------------------------------------------------------------------------------ -- -- -- GNAT COMPILER COMPONENTS -- -- -- -- S E M _ A G G R -- -- -- -- B o d y -- -- -- -- Copyright (C) 1992-2009, Free Software Foundation, Inc. -- -- -- -- GNAT is free software; you can redistribute it and/or modify it under -- -- terms of the GNU General Public License as published by the Free Soft- -- -- ware Foundation; either version 3, or (at your option) any later ver- -- -- sion. GNAT is distributed in the hope that it will be useful, but WITH- -- -- OUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY -- -- or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License -- -- for more details. You should have received a copy of the GNU General -- -- Public License distributed with GNAT; see file COPYING3. If not, go to -- -- http://www.gnu.org/licenses for a complete copy of the license. -- -- -- -- GNAT was originally developed by the GNAT team at New York University. -- -- Extensive contributions were provided by Ada Core Technologies Inc. -- -- -- ------------------------------------------------------------------------------ with Atree; use Atree; with Checks; use Checks; with Einfo; use Einfo; with Elists; use Elists; with Errout; use Errout; with Expander; use Expander; with Exp_Tss; use Exp_Tss; with Exp_Util; use Exp_Util; with Freeze; use Freeze; with Itypes; use Itypes; with Lib; use Lib; with Lib.Xref; use Lib.Xref; with Namet; use Namet; with Namet.Sp; use Namet.Sp; with Nmake; use Nmake; with Nlists; use Nlists; with Opt; use Opt; with Sem; use Sem; with Sem_Aux; use Sem_Aux; with Sem_Cat; use Sem_Cat; with Sem_Ch3; use Sem_Ch3; with Sem_Ch13; use Sem_Ch13; with Sem_Eval; use Sem_Eval; with Sem_Res; use Sem_Res; with Sem_Util; use Sem_Util; with Sem_Type; use Sem_Type; with Sem_Warn; use Sem_Warn; with Sinfo; use Sinfo; with Snames; use Snames; with Stringt; use Stringt; with Stand; use Stand; with Targparm; use Targparm; with Tbuild; use Tbuild; with Uintp; use Uintp; package body Sem_Aggr is type Case_Bounds is record Choice_Lo : Node_Id; Choice_Hi : Node_Id; Choice_Node : Node_Id; end record; type Case_Table_Type is array (Nat range <>) of Case_Bounds; -- Table type used by Check_Case_Choices procedure ----------------------- -- Local Subprograms -- ----------------------- procedure Sort_Case_Table (Case_Table : in out Case_Table_Type); -- Sort the Case Table using the Lower Bound of each Choice as the key. -- A simple insertion sort is used since the number of choices in a case -- statement of variant part will usually be small and probably in near -- sorted order. procedure Check_Can_Never_Be_Null (Typ : Entity_Id; Expr : Node_Id); -- Ada 2005 (AI-231): Check bad usage of null for a component for which -- null exclusion (NOT NULL) is specified. Typ can be an E_Array_Type for -- the array case (the component type of the array will be used) or an -- E_Component/E_Discriminant entity in the record case, in which case the -- type of the component will be used for the test. If Typ is any other -- kind of entity, the call is ignored. Expr is the component node in the -- aggregate which is known to have a null value. A warning message will be -- issued if the component is null excluding. -- -- It would be better to pass the proper type for Typ ??? procedure Check_Expr_OK_In_Limited_Aggregate (Expr : Node_Id); -- Check that Expr is either not limited or else is one of the cases of -- expressions allowed for a limited component association (namely, an -- aggregate, function call, or <> notation). Report error for violations. ------------------------------------------------------ -- Subprograms used for RECORD AGGREGATE Processing -- ------------------------------------------------------ procedure Resolve_Record_Aggregate (N : Node_Id; Typ : Entity_Id); -- This procedure performs all the semantic checks required for record -- aggregates. Note that for aggregates analysis and resolution go -- hand in hand. Aggregate analysis has been delayed up to here and -- it is done while resolving the aggregate. -- -- N is the N_Aggregate node. -- Typ is the record type for the aggregate resolution -- -- While performing the semantic checks, this procedure builds a new -- Component_Association_List where each record field appears alone in a -- Component_Choice_List along with its corresponding expression. The -- record fields in the Component_Association_List appear in the same order -- in which they appear in the record type Typ. -- -- Once this new Component_Association_List is built and all the semantic -- checks performed, the original aggregate subtree is replaced with the -- new named record aggregate just built. Note that subtree substitution is -- performed with Rewrite so as to be able to retrieve the original -- aggregate. -- -- The aggregate subtree manipulation performed by Resolve_Record_Aggregate -- yields the aggregate format expected by Gigi. Typically, this kind of -- tree manipulations are done in the expander. However, because the -- semantic checks that need to be performed on record aggregates really go -- hand in hand with the record aggregate normalization, the aggregate -- subtree transformation is performed during resolution rather than -- expansion. Had we decided otherwise we would have had to duplicate most -- of the code in the expansion procedure Expand_Record_Aggregate. Note, -- however, that all the expansion concerning aggregates for tagged records -- is done in Expand_Record_Aggregate. -- -- The algorithm of Resolve_Record_Aggregate proceeds as follows: -- -- 1. Make sure that the record type against which the record aggregate -- has to be resolved is not abstract. Furthermore if the type is a -- null aggregate make sure the input aggregate N is also null. -- -- 2. Verify that the structure of the aggregate is that of a record -- aggregate. Specifically, look for component associations and ensure -- that each choice list only has identifiers or the N_Others_Choice -- node. Also make sure that if present, the N_Others_Choice occurs -- last and by itself. -- -- 3. If Typ contains discriminants, the values for each discriminant is -- looked for. If the record type Typ has variants, we check that the -- expressions corresponding to each discriminant ruling the (possibly -- nested) variant parts of Typ, are static. This allows us to determine -- the variant parts to which the rest of the aggregate must conform. -- The names of discriminants with their values are saved in a new -- association list, New_Assoc_List which is later augmented with the -- names and values of the remaining components in the record type. -- -- During this phase we also make sure that every discriminant is -- assigned exactly one value. Note that when several values for a given -- discriminant are found, semantic processing continues looking for -- further errors. In this case it's the first discriminant value found -- which we will be recorded. -- -- IMPORTANT NOTE: For derived tagged types this procedure expects -- First_Discriminant and Next_Discriminant to give the correct list -- of discriminants, in the correct order. -- -- 4. After all the discriminant values have been gathered, we