00
Brian Anderson
0e5f76a718 Fix ty_uniq case in maybe_auto_unbox 
Closes #961
Issue #409
 2011-09-22 18:14:18 -07:00
Brian Anderson
1a48023a79 Add take glue for unique boxes 
Closes #962
Issue #409
 2011-09-22 18:05:36 -07:00
Brian Anderson
77fcab043e Remove nonsensical load and store from trans_uniq::copy_val 
Issue #409
 2011-09-22 17:50:27 -07:00
Brian Anderson
7a76323459 Allow vectors to contain unique boxes. Closes #952 
Issue #409
 2011-09-22 17:00:53 -07:00
Patrick Walton
e372f943e2 rustc: Write raw type parameters instead of linearized type parameters in object body shapes 2011-09-22 16:13:03 -07:00
Brian Anderson
30a4eab380 Autoderef indexes and fields of unique boxes 
Issue #409
 2011-09-22 16:06:25 -07:00
Brian Anderson
d174d917e2 Calculate the correct kind for unique boxes 
Issue #409
 2011-09-22 15:28:49 -07:00
Brian Anderson
3f41563cc9 Drop the previous value when copying one unique box local to another 
Issue #409
 2011-09-22 13:53:34 -07:00
Brian Anderson
67bac873e0 Add assignment to unique box locals 
Issue #409
 2011-09-22 13:53:33 -07:00
Brian Anderson
268a9fe5fb Initialize unique box locals from other locals 
Issue #409
 2011-09-22 13:53:33 -07:00
Brian Anderson
5d5136df9f Convert trans_uniq asserts to preconditions 
Issue #409
 2011-09-22 13:53:33 -07:00
Brian Anderson
c4f02a7925 Sprinkle some asserts through trans_uniq 2011-09-22 13:53:33 -07:00
Brian Anderson
8492eed137 Move uniq trans code to trans_uniq module 
Issue #409
 2011-09-22 13:53:30 -07:00
Brian Anderson
0f0ba33d82 Factor alloc_uniq from trans_uniq 
Issue #409
 2011-09-22 13:52:30 -07:00
Brian Anderson
f6ab12a3b1 Don't unify unique boxes with different mutability 
Issue #409
 2011-09-22 11:10:48 -07:00
Brian Anderson
ea67099234 Add support for mutable unique boxes 
Issue #409
 2011-09-22 10:26:34 -07:00
Brian Anderson
1b3023e4d0 Convert ty::ty_uniq to contain a mutable type 
Issue #409
 2011-09-22 10:26:34 -07:00
Brian Anderson
4d088bd528 Call drop glue on the thing in a unique box 
Issue #409
 2011-09-22 10:26:34 -07:00
Brian Anderson
42fd2a9d89 Remove unused arg from revoke_clean 2011-09-21 15:39:46 -07:00
Brian Anderson
2cc01e2ca2 Add ability to deref unique boxes. Make unique boxes immediates. 
Issue #409
 2011-09-21 14:40:55 -07:00
Marijn Haverbeke
80778f642f Revert "rustc: Make bottom types immediates. Add a |log ret| testcase. Closes #935." 
This reverts commit f19ab1ff3c.
 2011-09-21 21:39:31 +02:00
Patrick Walton
f19ab1ff3c rustc: Make bottom types immediates. Add a |log ret| testcase. Closes #935. 2011-09-21 10:55:28 -07:00
Patrick Walton
fcd17731cf rustc: Don't try to spill bottom types, since we can't make any assumptions about the type of the ValueRef 2011-09-21 10:55:27 -07:00
Marijn Haverbeke
a034f87146 Revert "Implement pattern ranges for all numeric types." 
This reverts commit ce0f054f9d.
 2011-09-21 18:42:09 +02:00
Marijn Haverbeke
410f73fdb2 Don't write to retptr when returning nil, pass undef for unused params 2011-09-21 17:57:19 +02:00
Marijn Haverbeke
07eb29dbce Revert "rustc: Make bottom types immediates. Add a |log ret| testcase. Closes #935." 
This reverts commit e6a84f252a.
 2011-09-21 11:05:54 +02:00
Josh Matthews
ce0f054f9d Implement pattern ranges for all numeric types. 2011-09-21 09:36:12 +02:00
Patrick Walton
e6a84f252a rustc: Make bottom types immediates. Add a |log ret| testcase. Closes #935. 2011-09-20 22:52:26 -07:00
Brian Anderson
f809e22697 Represent unique creation as a unop in the AST instead of its own expr 
Like the box unop.

