This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[Ada] implement many new Ada 2005 features

Tested under i686-linux, commited on mainline.

Major changes to add support for several Ada 2005 new features.
New versions of ACATS will include tests for these features.

addresses calls to inherited tagged primitives that inherit a default expression
given by a function call with a controlling result of the ancestor type.
In the case where the actual is classified as tag-indeterminate, the call
is defined to invoke the corresponding primitive of the descendant type.
Since this is equivalent to performing a dispatching call, we implement
it as dispatching. Since there is no extant actual for controlling the
dispatching, the tag attribute of the type associated with the outer
call is used as the controlling argument for the calls. Normally a
controlling tag is provided by a tagged object, but creating such an
object would be problematic in general, since the object might have
default initialization or require a discriminant constraint. Although
this capability for dispatching directly based on a tag is added here for
a rather specialized situation, it's anticipated that it will be useful
in the implementation of part of amendment AI-260, which defines a
notion of predefined generic dispatching constructors that require
dispatching to a formal function based on a value of type Ada.Tag.

makes it clear that an attempt to convert a null string from Ada
to C without null-termination must fail (because the result must have a
lower bound of zero, which is not possible for an unsigned index result
with no characters in the string. A previous patch (from quite a while
ago) had made this change for character, but not for wide character.

Implementation of the new Ada 2005 Stream_Size attribute as described in

Implementation of the new reserved words of Ada 2005 as described in

Support for wide wide characters
The full ai text that describes the new packages can be found in

This is one of the larger AI's from Ada 2005 in terms of impact on
the language and implementation, but won't affect most Ada programmers
very much (since we don't think too many programmers will be writing
programs using old italic letters or linear-B ideograms :-)

The most important point is that whereas Ada 95 supported only the
basic multilingual plane (corresponding to 16-bit Unicode, and the
type Wide_Character), in Ada 2005, the full 10646 standard is

Implement the new packages required by Ada 2005 AI 301. A full text of this
AI is found in

Most of the work is done in Strings.[Wide_[Wide_]]Unbounded.Text_IO. We always
implemented these routines in GNAT (well the Wide_Wide one is new :-) so
the new AI packages just rename the routines here (that's more convenient
anyway since these routines need to break the unbounded abstraction for
efficiency, bot not the text_io abstraction). Speaking of efficiency, we
only did this efficiency-breaking (using the aux package) in the non-wide
case. Now we have added the corresponding aux packages for wide and wide
wide cases as well.

The container packages from Ada 2005 AI-302 are added to the makefile
for the run-time and to the list of Ada 2005 units recognized by

Implement class-wide aggregate expression extensions as described in

A formal scalar type has non-static bounds, but is considered constrained.
In the instance, it is constrained by the bounds of the actuals, if any.
By constrast, T'Base is unconstrained. It is therefore consistent to say
that if T is a formal scalar type, it does not statically match T'Base.
This means that the following is illegal:
 generic type T is range <>;
 package P is procedure Proc (X : T); end;
 package body P is procedure Proc (X : T'Base) is begin null; end; end P;

Also, completion of the merge of PolyORB code generation for the DSA.

Finally, a new switch is added to gnatbind that allows specification of the
default exit status value. This is needed on certain systems such as
VMS where there are two alternative exit modes: Posix (where the success
code is 0) and VMS (where the sucess code is 1). Posix mode is used
to build and bootstrap the compiler and tools.  VMS mode is most often
used by application developers to retain compatibility with legacy code.

2005-02-09  Robert Dewar  <>
	    Thomas Quinot  <>
	    Javier Miranda  <>
	    Pascal Obry  <>
	    Ed Schonberg  <>
	    Doug Rupp  <>
	    Gary Dismukes  <>
	    Richard Kenner  <>

	* New file.

	*, a-chahan.adb: Add declarations from AI-285

	* Add pragma Ada_05 for wide_wide_space to get warning in
	Ada 95 mode
	Add definition of Wide_Wide_Space for AI-285

	*, impunit.adb, sem_ch10.adb: Complete rewrite and new
	interface (to support Ada 95 and Ada 2005 units).
	Add Unbounded_IO files
	Add entries for Wide_Wide packages for AI-285
	Add list of containers packages to Ada 2005 unit list

	*, a-swuwti.adb,, a-suteio.adb: Updates to
	support new Unbounded_IO package cleanly.