can set the -- Etype of the record aggregate. If Typ contains no discriminants this -- is straightforward: the Etype of N is just Typ, otherwise a new -- implicit constrained subtype of Typ is built to be the Etype of N. -- -- 5. Gather the remaining record components according to the discriminant -- values. This involves recursively traversing the record type -- structure to see what variants are selected by the given discriminant -- values. This processing is a little more convoluted if Typ is a -- derived tagged types since we need to retrieve the record structure -- of all the ancestors of Typ. -- -- 6. After gathering the record components we look for their values in the -- record aggregate and emit appropriate error messages should we not -- find such values or should they be duplicated. -- -- 7. We then make sure no illegal component names appear in the record -- aggregate and make sure that the type of the record components -- appearing in a same choice list is the same. Finally we ensure that -- the others choice, if present, is used to provide the value of at -- least a record component. -- -- 8. The original aggregate node is replaced with the new named aggregate -- built in steps 3 through 6, as explained earlier. -- -- Given the complexity of record aggregate resolution, the primary goal of -- this routine is clarity and simplicity rather than execution and storage -- efficiency. If there are only positional components in the aggregate the -- running time is linear. If there are associations the running time is -- still linear as long as the order of the associations is not too far off -- the order of the components in the record type. If this is not the case -- the running time is at worst quadratic in the size of the association -- list. procedure Check_Misspelled_Component (Elements : Elist_Id; Component : Node_Id); -- Give possible misspelling diagnostic if Component is likely to be a -- misspelling of one of the components of the Assoc_List. This is called -- by Resolve_Aggr_Expr after producing an invalid component error message. procedure Check_Static_Discriminated_Subtype (T : Entity_Id; V : Node_Id); -- An optimization: determine whether a discriminated subtype has a static -- constraint, and contains array components whose length is also static, -- either because they are constrained by the discriminant, or because the -- original component bounds are static. ----------------------------------------------------- -- Subprograms used for ARRAY AGGREGATE Processing -- ----------------------------------------------------- function Resolve_Array_Aggregate (N : Node_Id; Index : Node_Id; Index_Constr : Node_Id; Component_Typ : Entity_Id; Others_Allowed : Boolean) return Boolean; -- This procedure performs the semantic checks for an array aggregate. -- True is returned if the aggregate resolution succeeds. -- -- The procedure works by recursively checking each nested aggregate. -- Specifically, after checking a sub-aggregate nested at the i-th level -- we recursively check all the subaggregates at the i+1-st level (if any). -- Note that for aggregates analysis and resolution go hand in hand. -- Aggregate analysis has been delayed up to here and it is done while -- resolving the aggregate. -- -- N is the current N_Aggregate node to be checked. -- -- Index is the index node corresponding to the array sub-aggregate that -- we are currently checking (RM 4.3.3 (8)). Its Etype is the -- corresponding index type (or subtype). -- -- Index_Constr is the node giving the applicable index constraint if -- any (RM 4.3.3 (10)). It "is a constraint provided by certain -- contexts [...] that can be used to determine the bounds of the array -- value specified by the aggregate". If Others_Allowed below is False -- there is no applicable index constraint and this node is set to Index. -- -- Component_Typ is the array component type. -- -- Others_Allowed indicates whether an others choice is allowed -- in the context where the top-level aggregate appeared. -- -- The algorithm of Resolve_Array_Aggregate proceeds as follows: -- -- 1. Make sure that the others choice, if present, is by itself and -- appears last in the sub-aggregate. Check that we do not have -- positional and named components in the array sub-aggregate (unless -- the named association is an others choice). Finally if an others -- choice is present, make sure it is allowed in the aggregate context. -- -- 2. If the array sub-aggregate contains discrete_choices: -- -- (A) Verify their validity. Specifically verify that: -- -- (a) If a null range is present it must be the only possible -- choice in the array aggregate. -- -- (b) Ditto for a non static range. -- -- (c) Ditto for a non static expression. -- -- In addition this step analyzes and resolves each discrete_choice, -- making sure that its type is the type of the corresponding Index. -- If we are not at the lowest array aggregate level (in the case of -- multi-dimensional aggregates) then invoke Resolve_Array_Aggregate -- recursively on each component expression. Otherwise, resolve the -- bottom level component expressions against the expected component -- type ONLY IF the component corresponds to a single discrete choice -- which is not an others choice (to see why read the DELAYED -- COMPONENT RESOLUTION below). -- -- (B) Determine the bounds of the sub-aggregate and lowest and -- highest choice values. -- -- 3. For positional aggregates: -- -- (A) Loop over the component expressions either recursively invoking -- Resolve_Array_Aggregate on each of these for multi-dimensional -- array aggregates or resolving the bottom level component -- expressions against the expected component type. -- -- (B) Determine the bounds of the positional sub-aggregates. -- -- 4. Try to determine statically whether the evaluation of the array -- sub-aggregate raises Constraint_Error. If yes emit proper -- warnings. The precise checks are the following: -- -- (A) Check that the index range defined by aggregate bounds is -- compatible with corresponding index subtype. -- We also check against the base type. In fact it could be that -- Low/High bounds of the base type are static whereas those of -- the index subtype are not. Thus if we can statically catch -- a problem with respect to the base type we are guaranteed -- that the same problem will arise with the index subtype -- -- (B) If we are dealing with a named aggregate containing an others -- choice and at least one discrete choice then make sure the range -- specified by the discrete choices does not overflow the -- aggregate bounds. We also check against the index type and base -- type bounds for the same reasons given in (A). -- -- (C) If we are dealing with a positional aggregate with an others -- choice make sure the number of positional elements specified -- does not overflow the aggregate bounds. We also check against -- the index type and base type bounds as mentioned in (A). -- -- Finally construct an N_Range node giving the sub-aggregate bounds. -- Set the Aggregate_Bounds field of the sub-aggregate to be this -- N_Range. The routine Array_Aggr_Subtype below uses such N_Ranges -- to build the appropriate aggregate subtype. Aggregate_Bounds -- information is needed during expansion. -- -- DELAYED COMPONENT RESOLUTION: The resolution of bottom level component -- expressions in an array aggregate may call Duplicate_Subexpr or some -- other routine that inserts code just outside the outermost aggregate. -- If the array aggregate contains discrete choices or an others choice, -- this may be wrong. Consider for instance the following example. -- -- type Rec is record -- V : Integer := 0; -- end record; -- -- type Acc_Rec is access Rec; -- Arr : array (1..3) of Acc_Rec := (1 .. 