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]

committed: Ada updates

Tested on x86-linux

Sorry for the big update in one chunk, I was in vacation last week, and
I don't have any script to split these commits in smaller chunks (nor
do I see any simple way to achieve such task reliably).

If people have scripts that work reliably to be shared, feel free to share

2004-03-15  Jerome Guitton  <>

	* Fix multicast options.

	* Move unchecked conversion from ATSD_Access to Address
	in the spec.

2004-03-15  Robert Dewar  <>

	* sem_prag.adb: Make sure No_Strict_Aliasing flag is set right when
	pragma used for a private type.

	* lib-xref.adb (Generate_Reference): Do not generate warning if
	reference is in a different unit from the pragma Unreferenced.

	* 5vtpopde.adb: Minor reformatting
	Fix casing of To_Task_ID

	* sem_ch13.adb (Validate_Unchecked_Conversion): Set No_Strict_Aliasing
	flag if we have an unchecked conversion to an access type in the same

2004-03-15  Geert Bosch  <>

	* a-ngcoty.adb (Modulus): In alternate formula for large real or
	imaginary parts, use Double precision throughout.

	* a-tifiio.adb (Put_Scaled): Remove remaining pragma Debug. Not only
	we want to be able to compile run-time with -gnata for testing, but
	this may also be instantiated in user code that is compiled with -gnata.

2004-03-15  Olivier Hainque  <>

	* (Exception_Code): New type, to represent Import/Export
	codes. Having a separate type for this is useful to enforce consistency
	throughout the various run-time units.
	(Exception_Data): Use Exception_Code for Import_Code.

	*, s-vmexta.adb: Use Exception_Code instead of a mix of
	Natural and Integer in various places.
	(Register_VMS_Exception): Use Base_Code_In to compute the exception code
	with the severity bits masked off.
	(Register_VMS_Exception): Handle the additional exception data pointer

	* raise.c (_GNAT_Exception structure): Remove the handled_by_others
	component, now reflected by an exported accessor.
	(is_handled_by): New routine to compute whether the propagated
	occurrence matches some handler choice specification. Extracted out of
	get_action_description_for, and expanded to take care of the VMS
	(get_action_description_for): Use is_handled_by instead of an explicit
	complex condition to decide if the current choice at hand catches the
	propagated occurrence.

	* raise.h (Exception_Code): New type for C.

	* (RE_Id, RE_Unit_Table): Add
	System.Standard_Library.Exception_Code, to allow references from the
	pragma import/export expander.

	* a-exexpr.adb (Is_Handled_By_Others, Language_For, Import_Code_For):
	New accessors to allow easy access to GNAT exception data
	(GNAT_GCC_Exception record, Propagate_Exception): Get rid of the
	redundant Handled_By_Others component, helper for the personality
	routine which will now be able to call the appropriate exception data
	accessor instead.

	* cstand.adb (Create_Standard): Adjust the type of the Import_Code
	component of Standard_Exception_Type to be the closest possible to
	Exception_Code in System.Standard_Library, that we cannot get at this
	point. Expand a ??? comment to notify that this type node should
	probably be rewritten later on.

	* exp_prag.adb (Expand_Pragma_Import_Export_Exception): Adjust the
	registration call to include a pointer to the exception object in the

	* init.c (__gnat_error_handler): Use Exception_Code and Base_Code_In
	instead of int and explicit bitmasks.

2004-03-15  Vincent Celier  <>

	* Add new GNAT BIND qualifier /STATIC. Makes /NOSHARED
	equivalent to /STATIC and /NOSTATIC equivalent to /SHARED.

	* a-tasatt.adb (To_Access_Code): Remove this UC instantiation, no
	longer needed now that it is in the spec of

	* adaint.h, adaint.c: (__gnat_create_output_file): New function

	* gnatcmd.adb: Fix bug introduced in previous rev: /= instead of =

	*, g-os_lib.adb (Create_Output_Text_File): New function.

	* make.adb (Gnatmake): Do not check the executable suffix; it is being
	taken care of in Scan_Make_Arg.
	(Scan_Make_Arg): Add the executable suffix only if the argument
	following -o, in canonical case, does not end with the executable
	suffix.  When in verbose mode and executable file name does not end
	with executable suffix, output the executable name, in canonical case.

