2005-11-14 Gary Dismukes <dismukes@adacore.com>
Ed Schonberg <schonberg@adacore.com>
Gary Dismukes <dismukes@adacore.com>
Thomas Quinot <quinot@adacore.com>
* sem_ch12.ads, sem_ch12.adb (Map_Entities): Exclude entities whose
names are internal, because they will not have a corresponding partner
in the actual package.
(Analyze_Formal_Package): Move the setting of the formal package spec's
Generic_Parent field so that it occurs prior to analyzing the package,
to allow proper operation of Install_Parent_Private_Declarations.
(Analyze_Package_Instantiation): Set the instantiated package entity's
Package_Instantiation field.
(Get_Package_Instantiation_Node): Move declaration to package spec.
Retrieve the N_Package_Instantiation node when the Package_Instantiation
field is present.
(Check_Generic_Child_Unit): Within an inlined call, the only possible
instantiation is Unchecked_Conversion, for which no parents are needed.
(Inline_Instance_Body): Deinstall and record the use_clauses for all
parent scopes of a scope being removed prior to inlining an instance
body.
(Analyze_Package_Instantiation): Do not perform front-end inlining when
the current context is itself an instance within a non-instance child
unit, to prevent scope stack errors.
(Save_References): If the node is an aggregate that is an actual in a
call, rewrite as a qualified expression to preserve some type
information, to resolve possible ambiguities in the instance.
(Instance_Parent_Unit): New global variable to record the ultimate
parent unit associated with a generic child unit instance (associated
with the existing Parent_Unit_Visible flag).
(type Instance_Env): New component Instance_Parent_Unit for stacking
parents recorded in the global Instance_Parent_Unit.
(Init_Env): Save value of Instance_Parent_Unit in the Instance_Env
stack.
(Install_Spec): Save the parent unit entity in Instance_Parent_Unit when
it's not a top-level unit, and only do this if Instance_Parent_Unit is
not already set. Replace test of Is_Child_Unit with test of parent's
scope against package Standard. Add comments and a ??? comment.
(Remove_Parent): Revise condition for resetting Is_Immediately_Visible
on a child instance parent to test that the parent equals
Instance_Parent rather than simply checking that the unit is not a
child unit.
(Restore_Env): Restore value of Instance_Parent_Unit from Instance_Env.
(Validate_Derived_Interface_Type_Instance): Verify that all ancestors of
a formal interface are ancestors of the corresponding actual.
(Validate_Formal_Interface_Type): Additional legality checks.
(Analyze_Formal_Derived_Interface_Type): New procedure to handle formal
interface types with ancestors.
(Analyze_Formal_Package): If formal is a renaming, use renamed entity
to diagnose attempts to use generic within its own declaration.
From-SVN: r106999
2005-11-14 Robert Dewar <dewar@adacore.com>
Ed Schonberg <schonberg@adacore.com>
* sem_ch10.adb: Change name Is_Package to Is_Package_Or_Generic_Package
Do not give obsolescent warning on with of subprogram (since we
diagnose calls)
(Analyze_With_Clause): Add test for obsolescent package
(Install_Context_Clauses): If the unit is the body of a child unit, do
not install twice the private declarations of the parents, to prevent
circular lists of Use_Clauses in a parent.
(Implicit_With_On_Parent): Do add duplicate with_clause on parent when
compiling body of child unit.
Use new class N_Subprogram_Instantiation
(Expand_With_Clause): If this is a private with_clause for a child unit,
appearing in the context of a package declaration, then the implicit
with_clauses generated for parent units are private as well.
(License_Check): Do not generate message if with'ed unit is internal
From-SVN: r106998
2005-11-14 Ed Schonberg <schonberg@adacore.com>
* sem_aggr.adb (Resolve_Aggregate): An others choice is legal on the
rhs of an assignment even if the type is unconstrained, when the
context is non-expanding.
In an inlined body, if the context type is private,
resolve with its full view, which must be a composite type.
From-SVN: r106997
2005-11-14 Emmanuel Briot <briot@adacore.com>
* prj-pp.adb (Print): Do not output the with statement if the
associated name is empty, which happens for virtual extending projects.
(Print): Preserve the "extends all" attribute when printing the project.