3 => new Rec); -- -- Then the transformation of "new Rec" that occurs during resolution -- entails the following code modifications -- -- P7b : constant Acc_Rec := new Rec; -- RecIP (P7b.all); -- Arr : array (1..3) of Acc_Rec := (1 .. 3 => P7b); -- -- This code transformation is clearly wrong, since we need to call -- "new Rec" for each of the 3 array elements. To avoid this problem we -- delay resolution of the components of non positional array aggregates -- to the expansion phase. As an optimization, if the discrete choice -- specifies a single value we do not delay resolution. function Array_Aggr_Subtype (N : Node_Id; Typ : Node_Id) return Entity_Id; -- This routine returns the type or subtype of an array aggregate. -- -- N is the array aggregate node whose type we return. -- -- Typ is the context type in which N occurs. -- -- This routine creates an implicit array subtype whose bounds are -- those defined by the aggregate. When this routine is invoked -- Resolve_Array_Aggregate has already processed aggregate N. Thus the -- Aggregate_Bounds of each sub-aggregate, is an N_Range node giving the -- sub-aggregate bounds. When building the aggregate itype, this function -- traverses the array aggregate N collecting such Aggregate_Bounds and -- constructs the proper array aggregate itype. -- -- Note that in the case of multidimensional aggregates each inner -- sub-aggregate corresponding to a given array dimension, may provide a -- different bounds. If it is possible to determine statically that -- some sub-aggregates corresponding to the same index do not have the -- same bounds, then a warning is emitted. If such check is not possible -- statically (because some sub-aggregate bounds are dynamic expressions) -- then this job is left to the expander. In all cases the particular -- bounds that this function will chose for a given dimension is the first -- N_Range node for a sub-aggregate corresponding to that dimension. -- -- Note that the Raises_Constraint_Error flag of an array aggregate -- whose evaluation is determined to raise CE by Resolve_Array_Aggregate, -- is set in Resolve_Array_Aggregate but the aggregate is not -- immediately replaced with a raise CE. In fact, Array_Aggr_Subtype must -- first construct the proper itype for the aggregate (Gigi needs -- this). After constructing the proper itype we will eventually replace -- the top-level aggregate with a raise CE (done in Resolve_Aggregate). -- Of course in cases such as: -- -- type Arr is array (integer range <>) of Integer; -- A : Arr := (positive range -1 .. 2 => 0); -- -- The bounds of the aggregate itype are cooked up to look reasonable -- (in this particular case the bounds will be 1 .. 2). procedure Aggregate_Constraint_Checks (Exp : Node_Id; Check_Typ : Entity_Id); -- Checks expression Exp against subtype Check_Typ. If Exp is an -- aggregate and Check_Typ a constrained record type with discriminants, -- we generate the appropriate discriminant checks. If Exp is an array -- aggregate then emit the appropriate length checks. If Exp is a scalar -- type, or a string literal, Exp is changed into Check_Typ'(Exp) to -- ensure that range checks are performed at run time. procedure Make_String_Into_Aggregate (N : Node_Id); -- A string literal can appear in a context in which a one dimensional -- array of characters is expected. This procedure simply rewrites the -- string as an aggregate, prior to resolution. --------------------------------- -- Aggregate_Constraint_Checks -- --------------------------------- procedure Aggregate_Constraint_Checks (Exp : Node_Id; Check_Typ : Entity_Id) is Exp_Typ : constant Entity_Id := Etype (Exp); begin if Raises_Constraint_Error (Exp) then return; end if; -- Ada 2005 (AI-230): Generate a conversion to an anonymous access -- component's type to force the appropriate accessibility checks. -- Ada 2005 (AI-231): Generate conversion to the null-excluding -- type to force the corresponding run-time check if Is_Access_Type (Check_Typ) and then ((Is_Local_Anonymous_Access (Check_Typ)) or else (Can_Never_Be_Null (Check_Typ) and then not Can_Never_Be_Null (Exp_Typ))) then Rewrite (Exp, Convert_To (Check_Typ, Relocate_Node (Exp))); Analyze_And_Resolve (Exp, Check_Typ); Check_Unset_Reference (Exp); end if; -- This is really expansion activity, so make sure that expansion -- is on and is allowed. if not Expander_Active or else In_Spec_Expression then return; end if; -- First check if we have to insert discriminant checks if Has_Discriminants (Exp_Typ) then Apply_Discriminant_Check (Exp, Check_Typ); -- Next emit length checks for array aggregates elsif Is_Array_Type (Exp_Typ) then Apply_Length_Check (Exp, Check_Typ); -- Finally emit scalar and string checks. If we are dealing with a -- scalar literal we need to check by hand because the Etype of -- literals is not necessarily correct. elsif Is_Scalar_Type (Exp_Typ) and then Compile_Time_Known_Value (Exp) then if Is_Out_Of_Range (Exp, Base_Type (Check_Typ)) then Apply_Compile_Time_Constraint_Error (Exp, "value not in range of}?", CE_Range_Check_Failed, Ent => Base_Type (Check_Typ), Typ => Base_Type (Check_Typ)); elsif Is_Out_Of_Range (Exp, Check_Typ) then Apply_Compile_Time_Constraint_Error (Exp, "value not in range of}?", CE_Range_Check_Failed, Ent => Check_Typ, Typ => Check_Typ); elsif not Range_Checks_Suppressed (Check_Typ) then Apply_Scalar_Range_Check (Exp, Check_Typ); end if; -- Verify that target type is also scalar, to prevent view anomalies -- in instantiations. elsif (Is_Scalar_Type (Exp_Typ) or else Nkind (Exp) = N_String_Literal) and then Is_Scalar_Type (Check_Typ) and then Exp_Typ /= Check_Typ then if Is_Entity_Name (Exp) and then Ekind (Entity (Exp)) = E_Constant then -- If expression is a constant, it is worthwhile checking whether -- it is a bound of the type. if (Is_Entity_Name (Type_Low_Bound (Check_Typ)) and then Entity (Exp) = Entity (Type_Low_Bound (Check_Typ))) or else (Is_Entity_Name (Type_High_Bound (Check_Typ)) and then Entity (Exp) = Entity (Type_High_Bound (Check_Typ))) then return; else Rewrite (Exp, Convert_To (Check_Typ, Relocate_Node (Exp))); Analyze_And_Resolve (Exp, Check_Typ); Check_Unset_Reference (Exp); end if; else Rewrite (Exp, Convert_To (Check_Typ, Relocate_Node (Exp))); Analyze_And_Resolve (Exp, Check_Typ); Check_Unset_Reference (Exp); end if; end if; end Aggregate_Constraint_Checks; ------------------------ -- Array_Aggr_Subtype -- ------------------------ function Array_Aggr_Subtype (N : Node_Id; Typ : Entity_Id) return Entity_Id is Aggr_Dimension : constant Pos := Number_Dimensions (Typ); -- Number of aggregate index dimensions Aggr_Range : array (1 .. Aggr_Dimension) of