Jean Privat [Fri, 7 Dec 2012 20:25:38 +0000 (15:25 -0500)]
nitg: implements implicit casts for unsafe nexpr
add the option --no-check-autocast to disable it.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Fri, 7 Dec 2012 21:09:33 +0000 (16:09 -0500)]
tests: declare Bool to base_types_nullable_formal_and_virtual
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Fri, 7 Dec 2012 20:03:49 +0000 (15:03 -0500)]
niti: implements implicit casts for unsafe nexpr
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Fri, 7 Dec 2012 20:01:01 +0000 (15:01 -0500)]
typing: new attribute AExpr::implicit_cast_to
If required, the following implicit cast ".as(XXX)"
Such a cast may by required after evaluating the expression when
a unsafe operation is detected (silently accepted by the Nit language).
The attribute is computed by `check_subtype`
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Fri, 7 Dec 2012 19:26:38 +0000 (14:26 -0500)]
typing: the return of check_subtype can distinguish safety and unsafely.
Thus visit_expr_subtype can return the correct type
as if an implicit cast was added.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Fri, 7 Dec 2012 06:54:47 +0000 (01:54 -0500)]
lib: fix ArrayMapValues::first
Thanks to the new MType::is_subtype test, bugs like these can be spotted.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Fri, 7 Dec 2012 06:53:26 +0000 (01:53 -0500)]
model: rewrite of MType::is_subtype
The previous writing was just wrong.
Mainly because the two first `resolve_for` always lost the parameter types.
But the rest of the method was also overly complex and buggy.
Note that the fixed behavior make a test pass because of the
"FIXME workaround" in typing.nit
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 6 Dec 2012 06:06:28 +0000 (01:06 -0500)]
lib: rewrite Collection::rand without explicit closure
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 6 Dec 2012 14:59:08 +0000 (09:59 -0500)]
tests: test equality on Float
The point is that some floats are equal even if the bit representation
differs. This helps to illustrate bad implementation of floats.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 6 Dec 2012 00:43:32 +0000 (19:43 -0500)]
tests: add base_attr_init_val_raf.nit
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 6 Dec 2012 00:26:04 +0000 (19:26 -0500)]
tests: add base_gen_variance.nit
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 6 Dec 2012 00:32:15 +0000 (19:32 -0500)]
niti: do not output object_id (not reliable)
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 5 Dec 2012 19:06:40 +0000 (14:06 -0500)]
nitg: check_init_instance use a call CHECK_NEW_XXX
This is analogous the the instantiation that call NEW_XXX.
The point is to move global stuff (like the collect of attributes)
to the link-time in separate compilers.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 5 Dec 2012 16:17:50 +0000 (11:17 -0500)]
nig: localize init_expr in attributes
Add tests to illustrate the behavior.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 5 Dec 2012 15:07:22 +0000 (10:07 -0500)]
nitg: factorize Compiler#generate_init_attr
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 5 Dec 2012 14:52:43 +0000 (09:52 -0500)]
nitg: check_init_instance require a type
Relying on recv.mcasttype is dangerous.
The check_init_instance should be analogous to the init_instance.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 5 Dec 2012 14:21:17 +0000 (09:21 -0500)]
bench: compilation_time: replace nitg.bin by nitg-e
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 5 Dec 2012 04:49:23 +0000 (23:49 -0500)]
nitg: two equal C types do not imply equal Nit types
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 5 Dec 2012 04:03:36 +0000 (23:03 -0500)]
nitg: add dangerous options to disable runtime tests
These options can be used to evaluate the runtime cost of some
features of Nit.
run_bench is updated to include these options.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 5 Dec 2012 02:31:50 +0000 (21:31 -0500)]
nirg-sep: introduce box_kind to implements the "is" equality test
Each class is associated with a box_kind value that indicate if it is
a box and what kind of box.
Thus on comparaison, the box_kind can be used to unbox and compare
the real value.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 5 Dec 2012 02:55:24 +0000 (21:55 -0500)]
nitg-sep: add option --no-inline-intern
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Tue, 4 Dec 2012 02:28:19 +0000 (21:28 -0500)]
nitg-sep: inline primitive call on intern methods
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Tue, 4 Dec 2012 14:06:43 +0000 (09:06 -0500)]
nitg-e: fix implementation of class_name_string
see
a9435075f8ea304415ca846d69f14b888b3ed37f for details.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Tue, 4 Dec 2012 13:14:41 +0000 (08:14 -0500)]
nitg: move engine selection to nitg.nit
No more ugly hijacking.