* prj-tree.ads (String_Value_Of): Add comment about returned value for
a virtual extending project.
From-SVN: r106996
2005-11-14 Vincent Celier <celier@adacore.com>
* prj-part.adb (Create_Virtual_Extending_Project): Put virtual project
into Prj.Tree.Tree_Private_Part.Projects_Htable for GPS.
From-SVN: r106995
2005-11-14 Vincent Celier <celier@adacore.com>
* prj-ext.adb: Take into account new environment variable
GPR_PROJECT_PATH. Warn if both GPR_PROJECT_PATH and ADA_PROJECT_PATH
are defined.
(Prj.Ext elaboration): For each directory in the ADA_PROJECT_PATH,
normalize its path name, making it absolute and resolving symbolic
links, and replace the original if resolved path is different.
From-SVN: r106994
2005-11-14 Robert Dewar <dewar@adacore.com>
* par-ch10.adb (P_Compilation_Unit): Add defenses against junk unit
syntax, which could cause compiler hangs.
From-SVN: r106993
2005-11-14 Pascal Obry <obry@adacore.com>
* mdll.adb (Build_Reloc_DLL): Fix parameter handling when a map file is
used.
(Ada_Build_Reloc_DLL): Fix parameter handling when a map file is used.
In both cases the last argument was dropped.
From-SVN: r106991
2005-11-14 Vincent Celier <celier@adacore.com>
* make.adb (Compile_Sources): Change verbose message to minimum
verbosity level High for "is in an Ada library", "is a read-only
library" and "is an internal library",
(Create_Binder_Mapping_File): Path name of ALI file for library project
must include the library directory, not the object directory.
(Scan_Make_Arg): Make sure that Switch.M.Scan_Make_Switches is called
for new switches -vl, -vm and -vh.
(Verbose_Msg): Add new defaulted parameter Minimum_Verbosity
(Check): Use minimum verbosity Medium for some Verbose_Msg calls
(Compile_Sources): Do not attempt to compile if an ALI file is missing
in a project that is externally built.
(Compute_All_Imported_Projects): New procedure
(Gnatmake): Check if importing libraries should be regenerated because
at least an imported library is more recent.
(Initialize): For each project compute the list of the projects it
imports directly or indirectly.
(Add_Library_Search_Dir): New procedure, used in place of
Add_Lib_Search_Dir in procedure Scan_Make_Arg so that absolute paths are
put in the search paths.
(Add_Source_Search_Dir): New procedure, used in place of
Add_Src_Search_Dir in procedure Scan_Make_Arg so that absolute paths are
put in the search paths.
(Mark_Directory): Resolve the absolute path the directory before marking
it.
* switch-m.adb (Scan_Make_Switches): Replace "raise Bad_Switch;" with
call to new procedure Bad_Switch. Call Scan_Pos with new parameter
Switch. Do not handle any exception.
(Scan_Make_Switches): Increment Ptr for new switches -vl, -vm and -vh
so that the switch is recognized as valid.
(Scan_Make_Switches): Implement new switches -vl, -vm and -vh.
From-SVN: r106989
2005-11-14 Ed Schonberg <schonberg@adacore.com>
* lib-xref.adb (Generate_Definition, Generate_Reference): Treat the
internal entity created for the declaration of a child subprogram body
with no spec as coming from source, to generate proper cross-reference
information.
From-SVN: r106988
2005-11-14 Robert Dewar <dewar@adacore.com>
* interfac.ads: Change declarations of IEEE float types so that we no
longer need a separate version of this package for VMS.
From-SVN: r106987
2005-11-14 Robert Dewar <dewar@adacore.com>
Vincent Celier <celier@adacore.com>
* gprep.adb: Implement -C switch to scan comments
* scng.adb: Scan comment symbol separately if Replace_In_Comments set
* scans.ads: Comment updates (including new use of Tok_Comment in
preprocessing)
* opt.ads: Add documentation for flags that are used by gprmake,
currently and in the next version of gprmake.
(Verbosity_Level): New variable
Add Replace_In_Comments switch
* vms_data.ads: Add VMS equivalent for new gnatmake switches -vl, -vm
and -vm.
Add /REPLACE_IN_COMMENTS for gnatprep -C switch
From-SVN: r106983
2005-11-14 Pascal Obry <obry@adacore.com>
* gnatbind.adb (Is_Cross_Compiler): New function returning True for
cross-compiler.