Node_Id := (others => Empty); -- Constrained N_Range of each index dimension in our aggregate itype Aggr_Low : array (1 .. Aggr_Dimension) of Node_Id := (others => Empty); Aggr_High : array (1 .. Aggr_Dimension) of Node_Id := (others => Empty); -- Low and High bounds for each index dimension in our aggregate itype Is_Fully_Positional : Boolean := True; procedure Collect_Aggr_Bounds (N : Node_Id; Dim : Pos); -- N is an array (sub-)aggregate. Dim is the dimension corresponding to -- (sub-)aggregate N. This procedure collects the constrained N_Range -- nodes corresponding to each index dimension of our aggregate itype. -- These N_Range nodes are collected in Aggr_Range above. -- -- Likewise collect in Aggr_Low & Aggr_High above the low and high -- bounds of each index dimension. If, when collecting, two bounds -- corresponding to the same dimension are static and found to differ, -- then emit a warning, and mark N as raising Constraint_Error. ------------------------- -- Collect_Aggr_Bounds -- ------------------------- procedure Collect_Aggr_Bounds (N : Node_Id; Dim : Pos) is This_Range : constant Node_Id := Aggregate_Bounds (N); -- The aggregate range node of this specific sub-aggregate This_Low : constant Node_Id := Low_Bound (Aggregate_Bounds (N)); This_High : constant Node_Id := High_Bound (Aggregate_Bounds (N)); -- The aggregate bounds of this specific sub-aggregate Assoc : Node_Id; Expr : Node_Id; begin -- Collect the first N_Range for a given dimension that you find. -- For a given dimension they must be all equal anyway. if No (Aggr_Range (Dim)) then Aggr_Low (Dim) := This_Low; Aggr_High (Dim) := This_High; Aggr_Range (Dim) := This_Range; else if Compile_Time_Known_Value (This_Low) then if not Compile_Time_Known_Value (Aggr_Low (Dim)) then Aggr_Low (Dim) := This_Low; elsif Expr_Value (This_Low) /= Expr_Value (Aggr_Low (Dim)) then Set_Raises_Constraint_Error (N); Error_Msg_N ("sub-aggregate low bound mismatch?", N); Error_Msg_N ("\Constraint_Error will be raised at run-time?", N); end if; end if; if Compile_Time_Known_Value (This_High) then if not Compile_Time_Known_Value (Aggr_High (Dim)) then Aggr_High (Dim) := This_High; elsif Expr_Value (This_High) /= Expr_Value (Aggr_High (Dim)) then Set_Raises_Constraint_Error (N); Error_Msg_N ("sub-aggregate high bound mismatch?", N); Error_Msg_N ("\Constraint_Error will be raised at run-time?", N); end if; end if; end if; if Dim < Aggr_Dimension then -- Process positional components if Present (Expressions (N)) then Expr := First (Expressions (N)); while Present (Expr) loop Collect_Aggr_Bounds (Expr, Dim + 1); Next (Expr); end loop; end if; -- Process component associations if Present (Component_Associations (N)) then Is_Fully_Positional := False; Assoc := First (Component_Associations (N)); while Present (Assoc) loop Expr := Expression (Assoc); Collect_Aggr_Bounds (Expr, Dim + 1); Next (Assoc); end loop; end if; end if; end Collect_Aggr_Bounds; -- Array_Aggr_Subtype variables Itype : Entity_Id; -- The final itype of the overall aggregate Index_Constraints : constant List_Id := New_List; -- The list of index constraints of the aggregate itype -- Start of processing for Array_Aggr_Subtype begin -- Make sure that the list of index constraints is properly attached to -- the tree, and then collect the aggregate bounds. Set_Parent (Index_Constraints, N); Collect_Aggr_Bounds (N, 1); -- Build the list of constrained indices of our aggregate itype for J in 1 .. Aggr_Dimension loop Create_Index : declare Index_Base : constant Entity_Id := Base_Type (Etype (Aggr_Range (J))); Index_Typ : Entity_Id; begin -- Construct the Index subtype, and associate it with the range -- construct that generates it. Index_Typ := Create_Itype (Subtype_Kind (Ekind (Index_Base)), Aggr_Range (J)); Set_Etype (Index_Typ, Index_Base); if Is_Character_Type (Index_Base) then Set_Is_Character_Type (Index_Typ); end if; Set_Size_Info (Index_Typ, (Index_Base)); Set_RM_Size (Index_Typ, RM_Size (Index_Base)); Set_First_Rep_Item (Index_Typ, First_Rep_Item (Index_Base)); Set_Scalar_Range (Index_Typ, Aggr_Range (J)); if Is_Discrete_Or_Fixed_Point_Type (Index_Typ) then Set_RM_Size (Index_Typ, UI_From_Int (Minimum_Size (Index_Typ))); end if; Set_Etype (Aggr_Range (J), Index_Typ); Append (Aggr_Range (J), To => Index_Constraints); end Create_Index; end loop; -- Now build the Itype Itype := Create_Itype (E_Array_Subtype, N); Set_First_Rep_Item (Itype, First_Rep_Item (Typ)); Set_Convention (Itype, Convention (Typ)); Set_Depends_On_Private (Itype, Has_Private_Component (Typ)); Set_Etype (Itype, Base_Type (Typ)); Set_Has_Alignment_Clause (Itype, Has_Alignment_Clause (Typ)); Set_Is_Aliased (Itype, Is_Aliased (Typ)); Set_Depends_On_Private (Itype, Depends_On_Private (Typ)); Copy_Suppress_Status (Index_Check, Typ, Itype); Copy_Suppress_Status (Length_Check, Typ, Itype); Set_First_Index (Itype, First (Index_Constraints)); Set_Is_Constrained (Itype, True); Set_Is_Internal (Itype, True); -- A simple optimization: purely positional aggregates of static -- components should be passed to gigi unexpanded whenever possible, and -- regardless of the staticness of the bounds themselves. Subsequent -- checks in exp_aggr verify that type is not packed, etc. Set_Size_Known_At_Compile_Time (Itype, Is_Fully_Positional and then Comes_From_Source (N) and then Size_Known_At_Compile_Time (Component_Type (Typ))); -- We always need a freeze node for a packed array subtype, so that we -- can build the Packed_Array_Type corresponding to the subtype. If -- expansion is disabled, the packed array subtype is not built, and we -- must not generate a freeze node for the type, or else it will appear -- incomplete to gigi. if Is_Packed (Itype) and then not In_Spec_Expression and then Expander_Active then Freeze_Itype (Itype, N); end if; return Itype; end Array_Aggr_Subtype; -------------------------------- -- Check_Misspelled_Component -- -------------------------------- procedure Check_Misspelled_Component (Elements : Elist_Id; Component : Node_Id) is Max_Suggestions : constant := 2; Nr_Of_Suggestions : Natural := 0; Suggestion_1 : Entity_Id := Empty; Suggestion_2 : Entity_Id := Empty; Component_Elmt : Elmt_Id; begin -- All the components of List are matched against Component and a count -- is maintained of possible misspellings. When at the end of the -- the analysis there are one or two (not more!) possible misspellings, -- these misspellings will be suggested as possible correction. Component_Elmt := First_Elmt (Elements); while Nr_Of_Suggestions <= Max_Suggestions and then Present (Component_Elmt) loop if Is_Bad_Spelling_Of (Chars (Node (Component_Elmt)), Chars (Component)) then Nr_Of_Suggestions := Nr_Of_Suggestions + 1; case Nr_Of_Suggestions is when 1 => Suggestion_1 := Node (Component_Elmt); when 2 => Suggestion_2 := Node (Component_Elmt); when others => exit; end case; end if; Next_Elmt (Component_Elmt); end loop; -- Report at most two suggestions if Nr_Of_Suggestions = 1 then Error_Msg_NE -- CODEFIX ("\possible misspelling of&", Component, Suggestion_1); elsif Nr_Of_Suggestions = 2 then Error_Msg_Node_2 := Suggestion_2; Error_Msg_NE -- CODEFIX ("\possible misspelling of& or&", Component, Suggestion_1); end