Signed-off-by: Jean Privat <jean@pryen.org>
Alexandre Terrasa [Mon, 3 Dec 2012 22:27:23 +0000 (17:27 -0500)]
nitg-sep: fix calloc_array to use the live generic type of self
Signed-off-by: Alexandre Terrasa <alexandre@moz-code.org>
Alexandre Terrasa [Mon, 3 Dec 2012 18:45:16 +0000 (13:45 -0500)]
Merge branch 'master' of https://github.com/privat/nit
Alexandre Terrasa [Mon, 3 Dec 2012 18:35:33 +0000 (13:35 -0500)]
nitg-sep: better handling of nullable types
Signed-off-by: Alexandre Terrasa <alexandre@moz-code.org>
Alexandre Terrasa [Mon, 3 Dec 2012 17:31:36 +0000 (12:31 -0500)]
nitg-sep: colorize live generic types entries
Signed-off-by: Alexandre Terrasa <alexandre@moz-code.org>
Conflicts:
src/separate_compiler.nit
Jean Privat [Mon, 3 Dec 2012 04:59:28 +0000 (23:59 -0500)]
bench: add a boostrap test to generate compilers
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Mon, 3 Dec 2012 04:42:21 +0000 (23:42 -0500)]
bench: add nitg-e as en engine
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Mon, 3 Dec 2012 04:29:03 +0000 (23:29 -0500)]
tests: add nitg-s and nitg-e as specific engines
Notes:
* the base_gen* bug need to be investigated
* make nitg-s only run the base tests
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Mon, 3 Dec 2012 03:52:11 +0000 (22:52 -0500)]
tests: put stderr before stdout
The result files contain in order:
- compiler messages
- error output
- standard output
This improve the compatibility of res with niti since compiler
messages are issued on the error output.
Update the sav/* file to reflect the change.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Mon, 3 Dec 2012 03:52:44 +0000 (22:52 -0500)]
niti: write fatal error on stderr
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Mon, 3 Dec 2012 03:53:07 +0000 (22:53 -0500)]
tests: niti skips test_mem
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Mon, 3 Dec 2012 02:27:41 +0000 (21:27 -0500)]
nitg-e: remove autobox since the inherited one is identical
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Mon, 3 Dec 2012 02:27:06 +0000 (21:27 -0500)]
nitg-sep: BOX_XX does not require a type argument, the class is enough
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Mon, 3 Dec 2012 02:25:01 +0000 (21:25 -0500)]
nitg-sep: partially revert the last commit; do not change nitattribute_t
As usual, it is the little change that could not potentially break things
that break the build.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sun, 2 Dec 2012 19:55:24 +0000 (14:55 -0500)]
nitg-sep: improve attribute implementation for primitives
primitives are boxed, but the box is reused on write and the unboxed value
is returned on read.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sun, 2 Dec 2012 04:35:07 +0000 (23:35 -0500)]
nitg-e: type_test do it easy if the result is statically true
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sun, 2 Dec 2012 04:33:54 +0000 (23:33 -0500)]
nitg-e: type_test do not box the value.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sun, 2 Dec 2012 04:31:36 +0000 (23:31 -0500)]
nitg-sep: monomorphic_send if the mcasttype is primitive
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sun, 2 Dec 2012 04:30:03 +0000 (23:30 -0500)]
nitg: implements force_garbage_collection
Just call GC_gcollect
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sun, 2 Dec 2012 04:29:35 +0000 (23:29 -0500)]
nitg: stop displaying debug info on Unimplemented expr
Runtime still print NOT YET IMPLEMENTED however.
Signed-off-by: Jean Privat <jean@pryen.org>
Alexandre Terrasa [Sat, 1 Dec 2012 22:10:31 +0000 (17:10 -0500)]
nitg-sep: handles virtual types in typing and init_instance
Signed-off-by: Alexandre Terrasa <alexandre@moz-code.org>
Alexandre Terrasa [Sat, 1 Dec 2012 22:09:31 +0000 (17:09 -0500)]
nitg-sep: autobox self before call on self->type
Signed-off-by: Alexandre Terrasa <alexandre@moz-code.org>
Alexandre Terrasa [Sat, 1 Dec 2012 20:48:16 +0000 (15:48 -0500)]
nitg-sep: fix recursive unanchored type declarations
Signed-off-by: Alexandre Terrasa <alexandre@moz-code.org>
Jean Privat [Sat, 1 Dec 2012 14:40:25 +0000 (09:40 -0500)]
nitg: protect empty types `bugtype` in `send`.
abcbd319a121ec7c9fd3322e60fa5c2c1bb9ca46 moved the bug test after
the null check without protecting with brackets.
08627c569f04dc95108a445afb0152d293eebdf6 added a call to bugtype.
10ace83587bb20a66980c3a7be60f20e596e9ebe added an exit after the message,
but since there is no brackets, the exit is done when the receiver is null
a has a empty concrete type.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 1 Dec 2012 05:24:08 +0000 (00:24 -0500)]
nitg-sep: signature adaptation is done on the declared signature
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 1 Dec 2012 04:33:50 +0000 (23:33 -0500)]
nitg-sep: do not put attributes stuff in the class structure.