(Scan_Bind_Arg): Fail with an error message if -M option is used
on a native compiler.
From-SVN: r106982
2005-11-14 Robert Dewar <dewar@adacore.com>
* g-debpoo.adb (Set_Valid): Use Integer_Address instead of
Storage_Offset to avoid wrap around causing invalid results.
From-SVN: r106981
2005-11-14 Ed Schonberg <schonberg@adacore.com>
* freeze.adb (Freeze_Entity): For an access formal that is an access
to subprogram, freeze the anonymous subprogram type at the same time,
to prevent later freezing in the wrong scope, such as the enclosing
subprogram body.
(Freeze_Entity): Freeze the equivalent_type of an access_to_protected_
subprogram whenever available.
From-SVN: r106979
2005-11-14 Javier Miranda <miranda@adacore.com>
* exp_strm.adb (Build_Stream_Attr_Profile, Build_Stream_Function,
Build_Stream_Procedure): Add the null-excluding attribute to the first
formal.
This has no semantic meaning under Ada95 mode but it is a
requirement under Ada05 mode.
* par-ch3.adb (P_Access_Definition): Addition of warning message if
the null exclusion is used under Ada95 mode
(P_Null_Exclusion): The qualifier has no semantic meaning in Ada 95.
(P_Access_Definition): Remove assertion that forbids the use of
the null-exclusion feature in Ada95.
From-SVN: r106977
2005-11-14 Thomas Quinot <quinot@adacore.com>
Robert Dewar <dewar@adacore.com>
Ed Schonberg <schonberg@adacore.com>
* exp_intr.adb (Expand_Unc_Deallocation): If GIGI needs an actual
subtype to compute the size of the designated object at run-time,
create such a subtype and store it in the Actual_Designated_Subtype
attribute of the N_Free_Statement.
Generate itype for classwide designated object in both cases of
user-specified storage pool: specific and class-wide, not only in the
specific case.
Raise CE when trying to set a not null access type object to null.
(Expand_Dispatching_Constructor_Call): Retrieve subprogram actual with
an explicit loop, because freeze nodes make its position variable.
* sem_intr.adb (Check_Intrinsic_Call): Given warning for freeing not
null object.
From-SVN: r106976
2005-11-14 Robert Dewar <dewar@adacore.com>
* exp_fixd.adb: Use Universal_Real instead of Long_Long_Float when we
need a high precision float type for the generated code (prevents
gratuitous Vax_Float stuff when pragma Float_Representation (Vax_Float)
used).
* exp_imgv.adb: Use Universal_Real instead of Long_Long_Float when we
need a high precision float type for the generated code (prevents
gratuitous Vax_Float stuff when pragma Float_Representation (Vax_Float)
used).
(Expand_Width_Attribute): In configurable run-time, the attribute is not
allowed on non-static enumeration subtypes. Force a load error to emit
the correct diagnostic.
From-SVN: r106975
2005-11-14 Pascal Obry <obry@adacore.com>
* expect.c (__gnat_kill) [Win32]: Fix implementation, the pid returned
by spawnve is a process handle, no need to convert. Add a parameter
close to control wether the process handle must be closed.
(__gnat_waitpid): Fix implementation, the pid returned by spawnve is
a process handle, not need to convert.
(__gnat_kill) [*]: Add dummy parameter close to match the Win32 spec.
* g-expect.adb: (Kill): Document the new close parameter.
(Close): Do not release the process handle in the kill there as
waitpid() is using it.
(Send_Signal): Release the process handle.
From-SVN: r106974
2005-11-14 Thomas Quinot <quinot@adacore.com>
* exp_dist.adb (Append_Array_Traversal): Modify constrained case to
generate a set of nested array aggregates instead of a single flat
aggregate for multi-dimensional arrays.
From-SVN: r106973
2005-11-14 Cyrille Comar <comar@adacore.com>
Thomas Quinot <quinot@adacore.com>
* exp_ch5.adb (Expand_N_Assignment_Statement, Tagged_Case): For an
assignment of a value of a tagged type that has been rewritten to a
block statement, it is known by construction that no checks are
necessary for the statements within the block: analyze it with checks
suppressed.