if; end Check_Misspelled_Component; ---------------------------------------- -- Check_Expr_OK_In_Limited_Aggregate -- ---------------------------------------- procedure Check_Expr_OK_In_Limited_Aggregate (Expr : Node_Id) is begin if Is_Limited_Type (Etype (Expr)) and then Comes_From_Source (Expr) and then not In_Instance_Body then if not OK_For_Limited_Init (Etype (Expr), Expr) then Error_Msg_N ("initialization not allowed for limited types", Expr); Explain_Limited_Type (Etype (Expr), Expr); end if; end if; end Check_Expr_OK_In_Limited_Aggregate; ---------------------------------------- -- Check_Static_Discriminated_Subtype -- ---------------------------------------- procedure Check_Static_Discriminated_Subtype (T : Entity_Id; V : Node_Id) is Disc : constant Entity_Id := First_Discriminant (T); Comp : Entity_Id; Ind : Entity_Id; begin if Has_Record_Rep_Clause (T) then return; elsif Present (Next_Discriminant (Disc)) then return; elsif Nkind (V) /= N_Integer_Literal then return; end if; Comp := First_Component (T); while Present (Comp) loop if Is_Scalar_Type (Etype (Comp)) then null; elsif Is_Private_Type (Etype (Comp)) and then Present (Full_View (Etype (Comp))) and then Is_Scalar_Type (Full_View (Etype (Comp))) then null; elsif Is_Array_Type (Etype (Comp)) then if Is_Bit_Packed_Array (Etype (Comp)) then return; end if; Ind := First_Index (Etype (Comp)); while Present (Ind) loop if Nkind (Ind) /= N_Range or else Nkind (Low_Bound (Ind)) /= N_Integer_Literal or else Nkind (High_Bound (Ind)) /= N_Integer_Literal then return; end if; Next_Index (Ind); end loop; else return; end if; Next_Component (Comp); end loop; -- On exit, all components have statically known sizes Set_Size_Known_At_Compile_Time (T); end Check_Static_Discriminated_Subtype; -------------------------------- -- Make_String_Into_Aggregate -- -------------------------------- procedure Make_String_Into_Aggregate (N : Node_Id) is Exprs : constant List_Id := New_List; Loc : constant Source_Ptr := Sloc (N); Str : constant String_Id := Strval (N); Strlen : constant Nat := String_Length (Str); C : Char_Code; C_Node : Node_Id; New_N : Node_Id; P : Source_Ptr; begin P := Loc + 1; for J in 1 .. Strlen loop C := Get_String_Char (Str, J); Set_Character_Literal_Name (C); C_Node := Make_Character_Literal (P, Chars => Name_Find, Char_Literal_Value => UI_From_CC (C)); Set_Etype (C_Node, Any_Character); Append_To (Exprs, C_Node); P := P + 1; -- Something special for wide strings??? end loop; New_N := Make_Aggregate (Loc, Expressions => Exprs); Set_Analyzed (New_N); Set_Etype (New_N, Any_Composite); Rewrite (N, New_N); end Make_String_Into_Aggregate; ----------------------- -- Resolve_Aggregate -- ----------------------- procedure Resolve_Aggregate (N : Node_Id; Typ : Entity_Id) is Pkind : constant Node_Kind := Nkind (Parent (N)); Aggr_Subtyp : Entity_Id; -- The actual aggregate subtype. This is not necessarily the same as Typ -- which is the subtype of the context in which the aggregate was found. begin -- Ignore junk empty aggregate resulting from parser error if No (Expressions (N)) and then No (Component_Associations (N)) and then not Null_Record_Present (N) then return; end if; -- Check for aggregates not allowed in configurable run-time mode. -- We allow all cases of aggregates that do not come from source, since -- these are all assumed to be small (e.g. bounds of a string literal). -- We also allow aggregates of types we know to be small. if not Support_Aggregates_On_Target and then Comes_From_Source (N) and then (not Known_Static_Esize (Typ) or else Esize (Typ) > 64) then Error_Msg_CRT ("aggregate", N); end if; -- Ada 2005 (AI-287): Limited aggregates allowed if Is_Limited_Type (Typ) and then Ada_Version < Ada_05 then Error_Msg_N ("aggregate type cannot be limited", N); Explain_Limited_Type (Typ, N); elsif Is_Class_Wide_Type (Typ) then Error_Msg_N ("type of aggregate cannot be class-wide", N); elsif Typ = Any_String or else Typ = Any_Composite then Error_Msg_N ("no unique type for aggregate", N); Set_Etype (N, Any_Composite); elsif Is_Array_Type (Typ) and then Null_Record_Present (N) then Error_Msg_N ("null record forbidden in array aggregate", N); elsif Is_Record_Type (Typ) then Resolve_Record_Aggregate (N, Typ); elsif Is_Array_Type (Typ) then -- First a special test, for the case of a positional aggregate -- of characters which can be replaced by a string literal. -- Do not perform this transformation if this was a string literal to -- start with, whose components needed constraint checks, or if the -- component type is non-static, because it will require those checks -- and be transformed back into an aggregate. if Number_Dimensions (Typ) = 1 and then Is_Standard_Character_Type (Component_Type (Typ)) and then No (Component_Associations (N)) and then not Is_Limited_Composite (Typ) and then not Is_Private_Composite (Typ) and then not Is_Bit_Packed_Array (Typ) and then Nkind (Original_Node (Parent (N))) /= N_String_Literal and then Is_Static_Subtype (Component_Type (Typ)) then declare Expr : Node_Id; begin Expr := First (Expressions (N)); while Present (Expr) loop exit when Nkind (Expr) /= N_Character_Literal; Next (Expr); end loop; if No (Expr) then Start_String; Expr := First (Expressions (N)); while Present (Expr) loop Store_String_Char (UI_To_CC (Char_Literal_Value (Expr))); Next (Expr); end loop; Rewrite (N, Make_String_Literal (Sloc (N), End_String)); Analyze_And_Resolve (N, Typ); return; end if; end; end if; -- Here if we have a real aggregate to deal with Array_Aggregate : declare Aggr_Resolved : Boolean; Aggr_Typ : constant Entity_Id := Etype (Typ); -- This is the unconstrained array type, which is the type against -- which the aggregate is to be resolved. Typ itself is the array -- type of the context which may not be the same subtype as the -- subtype for the final aggregate. begin -- In the following we determine whether an others choice is -- allowed inside the array aggregate. The test checks the context -- in which the array aggregate occurs. If the context does not -- permit it, or the aggregate type is unconstrained, an others -- choice is not allowed. -- If expansion is disabled (generic context, or semantics-only -- mode) actual subtypes cannot be constructed, and the type of an -- object may be its unconstrained nominal type. However, if the -- context is an assignment, we assume that "others" is allowed, -- because the target of the assignment will have a constrained -- subtype when fully compiled. -- Note that there is no node for Explicit_Actual_Parameter. -- To test for this context we therefore have to test for node -- N_Parameter_Association which itself appears only if there is a -- formal parameter. Consequently we also need to test for -- N_Procedure_Call_Statement or N_Function_Call. Set_Etype (N, Aggr_Typ); -- May be overridden later on if Is_Constrained (Typ) and then (Pkind = N_Assignment_Statement or else Pkind = N_Parameter_Association or else Pkind = N_Function_Call or else Pkind = N_Procedure_Call_Statement or else Pkind = N_Generic_Association or else Pkind = N_Formal_Object_Declaration or else Pkind = N_Simple_Return_Statement or else Pkind = N_Object_Declaration or else Pkind = N_Component_Declaration