Attributes are already in the instance structures.
It seems that this piece of code comes from the global compiler where
there is no class structures, only instance ones.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 1 Dec 2012 04:32:43 +0000 (23:32 -0500)]
nitg-e: use -1 as hole in type_table
-1 is an invalid class id, 0 is the first valid class id.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 1 Dec 2012 04:31:36 +0000 (23:31 -0500)]
nitg: exit on bugtype
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Fri, 30 Nov 2012 21:11:39 +0000 (16:11 -0500)]
nitg-erase: erase MParameterType and workaround MVirtualType
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Fri, 30 Nov 2012 20:41:33 +0000 (15:41 -0500)]
nitg-erase: put 0 in arrays of int, not NULL
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Fri, 30 Nov 2012 20:26:54 +0000 (15:26 -0500)]
nitg-erase: reduce duplication with nitg-sep
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Fri, 30 Nov 2012 20:25:41 +0000 (15:25 -0500)]
nitg: use new_visitor instead of new XXXCompilerVisitor
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Fri, 30 Nov 2012 21:12:10 +0000 (16:12 -0500)]
nitg-erase: update sav/nitg.sav (option --erasure)
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Fri, 30 Nov 2012 18:50:31 +0000 (13:50 -0500)]
Merge remote-tracking branch 'alexandre/at/erasure'
Alexandre Terrasa [Fri, 30 Nov 2012 03:48:54 +0000 (22:48 -0500)]
nitg-erase: nitg with erased generic types
Signed-off-by: Alexandre Terrasa <alexandre@moz-code.org>
Jean Privat [Thu, 29 Nov 2012 21:42:26 +0000 (16:42 -0500)]
nitg: call bugtype on unexpected empty concrete type
The idea is to not execute silently dead code.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 29 Nov 2012 21:36:33 +0000 (16:36 -0500)]
nitg: check NPE on attribute access and call
NPE must also be checked on value with an empty concrete type.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 29 Nov 2012 21:42:15 +0000 (16:42 -0500)]
nitg: check null receiver if MNullableType or if MNullType
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 29 Nov 2012 21:34:55 +0000 (16:34 -0500)]
nitg: Visitor::send skip if the primitive receiver is dead
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 29 Nov 2012 21:32:57 +0000 (16:32 -0500)]
nitg: promote check_recv_notnull to nitg
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 29 Nov 2012 17:26:29 +0000 (12:26 -0500)]
nitg: explicit NOT YET IMPLEMENTED at runtime for nitni
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 29 Nov 2012 17:02:34 +0000 (12:02 -0500)]
nitg: C comment CustomizedRuntimeFunction indicate the type of "self"
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 29 Nov 2012 16:22:03 +0000 (11:22 -0500)]
nitg: type_test handle MNullType the same way than MNullableType
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 29 Nov 2012 15:17:49 +0000 (10:17 -0500)]
poset: implements "has" to avoid using the NaiveCollection implementation
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 29 Nov 2012 14:55:55 +0000 (09:55 -0500)]
tests: cleanup sav/ files for nitg&niti
Some result files are useless or misplaced.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 29 Nov 2012 14:08:10 +0000 (09:08 -0500)]
tests: nitg can compile empty modules
create a new method compile_main_function to factorize the work
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 29 Nov 2012 13:32:29 +0000 (08:32 -0500)]
nitg: generate a runtime error on on ADeferredMethPropdef
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 29 Nov 2012 04:20:10 +0000 (23:20 -0500)]
nitg: add type-checks on covariant parameters
The compilation of a MMethodDef will add required type checks before
compiling the body of the method.
The type checks are implemented trough v.type_test so this patch
is compatible with other engines.
problem: most nitg-sep tests become [todo] since virtual type tests
are not implemented is the engine but become mandatory in many methods
of Comparable.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 29 Nov 2012 04:08:33 +0000 (23:08 -0500)]
nitg: type_test assert that the cast-type was detected by RTA
This add nothing but might help to find a breakage of RTA or nitg.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 29 Nov 2012 04:06:34 +0000 (23:06 -0500)]
rta: consider type tests for covariant signatures
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 29 Nov 2012 17:01:27 +0000 (12:01 -0500)]
rta: forgot to call Array#with_native on ASuperstringExpr
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 29 Nov 2012 02:17:41 +0000 (21:17 -0500)]
nitg-sep: implements correctly class_name_string
1- do use the parameter instead of the string "self".