(Expand_N_If_Statement): When killing a dead then-branch in an
if-statement that has elsif_parts, recompute the Current_Value node
for any entity whose value is known from the condition of the first
elsif_part.
(Expand_N_Return_Statement): When returning a mutable record, convert
the return value into its actual subtype in order to help the backend
to return the actual size instead of the maximum. This is another
aftermath of not returning mutable records on the sec-stack anymore.
* sem_ch5.ads, sem_ch5.adb (Analyze_Iteration_Scheme): Minor change to
handling of error msg for suspicious reverse range iteration.
(Check_Possible_Current_Value_Condition): Move declaration from body to
spec, to allow this subprogram to be called from exp_ch5.
From-SVN: r106972
2005-11-14 Javier Miranda <miranda@adacore.com>
Robert Dewar <dewar@adacore.com>
Hristian Kirtchev <kirtchev@adacore.com>
* sem_disp.adb: Change name Is_Package to Is_Package_Or_Generic_Package
(Check_Dispatching_Operation): Protect the frontend againts
previously detected errors.
* Makefile.rtl: Add new instantiations of system.fat_gen
* s-fatflt.ads, s-fatlfl.ads, s-fatllf.ads, s-fatsfl.ads:
Change name of instantiated package for better consistency
with newly added system.fat_gen instantiations.
* s-filofl.ads, s-fishfl.ads, s-fvadfl.ads, s-fvaffl.ads,
s-fvagfl.ads: New files.
From-SVN: r106971
2005-11-14 Robert Dewar <dewar@adacore.com>
Hristian Kirtchev <kirtchev@adacore.com>
* sem_attr.adb: Implement Machine_Rounding attribute
(Analyze_Access_Attribute): The access attribute may appear within an
aggregate that has been expanded into a loop.
(Check_Task_Prefix): Add semantic check for attribute 'Callable and
'Terminated whenever the prefix is of a task interface class-wide type.
(Analyze_Attribute): Add semantic check for attribute 'Identity whenever
the prefix is of a task interface class-wide type.
* s-vaflop-vms-alpha.adb: Valid_D, Valid_F, Valid_G: Make Val constant
to avoid warnings.
* s-fatgen.ads, s-fatgen.adb (Machine_Rounding): New function
Remove pragma Inline for [Unaligned_]Valid.
Add comments that Valid routines do not work for Vax_Float
* exp_attr.adb: Implement Machine_Rounding attribute
* snames.h: Add entry for Machine_Rounding attribute
From-SVN: r106970
2005-11-14 Ed Schonberg <schonberg@adacore.com>
Cyrille Comar <comar@adacore.com>
* exp_aggr.adb (Build_Record_Aggr_Code): Do not create master entity
for task component, in the case of a limited aggregate. The enclosed
object declaration will create it earlier. Otherwise, in the case of a
nested aggregate, the object may appear in the wrong scope.
(Convert_Aggr_In_Object_Decl): Create a transient scope when needed.
(Gen_Assign): If the component being assigned is an array type and the
expression is itself an aggregate, wrap the assignment in a block to
force finalization actions on the temporary created for each row of the
enclosing object.
(Build_Record_Aggr_Code): Significant rewrite insuring that ctrl
structures are initialized after all discriminants are set so that
they can be accessed even when their offset is dynamic.
From-SVN: r106969
2005-11-14 Robert Dewar <dewar@adacore.com>
Ed Schonberg <schonberg@adacore.com>
* sem_elab.adb: Change name Is_Package to Is_Package_Or_Generic_Package
(Check_Elab_Call): A call within a protected body is never an
elaboration call, and does not require checking.
(Same_Elaboration_Scope): Take into account protected types for both
entities.
(Activate_Elaborate_All_Desirable): New procedure
* ali.ads, ali.adb: Implement new AD/ED for Elaborate_All/Elaborate
desirable
* binde.adb: Implement new AD/ED for Elaborate_All/Elaborate desirable
(Elab_Error_Msg): Use -da to include internal unit links, not -de.
* lib-writ.ads, lib-writ.adb:
Implement new AD/ED for Elaborate_All/Elaborate desirable
Use new Elaborate_All_Desirable flag in N_With_Clause node
* sinfo.ads, sinfo.adb (Actual_Designated_Subtype): New attribute for
N_Free_Statement nodes.