or else Pkind = N_Parameter_Specification or else Pkind = N_Qualified_Expression or else Pkind = N_Aggregate or else Pkind = N_Extension_Aggregate or else Pkind = N_Component_Association) then Aggr_Resolved := Resolve_Array_Aggregate (N, Index => First_Index (Aggr_Typ), Index_Constr => First_Index (Typ), Component_Typ => Component_Type (Typ), Others_Allowed => True); elsif not Expander_Active and then Pkind = N_Assignment_Statement then Aggr_Resolved := Resolve_Array_Aggregate (N, Index => First_Index (Aggr_Typ), Index_Constr => First_Index (Typ), Component_Typ => Component_Type (Typ), Others_Allowed => True); else Aggr_Resolved := Resolve_Array_Aggregate (N, Index => First_Index (Aggr_Typ), Index_Constr => First_Index (Aggr_Typ), Component_Typ => Component_Type (Typ), Others_Allowed => False); end if; if not Aggr_Resolved then Aggr_Subtyp := Any_Composite; else Aggr_Subtyp := Array_Aggr_Subtype (N, Typ); end if; Set_Etype (N, Aggr_Subtyp); end Array_Aggregate; elsif Is_Private_Type (Typ) and then Present (Full_View (Typ)) and then In_Inlined_Body and then Is_Composite_Type (Full_View (Typ)) then Resolve (N, Full_View (Typ)); else Error_Msg_N ("illegal context for aggregate", N); end if; -- If we can determine statically that the evaluation of the aggregate -- raises Constraint_Error, then replace the aggregate with an -- N_Raise_Constraint_Error node, but set the Etype to the right -- aggregate subtype. Gigi needs this. if Raises_Constraint_Error (N) then Aggr_Subtyp := Etype (N); Rewrite (N, Make_Raise_Constraint_Error (Sloc (N), Reason => CE_Range_Check_Failed)); Set_Raises_Constraint_Error (N); Set_Etype (N, Aggr_Subtyp); Set_Analyzed (N); end if; end Resolve_Aggregate; ----------------------------- -- Resolve_Array_Aggregate -- ----------------------------- function Resolve_Array_Aggregate (N : Node_Id; Index : Node_Id; Index_Constr : Node_Id; Component_Typ : Entity_Id; Others_Allowed : Boolean) return Boolean is Loc : constant Source_Ptr := Sloc (N); Failure : constant Boolean := False; Success : constant Boolean := True; Index_Typ : constant Entity_Id := Etype (Index); Index_Typ_Low : constant Node_Id := Type_Low_Bound (Index_Typ); Index_Typ_High : constant Node_Id := Type_High_Bound (Index_Typ); -- The type of the index corresponding to the array sub-aggregate along -- with its low and upper bounds. Index_Base : constant Entity_Id := Base_Type (Index_Typ); Index_Base_Low : constant Node_Id := Type_Low_Bound (Index_Base); Index_Base_High : constant Node_Id := Type_High_Bound (Index_Base); -- Ditto for the base type function Add (Val : Uint; To : Node_Id) return Node_Id; -- Creates a new expression node where Val is added to expression To. -- Tries to constant fold whenever possible. To must be an already -- analyzed expression. procedure Check_Bound (BH : Node_Id; AH : in out Node_Id); -- Checks that AH (the upper bound of an array aggregate) is <= BH -- (the upper bound of the index base type). If the check fails a -- warning is emitted, the Raises_Constraint_Error flag of N is set, -- and AH is replaced with a duplicate of BH. procedure Check_Bounds (L, H : Node_Id; AL, AH : Node_Id); -- Checks that range AL .. AH is compatible with range L .. H. Emits a -- warning if not and sets the Raises_Constraint_Error flag in N. procedure Check_Length (L, H : Node_Id; Len : Uint); -- Checks that range L .. H contains at least Len elements. Emits a -- warning if not and sets the Raises_Constraint_Error flag in N. function Dynamic_Or_Null_Range (L, H : Node_Id) return Boolean; -- Returns True if range L .. H is dynamic or null procedure Get (Value : out Uint; From : Node_Id; OK : out Boolean); -- Given expression node From, this routine sets OK to False if it -- cannot statically evaluate From. Otherwise it stores this static -- value into Value. function Resolve_Aggr_Expr (Expr : Node_Id; Single_Elmt : Boolean) return Boolean; -- Resolves aggregate expression Expr. Returns False if resolution -- fails. If Single_Elmt is set to False, the expression Expr may be -- used to initialize several array aggregate elements (this can happen -- for discrete choices such as "L .. H => Expr" or the others choice). -- In this event we do not resolve Expr unless expansion is disabled. -- To know why, see the DELAYED COMPONENT RESOLUTION note above. --------- -- Add -- --------- function Add (Val : Uint; To : Node_Id) return Node_Id is Expr_Pos : Node_Id; Expr : Node_Id; To_Pos : Node_Id; begin if Raises_Constraint_Error (To) then return To; end if; -- First test if we can do constant folding if Compile_Time_Known_Value (To) or else Nkind (To) = N_Integer_Literal then Expr_Pos := Make_Integer_Literal (Loc, Expr_Value (To) + Val); Set_Is_Static_Expression (Expr_Pos); Set_Etype (Expr_Pos, Etype (To)); Set_Analyzed (Expr_Pos, Analyzed (To)); if not Is_Enumeration_Type (Index_Typ) then Expr := Expr_Pos; -- If we are dealing with enumeration return -- Index_Typ'Val (Expr_Pos) else Expr := Make_Attribute_Reference (Loc, Prefix => New_Reference_To (Index_Typ, Loc), Attribute_Name => Name_Val, Expressions => New_List (Expr_Pos)); end if; return Expr; end if; -- If we are here no constant folding possible if not Is_Enumeration_Type (Index_Base) then Expr := Make_Op_Add (Loc, Left_Opnd => Duplicate_Subexpr (To), Right_Opnd => Make_Integer_Literal (Loc, Val)); -- If we are dealing with enumeration return -- Index_Typ'Val (Index_Typ'Pos (To) + Val) else To_Pos := Make_Attribute_Reference (Loc, Prefix => New_Reference_To (Index_Typ, Loc), Attribute_Name => Name_Pos, Expressions => New_List (Duplicate_Subexpr (To))); Expr_Pos := Make_Op_Add (Loc, Left_Opnd => To_Pos, Right_Opnd => Make_Integer_Literal (Loc, Val)); Expr := Make_Attribute_Reference (Loc, Prefix => New_Reference_To (Index_Typ, Loc), Attribute_Name => Name_Val, Expressions => New_List (Expr_Pos)); end if; return Expr; end Add; ----------------- -- Check_Bound -- ----------------- procedure Check_Bound (BH : Node_Id; AH : in out Node_Id) is Val_BH : Uint; Val_AH : Uint; OK_BH : Boolean; OK_AH : Boolean; begin Get (Value => Val_BH, From => BH, OK => OK_BH); Get (Value => Val_AH, From => AH, OK => OK_AH); if OK_BH and then OK_AH and then Val_BH < Val_AH then Set_Raises_Constraint_Error (N); Error_Msg_N ("upper bound out of range?", AH); Error_Msg_N ("\Constraint_Error will be raised at run-time?", AH); -- You need to set AH to BH or else in the case of enumerations -- indices we will not be able to resolve the aggregate bounds. AH := Duplicate_Subexpr (BH); end if; end Check_Bound; ------------------ -- Check_Bounds -- ------------------ procedure Check_Bounds (L, H : Node_Id; AL, AH : Node_Id) is Val_L : Uint; Val_H : Uint; Val_AL : Uint; Val_AH : Uint; OK_L : Boolean; OK_H : Boolean; OK_AL : Boolean; OK_AH : Boolean; pragma Warnings (Off, OK_AL); pragma Warnings (Off, OK_AH); begin if Raises_Constraint_Error (N) or else Dynamic_Or_Null_Range (AL, AH) then return; end if; Get (Value => Val_L, From => L, OK => OK_L); Get (Value => Val_H, From => H, OK => OK_H); Get (Value => Val_AL, From => AL, OK => OK_AL); Get (Value => Val_AH, From => AH, OK => OK_AH); if OK_L and then Val_L > Val_AL then Set_Raises_Constraint_Error (N); Error_Msg_N ("lower