2- do not get the value at declaration time but at statement time.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 24 Nov 2012 05:28:02 +0000 (00:28 -0500)]
Merge remote-tracking branch 'alexandre/at/sep' into wip
Jean Privat [Sat, 24 Nov 2012 04:55:17 +0000 (23:55 -0500)]
nitg: allow casts on primitive types
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 24 Nov 2012 04:00:30 +0000 (23:00 -0500)]
tests: fix uninitialized attribute test_gen and test_boxing
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 24 Nov 2012 03:34:14 +0000 (22:34 -0500)]
tests: barf on the absence of Object is an acceptable outcome in base_no_object.nit
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 24 Nov 2012 03:33:10 +0000 (22:33 -0500)]
nitg-sep: detect null receiver on attribute access
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 24 Nov 2012 03:32:31 +0000 (22:32 -0500)]
tests: force Object on some tests.
What is the behavior on the absence of Object
is not something tested in those.
Signed-off-by: Jean Privat <jean@pryen.org>
Alexandre Terrasa [Sat, 24 Nov 2012 03:28:44 +0000 (22:28 -0500)]
tests: fix previously broken test base_attr_gen_alt1
Signed-off-by: Alexandre Terrasa <alexadre@moz-code.org>
Alexandre Terrasa [Sat, 24 Nov 2012 03:25:33 +0000 (22:25 -0500)]
nitg-sep: clean type_test_method
Signed-off-by: Alexandre Terrasa <alexadre@moz-code.org>
Alexandre Terrasa [Sat, 24 Nov 2012 03:25:12 +0000 (22:25 -0500)]
nitg-sep: make livetype retrieving recursive
Signed-off-by: Alexandre Terrasa <alexadre@moz-code.org>
Alexandre Terrasa [Sat, 24 Nov 2012 03:15:23 +0000 (22:15 -0500)]
nitg: enlarge visibility of global_compiler::MType::c_name_cache
Signed-off-by: Alexandre Terrasa <alexadre@moz-code.org>
Alexandre Terrasa [Fri, 23 Nov 2012 22:59:24 +0000 (17:59 -0500)]
nitg-sep: autobox primitive types before subtyping test
Signed-off-by: Alexandre Terrasa <alexadre@moz-code.org>
Alexandre Terrasa [Fri, 23 Nov 2012 22:03:52 +0000 (17:03 -0500)]
Merge branch 'master' of https://github.com/privat/nit into jp/sep
Alexandre Terrasa [Fri, 23 Nov 2012 21:58:50 +0000 (16:58 -0500)]
nitg-sep: check table color bounds
Signed-off-by: Alexandre Terrasa <alexadre@moz-code.org>
Jean Privat [Fri, 23 Nov 2012 21:17:57 +0000 (16:17 -0500)]
tests: count fixme as todo
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Fri, 23 Nov 2012 21:17:02 +0000 (16:17 -0500)]
tests: delete specific skip files
Move all skip rules to the $engine.skip file
Signed-off-by: Jean Privat <jean@pryen.org>
Alexandre Terrasa [Fri, 23 Nov 2012 03:59:47 +0000 (22:59 -0500)]
tests: move base_class_name from fixme to res
Signed-off-by: Alexandre Terrasa <alexadre@moz-code.org>
Alexandre Terrasa [Fri, 23 Nov 2012 03:59:11 +0000 (22:59 -0500)]
tests: move base_attr_gen_alt1 from fail to res
Signed-off-by: Alexandre Terrasa <alexadre@moz-code.org>
Jean Privat [Fri, 23 Nov 2012 03:46:15 +0000 (22:46 -0500)]
nitg-sep: handle send with varargs
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Fri, 23 Nov 2012 03:44:18 +0000 (22:44 -0500)]
nitg: extract varargize into its own method
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Fri, 23 Nov 2012 02:49:50 +0000 (21:49 -0500)]
nit-seg: global instances must be declared "extern" in the .h file
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 22 Nov 2012 20:33:38 +0000 (15:33 -0500)]
nitg-sep: workaround for covariant return type
Just force to use VIRTUAL function in order to force the implementations to
respond to the original signature from the introduction.
More work need to be done to cleanup and simplify the stuff.
Signed-off-by: Jean Privat <jean@pryen.org>
Alexandre Terrasa [Fri, 23 Nov 2012 01:32:39 +0000 (20:32 -0500)]
nitg-sep: make glob_sys global
Signed-off-by: Alexandre Terrasa <alexadre@moz-code.org>
Alexandre Terrasa [Fri, 23 Nov 2012 01:32:01 +0000 (20:32 -0500)]
nitg-sep: classnames
Signed-off-by: Alexandre Terrasa <alexadre@moz-code.org>
Alexandre Terrasa [Thu, 22 Nov 2012 22:02:08 +0000 (17:02 -0500)]
Merge branch 'master' of https://github.com/privat/nit into jp/master
Alexandre Terrasa [Thu, 22 Nov 2012 20:35:05 +0000 (15:35 -0500)]
nitg-sep: dirty workaround for NativeString
Signed-off-by: Alexandre Terrasa <alexadre@moz-code.org>
nitlanguage / nit.git / log