Define new class N_Subprogram_Instantiation
Add Elaborate_Desirable flag to N_With_Clause node
Add N_Delay_Statement (covering two kinds of delay)
* debug.adb: Introduce d.f flag for compiler
Add -da switch for binder
From-SVN: r106968
2005-11-14 Vincent Celier <celier@adacore.com>
* clean.adb (Check_Project): Look for Ada code in extending project,
even if Ada is not specified as a language.
Use new function DLL_Prefix for DLL_Name
(Clean_Interface_Copy_Directory): New procedure
(Clean_Library_Directory): New procedure
(Clean_Directory): Remove procedure, no longer used
(Clean_Project): Do not delete any file in an externally built project
* prj-env.adb (Set_Ada_Paths.Add.Recursive_Add): Add the object
directory of an extending project, even when there are no Ada source
present.
(Ada_Objects_Path.Add): Add Library_ALI_Dir, not Library_Dir to the path
(Set_Ada_Paths.Add.Recursive_Add): Ditto
* mlib-prj.adb (Check_Library): For all library projects, get the
library file timestamp.
(Build_Library): Copy ALI files in Library_ALI_Dir, not in Library_Dir
(Build_Library): Use new function DLL_Prefix for the DLL_Name
(Clean): Remove procedure, no longer used
(Ultimate_Extension_Of): New function
(Build_Library): When cleaning the library directory, only remove an
existing library file and any ALI file of a source of the project.
When cleaning the interface copy directory, remove any source that
could be a source of the project.
* prj.ads, prj.adb (Project_Empty): Add values of new components
Library_TS and All_Imported_Projects.
(Project_Empty): Add values for new components of Project_Data:
Library_ALI_Dir and Display_Library_ALI_Dir
* prj-attr.adb: New project level attribute name Library_ALI_Dir
* prj-nmsc.adb (Check_Library_Attributes): Take into account new
attribute Library_ALI_Dir.
(Check_Library_Attributes): The library directory cannot be the same as
any source directory of the project tree.
(Check_Stand_Alone_Library): The interface copy directory cannot be
the same as any source directory of the project tree.
* mlib.adb: Use Prj.Com.Fail, instead of Osint.Fail directly, to delete
all temporary files.
From-SVN: r106967
2005-11-14 Emmanuel Briot <briot@adacore.com>
* xref_lib.adb (Parse_Identifier_Info): It is possible for an entity
line in the ALI file to include both an instantiation reference, and a
returned value.
From-SVN: r106966
2005-11-14 Hristian Kirtchev <kirtchev@adacore.com>
Javier Miranda <miranda@adacore.com>
* rtsfind.ads, exp_util.adb, exp_util.ads, exp_disp.adb, exp_disp.ads,
exp_ch7.adb, sem_ch9.adb, snames.adb, snames.ads,
exp_ch9.adb, exp_ch9.ads, exp_ch6.adb, exp_ch3.adb, exp_ch3.ads,
einfo.ads, einfo.adb: Complete support for Ada 2005 interfaces.
* a-tags.ads, a-tags.adb: Major rewrite and additions to implement
properly new Ada 2005 interfaces (AI-345) and add run-time checks (via
assertions).
* exp_dbug.ads, exp_dbug.adb (Get_Secondary_DT_External_Name): New
subprogram that generates the external name associated with a
secondary dispatch table.
(Get_Secondary_DT_External_Name): New subprogram that generates the
external name associated with a secondary dispatch table.
From-SVN: r106965
2005-11-14 Matthew Gingell <gingell@adacore.com>
Olivier Hainque <hainque@adacore.com>
* tb-gcc.c (trace_callback): Work around problem with _Unwind_GetIP on
ia64 HP-UX.
* tracebak.c (ia64 configuration): Enable _Unwind_Backtrace driven
tracebacks on ia64 HP-UX and provide explanatory comment.
Enable backtraces on ia64 GNU/Linux.
(x86 configuration): Bump FRAME_LEVEL to 1 to ensure we retrieve a real
base pointer from builtin_frame_address. Adjust BASE_SKIP accordingly.