	*, g-utf_32.adb: New files.

	* Makefile.rtl: Add entry for g-utf_32
	Add new files for Unbounded_IO
	Adjust make file for new AI-285 wide wide packages
	Add AI-302 containers to the run time.

	* a-stwibo.adb,, a-stwisu.adb,,, a-strbou.adb,, a-strsup.adb: New
	subprograms for AI-301.

	* a-stwiun.adb, Minor reformatting.

	* Minor comment correction

	*, rtsfind.adb: Add definitions for Wide_Wide attributes
	Also extend Text_IO_Kludge to support Wide_Wide_Text_IO
	(Check_RPC): Update to match changes in expanded code.
	Clean up unused entity.

	*, exp_ch3.adb: Fix various places where Wide_Wide_String
	was not taken into account.
	This includes proper initialization with Normalize_Scalars.
	(Get_Simple_Init_Val): Major rewrite for initialize scalars and
	normalize scalars cases (particularly the latter) to do a better job
	of finding invalid representations.

	*, s-scaval.adb: Add values for zero invalid values

	*, s-strops.adb: Remove string normalize routines, never

	* exp_dist.adb: Add support for wide wide character type
	(Expand_Receiving_Stubs_Bodies): For a package declaration that has a
	private part, generate stub bodies at the end of the private part,
	not the visible part.
	(Add_RACW_Primitive_Operations_And_Bodies): Add last missing code for
	PolyORB support.
	(Add_Obj_RPC_Receiver_Completion): Add PCS-specific subprograms and
	generic wrapper to execute final processing after completing the
	expansion of the RPC receiver for an RACW.

	* snames.h,, snames.adb: Add definitions for wide_wide
	packages and attributes.
	(Preset_Names): Addition of the new reserved words of Ada 2005,
	that is interface, overriding and synchronized.
	(Get_Pragma_Id): Give support to the use of the new reserved word
	"interface" as a pragma name.
	(Is_Pragma_Name): Give support to the use of the new reserved word
	"interface" as a pragma name.
	(Preset_Names): Add stream_size string for the Stream_Size Ada2005
	attribute implementation.

	* exp_attr.adb (Expand_Attribute_Reference): Do not apply validity
	checks to entities that are output parameters of Asm operations.
	Handle the Stream_Size attribute.
	Add implementation of Wide_Wide_Value, Wide_Wide_Image, Wide_Wide_Width

	*, exp_imgv.adb: Add support for wide wide character type

	* sem_attr.adb (Eval_Attribute): Raise compile-time constraint error
	for second parameter being 0.0.
	Add support for wide wide character type.
	(Analyze_Attribute, Eval_Attribute): Handle the Stream_Size attribute.

	* s-valwch.adb,,, s-imgwch.adb,, s-wchstw.adb, s-wchwts.adb,,
	s-widwch.adb,, s-wwdcha.adb,,
	s-wwdenu.adb,, s-wwdwch.adb, Add support
	for wide wide character cases.

	* cstand.adb: Create entities for Wide_Wide_Character and

	*, i-c.adb: Fix not raising CE for null wide strings in
	accordance with AI-258.
	Add new declarations for 16/32 bit C character types (Part of AI285)

	*, einfo.adb (Is_Obsolescent, Is_Ada_2005): New flag
	(Obsolescent_Warning): New field
	(Rep_Clause): New local subprogram used to share code. Returns the rep
	clause for which the name is given in parameter.
	(Has_Stream_Size_Clause): New routine.
	(Stream_Size_Clause): Idem. Implementation is based on Rep_Clause.
	(Address_Clause): Implementation is now using Rep_Clause.
	(Alignment_Clause): Idem.
	(Size_Clause): Idem.

	* lib-xref.adb (Generate_Reference): Test for reference to Ada 2005
	entity in non-Ada 2005 mode and generate warning.

	* par-prag.adb: Add handling of one argument form for pragma Ada_05.
	(Prag): Code cleanup. Remove old gnat pragma "overriding"

	* sem_prag.adb: Add handling of one argument form for pragma Ada_05
	(Analyze_Pragma, case Elaborate, Elaborate_All): Do not disable warnings
	on the named unit if the pragma is not in the current compilation unit,
	so that elaboration calls in the current unit can set up an elaboration
	dependency on the named unit, as needed.
	(Analyze_Pragma, case Obsolescent): Allow pragma to be used for library
	subprogram as well as for subprograms declared within a package.
	(Analyze_Pragma, Sig_Flags): Code cleanup. Remove support for the GNAT
	pragma overriding.