	* (Access_Dummy_Wrapper): Add pragma No_Strict_Aliasing
	to avoid warnings when instantiating Ada.Task_Attributes.
	Minor reformating.

	* mlib-prj.adb (Process_Imported_Libraries): Get the imported libraries
	in the correct order.

	* prj-makr.adb (Process_Directory): No longer use GNAT.Expect, but
	redirect standard output and error to a file for the invocation of the
	compiler, then read the file.

	* prj-nmsc.adb (Find_Sources): Use the Display_Value for each
	directory, instead of the Value.
	(Find_Source_Dirs): Remove useless code & comments.

2004-03-15  Ed Schonberg  <>

	* exp_ch3.adb (Freeze_Record_Type): If a primitive operation of a
	tagged type is inherited, and the parent operation is not frozen yet,
	force generation of a freeze node for the inherited operation, so the
	corresponding dispatch entry is properly initialized.
	(Make_Predefined_Primitive_Specs): Check that return type is Boolean
	when looking for user-defined equality operation.

	* exp_ch4.adb (Expand_Composite_Equality): Check that return type is
	boolean when locating primitive equality of tagged component.

	* exp_ch5.adb (Expand_Assign_Array): If the left-hand side is a
	bit-aligned field and the right-hand side a string literal, introduce
	a temporary before expanding assignment into a loop.

	* exp_ch9.adb (Expand_N_Task_Type_Declaration): Copy expression for
	priority in full, to ensure that any expanded subepxressions of it are
	elaborated in the scope of the init_proc.

	* exp_prag.adb (Expand_Pragma_Import): Search for initialization call
	after object declaration, skipping over code that may have been
	generated for validity checks.

	* sem_ch12.adb (Validate_Private_Type_Instance): If type has unknown
	discriminants, ignore the known discriminants of its full view, if
	any, to check legality.

	* sem_ch3.adb (Complete_Private_Subtype): Do not create constrained
	component if type has unknown discriminants.
	(Analyze_Private_Extension_Declaration): Discriminant constraint is
	null if type has unknown discriminants.

	* sem_ch6.adb (Analyze_Generic_Subprogram_Body): Generate reference
	for end label when present.

	* s-fileio.adb (Open): When called with a C_Stream, use given name for
	temporary file, rather than an empty string.

2004-03-15  Ed Falis  <>

	* s-thread.adb: Removed, no longer used.

2004-03-15  Richard Kenner  <>

	* decl.c (target.h): Now include.
	(gnat_to_gnu_entity, case E_Access_Type): Use mode derived from ESIZE
	in new build_pointer_from_mode calls for non-fat/non-thin pointer.
	(validate_size): For POINTER_TYPE, get smallest size permitted on

	* fe.h: Sort Einfo decls and add Set_Mechanism.

	* (LIBGNAT_SRCS): Remove types.h.
	(ada/decl.o): Depends on target.h.

	* trans.c (tree_transform, N_Unchecked_Type_Conversion): Do not use

2004-03-15  Thomas Quinot  <>

	* 5ztpopsp.adb, 56tpopsp.adb: Fix spelling of Task_ID.

	* exp_ch4.adb (Expand_N_Indexed_Component): Do not call
	Insert_Dereference_Action when rewriting an implicit dereference into
	an explicit one, this will be taken care of during expansion of the
	explicit dereference.
	(Expand_N_Slice): Same. Always do the rewriting, even for the case
	of non-packed slices, since the dereference action generated by
	expansion of the explicit dereference is needed in any case.
	(Expand_N_Selected_Component): When rewriting an implicit dereference,
	analyze and resolve the rewritten explicit dereference so it is seen
	by the expander.
	(Insert_Dereference_Action): This procedure is now called only for the
	expansion of an N_Explcit_Dereference_Node. Do insert a check even for
	dereferences that do not come from source (including explicit
	dereferences resulting from rewriting implicit ones), but do not
	recursively insert a check for the dereference nodes contained within
	the check.
	(Insert_Dereference_Action): Clarify and correct comment.

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