From-SVN: r106964
2005-11-14 Olivier Hainque <hainque@adacore.com>
* cuintp.c (UI_To_gnu): Use a proper type for intermediate computations
to ensure bias adjustments take place when need be and to prevent
occurrences of intermediate overflows.
From-SVN: r106963
2005-11-14 Thomas Quinot <quinot@adacore.com>
Olivier Hainque <hainque@adacore.com>
Eric Botcazou <ebotcazou@adacore.com>
* decl.c:
Factor common code to build a storage type for an unconstrained object
from a fat or thin pointer type and a constrained object type.
(annotate_value): Handle BIT_AND_EXPR.
(annotate_rep): Don't restrict the back annotation of inherited
components to the type_annotate_only case.
(gnat_to_gnu_entity) <E_Array_Type>: Do not invoke create_type_decl if
we are not defining the type.
<E_Record_Type>: Likewise.
(gnat_to_gnu_entity) <object, renaming>: Adjust comments and structure
to get advantage of the new maybe_stabilize_reference interface, to
ensure that what we reference is indeed stabilized instead of relying
on assumptions on what the stabilizer does.
(gnat_to_gnu_entity) <E_Incomplete_Type>: If the entity is an incomplete
type imported through a limited_with clause, use its non-limited view.
(Has_Stdcall_Convention): New macro, to centralize the Windows vs others
differentiation.
(gnat_to_gnu_entity): Use Has_Stdcall_Convention instead of a spread mix
of #if sections + explicit comparisons of convention identifiers.
(gnat_to_gnu_entity) <E_Variable>: Decrement force_global if necessary
before early-returning for certain types when code generation is
disabled.
(gnat_to_gnu_entity) <object>: Adjust comment attached to the
nullification of gnu_expr we do for objects with address clause and
that we are not defining.
(elaborate_expression_1): Do not create constants when creating
variables needed by the debug info: the dwarf2 writer considers that
CONST_DECLs is used only to represent enumeration constants, and emits
nothing for them.
(gnat_to_gnu_entity) <object>: When turning a non-definition of an
object with an address clause into an indirect reference, drop the
initializing expression.
Include "expr.h".
(STACK_CHECK_BUILTIN): Delete.
(STACK_CHECK_PROBE_INTERVAL): Likewise.
(STACK_CHECK_MAX_FRAME_SIZE): Likewise.
(STACK_CHECK_MAX_VAR_SIZE): Likewise.
(gnat_to_gnu_entity): If gnat_entity is a renaming, do not mark the tree
corresponding to the renamed object as ignored for debugging purposes.
* trans.c (tree_transform, case N_Attribute_Reference, case Attr_Size &
related): For a prefix that is a dereference of a fat or thin pointer,
if there is an actual subtype provided by the front-end, use that
subtype to build an actual type with bounds template.
(tree_transform, case N_Free_Statement): If an Actual_Designated_Subtype
is provided by the front-end, use that subtype to compute the size of
the deallocated object.
(gnat_to_gnu): When adding a statement into an elaboration procedure,
check for a potential violation of a No_Elaboration_Code restriction.
(maybe_stabilize_reference): New function, like gnat_stabilize_reference
with extra arguments to control whether to recurse through non-values
and to let the caller know if the stabilization has succeeded.
(gnat_stabilize_reference): Now a simple wrapper around
maybe_stabilize, for common uses without restriction on lvalues and
without need to check for the success indication.
(gnat_to_gnu, call_to_gnu): Adjust calls to gnat_stabilize_reference, to
pass false instead of 0 as the FORCE argument which is a bool.
(Identifier_to_gnu): Remove checks ensuring that an renamed object
attached to a renaming pointer has been properly stabilized, as no such
object is attached otherwise.
(call_to_gnu): Invoke create_var_decl to create the temporary when the
function uses the "target pointer" return mechanism.
Reinstate conversion of the actual to the type of the formal
parameter before any other specific treatment based on the passing
mechanism. This turns out to be necessary in order for PLACEHOLDER
substitution to work properly when the latter type is unconstrained.
* gigi.h (build_unc_object_type_from_ptr): New subprogram, factoring a
common pattern.
(maybe_stabilize_reference): New function, like gnat_stabilize_reference
with extra arguments to control whether to recurse through non-values
and to let the caller know if the stabilization has succeeded.