	*, krunch.adb: Add special handling of Wide_Wide (krunched
	to z) to avoid some instances of duplication for Wide_Wide packages.

	*, namet.adb: Implement encoding (WWhhhhhhhh) for wide wide

	* scn.adb: Char_Literal_Value field is now a Uint

	* scng.adb: Significant rewrite to handle new Ada 2005 features
	allowing wide and wide wide characters in program text, e.g. for
	identifiers, as described in AI-285.
	(Set_Reserved): New procedure, makes setting up keywords cleaner.
	(Initialize_Scanner): Register the new reserved words of Ada 2005.
	(Scan): Give support to the new reserved words.

	* par-ch2.adb (P_Identifier): Compiling in Ada95 mode, generate a
	warning notifying that interface, overriding, and synchronized are
	new reserved words.
	(P_Pragma): Allow the use of the new reserved word "interface" as
	a pragma name.

	* gnatls.adb, gnatbind.adb,
	ali-util.adb, binde.adb,, ali.adb: Code cleanup. Rename
	identifiers named "interface" to "SAL_Interface".

	* bindgen.adb (Gen_Main_Ada): Add support for the new SEH
	(Structured Exception handling).
	(Gen_Main_C): Idem.

	* bindgen.adb: 
	(Gen_Main_Ada): Set the default exit code if specified.
	(Gen_Main_C): Likewise.
	Part of *DC20-006.
	(Gen_Output_File_C): Remove redundant output of gnat_exit_status.
	Code cleanup. Rename identifiers named "interface" to "SAL_Interface"

	* switch-b.adb, bindusg.adb,, Add handling of
	new -Xnnn switch.

	* mlib-prj.adb, mlib.adb: Code cleanup. Rename one identifier that
	has a collision with the new Ada 2005 "interface" reserved word.

	* par-ch3.adb (P_Defining_Identifier): Compiling in Ada95 mode,
	generate a warning notifying that interface, overriding, and
	synchronized are new reserved words.

	* (Token_Type): Addition of the tokens corresponding to the
	new reserved words of Ada 2005: Tok_Interface, Tok_Overriding
	and Tok_Synchronized.

	* sem_res.adb (Resolve_Actuals): Change error messages to refer to
	"dispatching" rather than "primitive" operations, since dispatching
	calls are now allowed to abstract formal subprograms (which are not
	Char_Literal_Value field is now a Uint
	(Resolve_Slice): If the prefix is an access to an unconstrained array,
	compute the actual subtype of the designated object to impose the proper
	index constraints.
	(Resolve_Selected_Component): Do not insert an access check if the
	prefix is an access type: such a node is expanded into an explicit
	dereference, on which the access check is performed anyway. Removes
	expensive duplicate checks.
	(Resolve_Call): Use new flag Is_Obsolescent and field
	Obsolescent_Warning so that pragma Obsolescent works on library
	Add support for wide wide character type
	(Resolve_Allocator): Replace the error message on wrong null-exclusion
	value by a warning message.
	(Resolve_Type_Conversion): If the mixed-mode expression is interpreted
	as fixed-point, and one of the operands is non-static and universal, it
	can only be an illegal exponentiation operation, in which case there is
	no real value to retrieve.

	* exp_strm.adb: Add support for wide wide character type
	(Build_Elementary_Input_Call): Compute the size of the stream element by
	querying the rep chain to find the Stream_Attribute attribute value.
	(Build_Elementary_Write_Call): Ditto.

	* sem_aggr.adb: Char_Literal_Value field is now a Uint
	Add support for wide wide character type
	Replace the error messages on wrong null-exclusion value by warnings
	as described in Ada 2005.
	(Resolve_Extension_Aggregate): Document the fact that the error
	message on class-wide expressions in extensions aggregates.

	* sem_case.adb: Add support for wide wide character type

	* sem_ch13.adb: Add support for wide wide character type
	(Analyze_Attribute_Definition_Clause): Handle the Stream_Size attribute.

	* sem_ch3.adb: Add support for wide wide character type
	(Process_Subtype): If constraint is illegal for the type, set Ekind of
	now-useless Itype, to prevent cascaded errors on a compiler built
	without -gnatp.