Issue #409
 2011-09-20 18:06:47 -07:00
Brian Anderson
7ae251789c Make creation of unique boxes work again 
Issue #409
 2011-09-20 17:51:17 -07:00
Brian Anderson
865dcb663d Parse unique box types 
Issue #409
 2011-09-20 17:51:17 -07:00
Patrick Walton
d607316b2f rustc: Pass actual tydescs to upcall_malloc 2011-09-20 13:48:22 -07:00
Patrick Walton
78147452b8 rustc: Remove trans_raw_malloc, as it's only called from one place 2011-09-20 13:38:54 -07:00
Brian Anderson
870117f44e Zero locals that are inited via alt expressions. Closes #945 2011-09-20 10:28:16 -07:00
Marijn Haverbeke
e33f116598 Remove some obsolete 'copy' operators from trans 2011-09-20 17:35:40 +02:00
Marijn Haverbeke
1eaddb772e Properly handle 'put;' in trans 
Closes #947
 2011-09-20 17:34:32 +02:00
Brian Anderson
7c6630a367 Build tuples in two phases to avoid cleaning up partial tuples 2011-09-19 17:59:53 -07:00
Brian Anderson
9626e2f71f Register tuple cleanups after the tuple is built 
Issue #936
 2011-09-19 17:59:47 -07:00
Brian Anderson
dd1cf63515 Build records in two phases to avoid cleanups on partial records 2011-09-19 14:28:04 -07:00
Brian Anderson
432e931a2a Don't register record cleanups until the record is built 2011-09-19 14:28:04 -07:00
Marijn Haverbeke
eafb6789a2 Stop creating function pairs for static functions 
These are now called directly. When taking their value, a null pointer is added
to create a pair.
 2011-09-18 22:05:58 +02:00
Tim Chevalier
8640e67e3f Add a precondition to GEP_tup_like 2011-09-17 10:18:30 -07:00
Tim Chevalier
2b98eccfee Add a type_is_tup_like predicate that takes a block ctxt, and make some fns pure 2011-09-17 10:18:11 -07:00
Tim Chevalier
8613f28a16 In trans, replace another impossible case with a precondition 2011-09-17 09:13:05 -07:00
Brian Anderson
1002623201 Make move-mode arguments unwind correctly. Closes #939 2011-09-16 14:38:04 -07:00
Tim Chevalier
99236d64d0 Dead code elim 2011-09-16 13:32:44 -07:00
Tim Chevalier
fa74df033d Add a constraint on trans::type_of_fn_from_ty 
And change callers appropriately.
 2011-09-16 13:07:08 -07:00
Tim Chevalier
0d4b51d10d Add a returns_non_ty_var predicate 
which just calls non_ty_var on the return type of t, assuming t is
a function type.
 2011-09-16 13:06:31 -07:00
Tim Chevalier
1b2c5c4c9b Make ty_fn_ret pure and get rid of a duplicate function 2011-09-16 13:06:31 -07:00
Marijn Haverbeke
43b219dbec Work around an apparent compiler bug to make my patch run 
Without this, it would fail when optimization is turned on. The result of
the Load(bcx, llclosure) call would end up in llenv in a corrupted form.
 2011-09-16 21:31:36 +02:00
Marijn Haverbeke
575aae407b Saner approach to lvalues and callable values in trans 
LValues no longer carry information about generics and objs, instead
there's an extended form of lvalue, lval_maybe_callee, only used by
call and bind, which holds this info.

This makes it possible to take the value of a method and get a working
closure, and will (with some more work) allow us to call statically
known functions without loading from their pair.

Closes #435
Closes #758
 2011-09-16 21:31:36 +02:00
Tim Chevalier
172df2e3a2 Add non_ty_var precondition for type_of_fn, plus minor cleanup 2011-09-16 10:58:02 -07:00
Marijn Haverbeke
ad470d741e Revert "Saner approach to lvalues and callable values in trans" 
This reverts commit 66153436c9.
 2011-09-16 18:37:46 +02:00
Marijn Haverbeke
133d36f452 Require body of else-less if expressions to be a value-less block 
For consistency with other constructs that could not possibly return a
value (say, loops).
 2011-09-16 18:21:06 +02:00