* utils2.c (gnat_build_constructor): Only sort the fields for possible
static output of record constructor if all the components are constant.
(gnat_build_constructor): For a record type, sort the list of field
initializers in increasing bit position order.
Factor common code to build a storage type for an unconstrained object
from a fat or thin pointer type and a constrained object type.
(build_unary_op) <ADDR_EXPR>: Always recurse down conversions between
types variants, and process special cases of VIEW_CONVERT expressions
as their NOP_EXPR counterpart to ensure we get to the
CORRESPONDING_VARs associated with CONST_DECls.
(build_binary_op) <MODIFY_EXPR>: Do not strip VIEW_CONVERT_EXPRs
on the right-hand side.
* utils.c (build_unc_object_type_from_ptr): New subprogram, factoring
a common pattern.
(convert) <VIEW_CONVERT_EXPR>: Return the inner operand directly if we
are converting back to its original type.
(convert) <JM input>: Fallthrough regular conversion code instead of
extracting the object if converting to a type variant.
(create_var_decl): When a variable has an initializer requiring code
generation and we are at the top level, check for a potential violation
of a No_Elaboration_Code restriction.
(create_var_decl): call expand_decl for CONST_DECLs, to set MODE, ALIGN
SIZE and SIZE_UNIT which we need for later back-annotations.
* utils.c: (convert) <STRING_CST>: Remove obsolete code.
<VIEW_CONVERT_EXPR>: Do not lift the conversion if the target type
is an unchecked union.
(pushdecl): Set DECL_NO_STATIC_CHAIN on imported nested functions.
(convert) <VIEW_CONVERT_EXPR>: When the types have the same
main variant, just replace the VIEW_CONVERT_EXPR.
<UNION_TYPE>: Revert 2005-03-02 change.
* repinfo.h, repinfo.ads: Add tcode for BIT_AND_EXPR.
* repinfo.adb (Print_Expr, Rep_Value): Handle Bit_And_Expressions.
From-SVN: r106961
2005-11-14 Jose Ruiz <ruiz@adacore.com>
* s-tassta.adb (Create_Task): Move the code in charge of resetting the
deferral level, when abort is not allowed, to a later stage (the
Task_Wrapper).
(Task_Wrapper): If Abort is not allowed, reset the deferral level since
it will not get changed by the generated code. It was previously done
in Create_Task.
From-SVN: r106960
* a-except.adb (Zero_Cost_Exceptions): Removed, no longer used.
(builtin_longjmp, Process_Raise_Exceeption): Move setjmp/longjmp
related code to a-exexpr.adb
(Save_Occurrence_And_Private): Move GCC EH related code to
a-exexpr-gcc.adb
(Raise_Current_Excep): Add new variable Id with pragma
volatile, to ensure that the variable lives on stack.
* a-exexpr-gcc.adb, raise-gcc.c: New file.
* a-exexpr.adb (builtin_longjmp, Propagate_Exception): Moved here code
from a-except.adb.
Move GCC EH related code to a-exexpr-gcc.adb
* Makefile.in: Add or update g-soccon LIBGNAT pairs for Linux/PPC and
64-bit Solaris
Split the Linux version of g-soccon into separate variants for 32 and 64
bit platforms.
(gnatlib): Use $(AR_FOR_TARGET) and $(RANLIB_FOR_TARGET)
vice $(AR) and $(RANLIB). Remove use of host variable $(RANLIB_FLAGS).
install-gnatlib: Use $(RANLIB_FOR_TARGET) vice $(RANLIB). Remove use
of host variable $(RANLIB_FLAGS).
(alpha64-dec-*vms*): Fix translations for 64 bit compiler.
Code clean up: remove unused/obsolete targets.
(EH_MECHANISM): New variable introduced to differenciate between the
two EH mechanisms statically.
(gnatlib-zcx, gnatlib-sjlj): Force EH_MECHANISM manually.
(LIBGNAT_OBJS): Add raise-gcc.o
(LIBGNAT_TARGET_PAIRS for ppc-vxworks): Use an specialized version of
s-osinte.adb, s-tpopsp.adb, and system.ads for the run time that
supports VxWorks 6 RTPs.