	* sem_ch8.adb: Add with and use of Sem_Disp.
	(Analyze_Subprogram_Renaming): Replace unclean uses of
	Corresponding_Spec with Corresponding_Formal_Spec (and delete setting
	of Corresponding_Spec to Empty).
	(Attribute_Renaming): Replace use of Corresponding_Spec with
	Corresponding_ Formal_Spec and simplify condition.
	(Use_One_Package): Check that scope of homonym of identifier is defined,
	before checking whether it is a wrapper package.
	Add support for wide wide character type

	* sem_eval.adb: Add support for wide wide character type.
	(Eval_Arithmetic_Op): Check for compile time known signed integer
	overflow in the non-static case.
	(Subtypes_Statically_Match): A formal scalar type and its base type do
	not statically match.

	* sem_util.adb (Collect_Primitive_Operations): Minor change of "/=" to
	"not in" for test of N_Formal_Subprogram_Declaration (which is now a
	(Unit_Declaration_Node): Ditto.
	(Is_Variable_Prefix):  For the case of an indexed component whose prefix
	has a packed array type, the prefix has been rewritten into a type
	conversion. Determine variable-ness from the converted expression.
	Handle wide wide character cases.

	* Add types Wide_Wide_Character and Wide_Wide_String

	*, stringt.adb: Handle full UTF-32 range.
	Remove ["0A"] from comment, since it can look like a line terminator.
	Currently we don't permit this, but this is under discussion by the
	ARG, and it is easy enough to use a different example.

	*,, s-wchcnv.adb: Add new subprograms for
	handling UTF-32 encoding for wide wide character.
	Implement new brackets coding ["hhhhhhhh"]
	Add UTF-8 encodings for full UTF-32 range

	* Add definition of Standard_Wide_Wide_Character_Size

	* types.h,, types.adb: Wide_Wide_Character now has full 31
	bit range Add full UTF-32 support.
	(RT_Exception_Code): Addition of CE_Null_Not_Allowed; used to
	notify that constraint error will be raised at run-time
	because a null value is assigned to a null-excluding object.
	Remove some obsolete declarations and make Char_Code

	* a-except.adb (Rcheck_30): New subprogram. Addition of the message
	corresponding to CE_Null_Not_Allowed, and adjust the output of all the
	Rcheck subprograms.

	* checks.adb (Check_Null_Not_Allowed): Replace the error message on
	wrong null-exclusion value by a warning message.
	(Enable_Range_Check): Do range check if the prefix is an
	explicit dereference whose designated object is an unconstrained array.
	Current algorithm for removing duplicate checks is over-eager in this

	* sem_ch5.adb (Analyze_Assignment): Replace the error messages on wrong
	null-exclusion value by a warning message

	* atree.h,, atree.adb: Remove Char_Code field support
	completely. Add support for Uint2 field

	sem_ch2.adb, exp_ch11.adb, exp_dbug.adb,
	exp_prag.adb: Char_Literal_Value field is now a Uint.

	* exp_util.adb (Insert_Actions): Replace
	N_Formal_Subprogram_Declaration by
	Char_Literal_Value field is now a Uint.

	*, sinfo.adb (Corresponding_Formal_Spec): New function
	defined for subprogram renaming declarations. When set, the field
	indicates the defining entity of a corresponding formal subprogram
	when the renaming corresponds to a formal subprogram association in an
	(Set_Corresponding_Formal_Spec): New procedure to return
	Corresponding_Formal_Spec field.
	Minor changes of "=" to "in" in tests of N_Formal_Subprogram_Declaration
	(which is now a subtype).
	Char_Literal_Value field is now a Uint

	*, exp_disp.adb (Make_DT): Generate code that moves the
	pointer to the base of the dispatch table.
	Minor changes to comments.
	(Controlling_Type): New function for determining the tagged type
	associated with a tagged primitive subprogram.
	(Expand_Dispatching_Call): Add support for a controlling actual that is
	directly a value of type Ada.Tag rather than a tagged object.

	*, i-cpp.adb,, a-tags.adb: Update documentation
	describing the new layout.
	(Dispatch_Table): The expander computes the actual array size, allocates
	the Dispatch_Table record accordingly, and generates code that displaces
	the base of the record after the Typeinfo_Ptr component. The access to
	these components is done by means of local functions.
	(Offset_To_Top): New function.
	(Typeinfo_Ptr): New function.
	(Get_TSD): Modified to access the new position of the TSD.
	(Set_TSD): Modified to save the TSD in its new position.