(EXTRA_GNATRTL_NONTASKING_OBJS for ppc-vxworks): Remove the use of
i-vxworks and i-vxwoio from the run time that supports VxWorks 6 RTPs.
* raise.c: Move all GCC EH-related routines to raise-gcc.c
From-SVN: r106959
* adaint.h, adaint.c (__gnat_is_cross_compiler): New constant.
(Gnat_Install_Locks, __gnatlib_install_locks): Removed, no longer used.
(convert_address): Update comments and list of platforms using this.
* s-tasini.adb (Gnat_Install_Locks, __gnatlib_install_locks): Removed,
no longer used.
From-SVN: r106957
2005-11-14 Jose Ruiz <ruiz@adacore.com>
* s-taprop-posix.adb (Initialize_Lock): Destroy mutex attribute before
raising the exception so the memory used is freed.
From-SVN: r106956
2005-11-14 Matthew Gingell <gingell@adacore.com>
* system-lynxos-ppc.ads, system-lynxos-x86.ads:
Increase default priority on Lynx from 15 to 17, and meet the Ada
requirement that Default_Priority be ((Priority'First +
Priority'Last) / 2) by increasing the range of Interrupt_Priority.
From-SVN: r106952
2005-11-14 Robert Dewar <dewar@adacore.com>
* exp_vfpt.adb: Handle /= case
(Expand_Vax_Conversion): Properly recognize Conversion_OK flag
so that we do not get duplicate scaling for fixed point conversions.
* s-vaflop.ads, s-vaflop.adb: (Ne_F): New function
From-SVN: r106951
2005-11-14 Robert Dewar <dewar@adacore.com>
Ed Schonberg <schonberg@adacore.com>
PR ada/18434
* types.ads: Include All_Checks in Suppress_Array
* checks.adb (Check_Needed): Remove kludge for a/=b rewritten as
not(a=b), since we no longer do this rewriting, and hence it is not
needed.
(Elaboration_Checks_Suppressed): Add special casing to
deal with different cases of static and dynamic elaboration checks (all
checks does not count in the first case, but does in the second).
(Expr_Known_Valid): Do not assume that the result of any arbitrary
function call is valid, since this is not the case.
(Ensure_Valid): Do not apply validity check to a real literal
in a universal or fixed context
* exp_ch4.adb (Expand_N_Op_Ne): Don't expand a/=b to not(a=b) for
elementary types using the operator in standard. It is cleaner not to
modify the programmers intent, especially in the case of floating-point.
(Rewrite_Comparison): Fix handling of /= (this was always wrong, but
it did not matter because we always rewrote a/=b to not(a=b).
(Expand_Allocator_Expression): For an allocator expression whose nominal
subtype is an unconstrained packed type, convert the expression to its
actual constrained subtype.
Implement warning for <= or >= where < or > not possible
Fix to Vax_Float tests (too early in many routines, causing premature
Vax_Float expansions.
* sem_prag.adb (Analyze_Pragma, case Obsolescent): Allow this pragma
to be used with packages and generic packages as well as with
subprograms.
(Suppress): Set All_Checks, but not Elaboration_Check, for case
of pragma Suppress (All_Checks)
(Analyze_Pragma, case Warnings): Implement first argument allowed to be
a string literal for precise control over warnings.
Avoid raise of pragma in case of unrecognized pragma and just return
instead.
* sem_prag.ads: Minor reformatting
* switch-c.adb (Scan_Front_End_Switches): Replace "raise Bad_Switch;"
with call to new procedure Bad_Switch. Call Scan_Pos with new parameter
Switch. Do not handle any exception.
Include -gnatwx as part of -gnatg (warn on redundant parens)
Allow optional = after -gnatm
(Scan_Front_End_Switches): The -gnatp switch sets All_Checks, but no
longer sets Elaboration_Checks.
Code to set warning mode moved to Sem_Warn
so that it can be shared by pragma processing.
* s-mastop-tru64.adb (Pop_Frame): Remove redundant parentheses in if
statement.
* s-taprop-solaris.adb:
Change some <= to =, to avoid new warning
* a-exexda.adb, prj-proc.adb:
Fix obvious typo (Num_Tracebacks compared <= 0 instead of < 0)
Fix obvious typo (Total_Errors_Detected <= 0 should be = 0)
From-SVN: r106950