	* par-ch12.adb (P_Formal_Subprogram_Declaration): Add parsing for the
	case of formal abstract subprograms. Add check and message for -gnat05.
	Update comments.

	* sem_ch12.adb: Add with and use for Sem_Disp.
	(Analyze_Associations): Minor change from "=" to "in" for use of
	N_Formal_Subtype_Declaration (which is now a subtype).
	(Set_Analyzed_Formal): Minor changes from "=" to "in" for uses of
	N_Formal_Subtype_Declaration (which is now a subtype).
	(Analyze_Formal_Subprogram): Add handling for
	N_Formal_Abstract_Subprogram, marking the formal as abstract and
	dispatching, setting the controlling status of the formal parameters
	and result, and issuing an error if there is no controlling type for
	the formal subprogram.
	(Instantiate_Formal_Subprogram): Rather than setting Corresponding_Spec,
	which is an unclean use of that field, we set the new field
	Corresponding_Formal_Spec to make the formal subprogram available to
	processing in Analyze_Subprogram_Declaration.
	(Analyze_Formal_{Discrete, Decimal_Fixed_Point, Fixed_Point,
	Floating_Point, Modular_Integer, Signed_Integer}_Type: Make formal type
	Constrained, so that it is is does not statically match its anonymous
	base type.

	* sem_ch6.adb (Analyze_Subprogram_Specification): Include test for
	abstract formal subprograms in error check for functions returning
	abstract types. Set scope of new designator for
	a parameterless subprogram, so that it is available when checking the
	body for nested subprograms, before full analysis of said body.
	(Analyze_Subprogram_Body): Warn on inlining bodies with nested
	subprogram only if inner one comes from source.
	(Analyze_Function_Call): If the call is given in object notation, the
	analysis of the name rewrites the node and analyzes it with the proper
	argument list. After analyzing the name, if the call has been rewritten
	and the result type is set, no further analysis is needed.
	(Analyze_Return_Type): Subsidiary to Process_Formals: analyze subtype
	mark in function specification, in a context where the formals are
	visible and hide outer homographs.

	* sem_disp.adb (Check_Controlling_Type): Relax the check for same scope
	as the tagged type for the cases of abstract formal subprograms and
	renamings of those.  Clean up spec comments.
	(Check_Dispatching_Context): Add error message to indicate "abstract
	procedure", covering the case of a call to a formal abstract procedure
	that has statically tagged operands.
	(Check_Dispatching_Call): Check for the case of an actual given by
	a tag-indeterminate function call whose type is an ancestor of the
	containing call's associated tagged type. This situation can occur
	for inherited primitives with function defaults. In this case we
	use the tagged type's tag directly as the controlling argument for
	the calls.
	(Expand_Call): Name change on call to Expand_Dispatch_Call.

	* sprint.adb (Sprint_Node_Actual): Split
	N_Formal_Subprogram_Declaration into two alternatives for the new
	cases N_Formal_Abstract_Subprogram_Declaration and
	Char_Literal_Value field is now a Uint.

	* trans.c: Get rid of junk Uint2 reference.
	Char_Literal_Value field is now a Uint.
	(gnat_to_gnu, case N_Aggregate): Check TYPE_UNCHECKED_UNION_P.
	(gigi): Correct third arg to gimplify_body.

	* ada-tree.h: (TYPE_UNCHECKED_UNION_P): New flag.
	(TYPE_LANG_FLAG_0): Check for record or union.

	* treepr.adb: Char_Literal_Value field is now a Uint

	* uintp.h,, uintp.adb: Add new routines UI_To_CC and

	*, widechar.adb (Is_UTF_32_Non_Graphic): New function
	Add full UTF-32 support
	Char_Code is now 32 bits

	*, sinput.adb (Skip_Line_Terminators): Extend to deal with
	wide character UTF_32 line terminators.
	Initialize Main_Source_File to avoid error when no main
	source is loaded.

	* errout.adb (Finalize): Do not check Num_SRef_Pragmas
	(Main_Source_File) when no main source has been loaded, to avoid
	potential crash.

Attachment: difs.4.bz2
Description: Binary data

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]