bound of aggregate out of range?", N); Error_Msg_N ("\Constraint_Error will be raised at run-time?", N); end if; if OK_H and then Val_H < Val_AH then Set_Raises_Constraint_Error (N); Error_Msg_N ("upper bound of aggregate out of range?", N); Error_Msg_N ("\Constraint_Error will be raised at run-time?", N); end if; end Check_Bounds; ------------------ -- Check_Length -- ------------------ procedure Check_Length (L, H : Node_Id; Len : Uint) is Val_L : Uint; Val_H : Uint; OK_L : Boolean; OK_H : Boolean; Range_Len : Uint; begin if Raises_Constraint_Error (N) then return; end if; Get (Value => Val_L, From => L, OK => OK_L); Get (Value => Val_H, From => H, OK => OK_H); if not OK_L or else not OK_H then return; end if; -- If null range length is zero if Val_L > Val_H then Range_Len := Uint_0; else Range_Len := Val_H - Val_L + 1; end if; if Range_Len < Len then Set_Raises_Constraint_Error (N); Error_Msg_N ("too many elements?", N); Error_Msg_N ("\Constraint_Error will be raised at run-time?", N); end if; end Check_Length; --------------------------- -- Dynamic_Or_Null_Range -- --------------------------- function Dynamic_Or_Null_Range (L, H : Node_Id) return Boolean is Val_L : Uint; Val_H : Uint; OK_L : Boolean; OK_H : Boolean; begin Get (Value => Val_L, From => L, OK => OK_L); Get (Value => Val_H, From => H, OK => OK_H); return not OK_L or else not OK_H or else not Is_OK_Static_Expression (L) or else not Is_OK_Static_Expression (H) or else Val_L > Val_H; end Dynamic_Or_Null_Range; --------- -- Get -- --------- procedure Get (Value : out Uint; From : Node_Id; OK : out Boolean) is begin OK := True; if Compile_Time_Known_Value (From) then Value := Expr_Value (From); -- If expression From is something like Some_Type'Val (10) then -- Value = 10 elsif Nkind (From) = N_Attribute_Reference and then Attribute_Name (From) = Name_Val and then Compile_Time_Known_Value (First (Expressions (From))) then Value := Expr_Value (First (Expressions (From))); else Value := Uint_0; OK := False; end if; end Get; ----------------------- -- Resolve_Aggr_Expr -- ----------------------- function Resolve_Aggr_Expr (Expr : Node_Id; Single_Elmt : Boolean) return Boolean is Nxt_Ind : constant Node_Id := Next_Index (Index); Nxt_Ind_Constr : constant Node_Id := Next_Index (Index_Constr); -- Index is the current index corresponding to the expression Resolution_OK : Boolean := True; -- Set to False if resolution of the expression failed begin -- If the array type against which we are resolving the aggregate -- has several dimensions, the expressions nested inside the -- aggregate must be further aggregates (or strings). if Present (Nxt_Ind) then if Nkind (Expr) /= N_Aggregate then -- A string literal can appear where a one-dimensional array -- of characters is expected. If the literal looks like an -- operator, it is still an operator symbol, which will be -- transformed into a string when analyzed. if Is_Character_Type (Component_Typ) and then No (Next_Index (Nxt_Ind)) and then Nkind_In (Expr, N_String_Literal, N_Operator_Symbol) then -- A string literal used in a multidimensional array -- aggregate in place of the final one-dimensional -- aggregate must not be enclosed in parentheses. if Paren_Count (Expr) /= 0 then Error_Msg_N ("no parenthesis allowed here", Expr); end if; Make_String_Into_Aggregate (Expr); else Error_Msg_N ("nested array aggregate expected", Expr); -- If the expression is parenthesized, this may be -- a missing component association for a 1-aggregate. if Paren_Count (Expr) > 0 then Error_Msg_N ("\if single-component aggregate is intended," & " write e.g. (1 ='> ...)", Expr); end if; return Failure; end if; end if; -- Ada 2005 (AI-231): Propagate the type to the nested aggregate. -- Required to check the null-exclusion attribute (if present). -- This value may be overridden later on. Set_Etype (Expr, Etype (N)); Resolution_OK := Resolve_Array_Aggregate (Expr, Nxt_Ind, Nxt_Ind_Constr, Component_Typ, Others_Allowed); -- Do not resolve the expressions of discrete or others choices -- unless the expression covers a single component, or the expander -- is inactive. elsif Single_Elmt or else not Expander_Active or else In_Spec_Expression then Analyze_And_Resolve (Expr, Component_Typ); Check_Expr_OK_In_Limited_Aggregate (Expr); Check_Non_Static_Context (Expr); Aggregate_Constraint_Checks (Expr, Component_Typ); Check_Unset_Reference (Expr); end if; if Raises_Constraint_Error (Expr) and then Nkind (Parent (Expr)) /= N_Component_Association then Set_Raises_Constraint_Error (N); end if; -- If the expression has been marked as requiring a range check, -- then generate it here. if Do_Range_Check (Expr) then Set_Do_Range_Check (Expr, False); Generate_Range_Check (Expr, Component_Typ, CE_Range_Check_Failed); end if; return Resolution_OK; end Resolve_Aggr_Expr; -- Variables local to Resolve_Array_Aggregate Assoc : Node_Id; Choice : Node_Id; Expr : Node_Id; Discard : Node_Id; pragma Warnings (Off, Discard); Aggr_Low : Node_Id := Empty; Aggr_High : Node_Id := Empty; -- The actual low and high bounds of this sub-aggregate Choices_Low : Node_Id := Empty; Choices_High : Node_Id := Empty; -- The lowest and highest discrete choices values for a named aggregate Nb_Elements : Uint := Uint_0; -- The number of elements in a positional aggregate Others_Present : Boolean := False; Nb_Choices : Nat := 0; -- Contains the overall number of named choices in this sub-aggregate Nb_Discrete_Choices : Nat := 0; -- The overall number of discrete choices (not counting others choice) Case_Table_Size : Nat; -- Contains the size of the case table needed to sort aggregate choices -- Start of processing for Resolve_Array_Aggregate begin -- Ignore junk empty aggregate resulting from parser error if No (Expressions (N)) and then No (Component_Associations (N)) and then not Null_Record_Present (N) then return False; end if; -- STEP 1: make sure the aggregate is correctly formatted if Present (Component_Associations (N)) then Assoc := First (Component_Associations (N)); while Present (Assoc) loop Choice := First (Choices (Assoc)); while Present (Choice) loop if Nkind (Choice) = N_Others_Choice then Others_Present := True; if Choice /= First (Choices (Assoc)) or else Present (Next (Choice)) then Error_Msg_N ("OTHERS must appear alone in a choice list", Choice); return Failure; end if; if Present (Next (Assoc)) then Error_Msg_N ("OTHERS must appear last in an aggregate", Choice); return Failure; end if; if Ada_Version = Ada_83 and then Assoc /= First (Component_Associations (N)) and then Nkind_In (Parent (N), N_Assignment_Statement, N_Object_Declaration) then Error_Msg_N ("(Ada 83) illegal context for OTHERS choice", N); end if; end if; Nb_Choices := Nb_Choices + 1; Next (Choice); end loop; Next (Assoc); end loop; end if; -- At this point we know that the others choice, if present, is by -- itself and appears last in the aggregate. Check if we have mixed -- positional and discrete associations (other than the others choice). if Present (Expressions (N)) and then (Nb_Choices > 1 or else (Nb_Choices = 1 and then not Others_Present)) then Error_Msg_N ("named association cannot follow positional association", First (Choices (First (Component_Associations (N))))); return Failure; end if; -- Test for the validity of an others choice if present if Others_Present and then not Others_Allowed then Error_Msg_N ("OTHERS choice not allowed here", First (Choices (First (Component_Associations (N))))); return Failure; end if; -- Protect against cascaded errors if Etype (Index_Typ) = Any_Type then return Failure; end if; -- STEP 2: Process named components if No (Expressions (N)) then if Others_Present then Case_Table_Size := Nb_Choices - 1; else Case_Table_Size := Nb_Choices; end if; Step_2 : declare Low : Node_Id; High : Node_Id; -- Denote the lowest and highest values in an aggregate choice Hi_Val : Uint; Lo_Val : Uint; -- High end of one range and Low end of the next. Should be -- contiguous if there is no hole in the list of values. Missing_Values : Boolean; -- Set True if missing index values S_Low : Node_Id := Empty; S_High : Node_Id := Empty; -- if a choice in an aggregate is a subtype indication these -- denote the lowest and highest values of the subtype Table : Case_Table_Type (1 .. Case_Table_Size); -- Used to sort all the different choice values Single_Choice : Boolean; -- Set to true every time there is a single discrete choice in a -- discrete association Prev_Nb_Discrete_Choices : Nat; -- Used to keep track of the number of discrete choices in the -- current association. begin -- STEP 2 (A): Check discrete choices validity Assoc := First (Component_Associations (N)); while Present (Assoc) loop Prev_Nb_Discrete_Choices := Nb_Discrete_Choices; Choice := First (Choices (Assoc)); loop Analyze (Choice); if Nkind (Choice) = N_Others_Choice then Single_Choice := False; exit; -- Test for subtype mark without constraint elsif Is_Entity_Name (Choice) and then Is_Type (Entity (Choice)) then if Base_Type (Entity (Choice)) /= Index_Base then Error_Msg_N ("invalid subtype mark in aggregate choice", Choice); return Failure; end if; -- Case of subtype indication elsif Nkind (Choice) = N_Subtype_Indication then Resolve_Discrete_Subtype_Indication (Choice, Index_Base); -- Does the subtype indication evaluation raise CE ? Get_Index_Bounds (Subtype_Mark (Choice), S_Low, S_High); Get_Index_Bounds (Choice, Low, High); Check_Bounds (S_Low, S_High, Low, High); -- Case of range or expression else Resolve (Choice, Index_Base); Check_Unset_Reference (Choice); Check_Non_Static_Context (Choice); -- Do not range check a choice. This check is redundant -- since this test is already done when we check that the -- bounds of the array aggregate are within range. Set_Do_Range_Check (Choice, False); end if; -- If we could not resolve the discrete choice stop here if Etype (Choice) = Any_Type then return Failure; -- If the discrete choice raises CE get its original bounds elsif Nkind (Choice) = N_Raise_Constraint_Error then Set_Raises_Constraint_Error (N); Get_Index_Bounds (Original_Node (Choice), Low, High); -- Otherwise get its bounds as usual else Get_Index_Bounds (Choice, Low, High); end if; if (Dynamic_Or_Null_Range (Low, High) or else (Nkind (Choice) = N_Subtype_Indication and then Dynamic_Or_Null_Range (S_Low, S_High))) and then Nb_Choices /= 1 then Error_Msg_N ("dynamic or empty choice in aggregate " & "must be the only choice", Choice); return Failure; end if; Nb_Discrete_Choices := Nb_Discrete_Choices + 1; Table (Nb_Discrete_Choices).Choice_Lo := Low; Table (Nb_Discrete_Choices).Choice_Hi := High; Next (Choice); if No (Choice) then -- Check if we have a single discrete choice and whether -- this discrete choice specifies a single value. Single_Choice := (Nb_Discrete_Choices = Prev_Nb_Discrete_Choices + 1) and then (Low = High); exit; end if; end loop; -- Ada 2005 (AI-231) if Ada_Version >= Ada_05 and then Known_Null (Expression (Assoc)) then Check_Can_Never_Be_Null (Etype (N), Expression (Assoc)); end if; -- Ada 2005 (AI-287): In case of default initialized component -- we delay the resolution to the expansion phase. if Box_Present (Assoc) then -- Ada 2005 (AI-287): In case of default initialization of a -- component the expander will generate calls to the -- corresponding initialization subprogram. null; elsif not Resolve_Aggr_Expr (Expression (Assoc), Single_Elmt => Single_Choice) then return Failure; -- Check incorrect use of dynamically tagged expression -- We differentiate here two cases because the expression may -- not be decorated. For example, the analysis and resolution -- of the expression associated with the others choice will be -- done later with the full aggregate. In such case we -- duplicate the expression tree to analyze the copy and -- perform the required check. elsif not Present (Etype (Expression (Assoc))) then declare Save_Analysis : constant Boolean := Full_Analysis; Expr : constant Node_Id := New_Copy_Tree (Expression (Assoc)); begin Expander_Mode_Save_And_Set (False); Full_Analysis := False; Analyze (Expr); Full_Analysis := Save_Analysis; Expander_Mode_Restore; if Is_Tagged_Type (Etype (Expr)) then Check_Dynamically_Tagged_Expression (Expr => Expr, Typ => Component_Type (Etype (N)), Related_Nod => N); end if; end; elsif Is_Tagged_Type (Etype (Expression (Assoc))) then Check_Dynamically_Tagged_Expression (Expr => Expression (Assoc), Typ => Component_Type (Etype (N)), Related_Nod => N); end if; Next (Assoc); end loop; -- If aggregate contains more than one choice then these must be -- static. Sort them and check that they are contiguous. if Nb_Discrete_Choices > 1 then Sort_Case_Table (Table); Missing_Values := False; Outer : for J in 1 .. Nb_Discrete_Choices - 1 loop if Expr_Value (Table (J).Choice_Hi) >= Expr_Value (Table (J + 1).Choice_Lo) then Error_Msg_N ("duplicate choice values in array aggregate", Table (J).Choice_Hi); return Failure; elsif not Others_Present then Hi_Val := Expr_Value (Table (J).Choice_Hi); Lo_Val := Expr_Value (Table (J + 1).Choice_Lo); -- If missing values, output error messages if Lo_Val - Hi_Val > 1 then -- Header message if not first missing value if not Missing_Values then Error_Msg_N ("missing index value(s) in array aggregate", N); Missing_Values := True; end if; -- Output values of missing indexes Lo_Val := Lo_Val - 1; Hi_Val := Hi_Val + 1; -- Enumeration type case if Is_Enumeration_Type (Index_Typ) then Error_Msg_Name_1 := Chars (Get_Enum_Lit_From_Pos (Index_Typ, Hi_Val, Loc)); if Lo_Val = Hi_Val then Error_Msg_N ("\ %", N); else Error_Msg_Name_2 := Chars (Get_Enum_Lit_From_Pos (Index_Typ, Lo_Val, Loc)); Error_Msg_N ("\ % .. %", N); end if; -- Integer types case else Error_Msg_Uint_1 := Hi_Val; if Lo_Val = Hi_Val then Error_Msg_N ("\ ^", N); else Error_Msg_Uint_2 := Lo_Val; Error_Msg_N ("\ ^ .. ^", N); end if; end if; end if; end if; end loop Outer; if Missing_Values then Set_Etype (N, Any_Composite); return Failure; end if; end if; -- STEP 2 (B): Compute aggregate bounds and min/max choices values if Nb_Discrete_Choices > 0 then Choices_Low := Table (1).Choice_Lo; Choices_High := Table (Nb_Discrete_Choices).Choice_Hi; end if; -- If Others is present, then bounds of aggregate come from the -- index constraint (not the choices in the aggregate itself). if Others_Present then Get_Index_Bounds (Index_Constr, Aggr_Low, Aggr_High); -- No others clause present else -- Special processing if others allowed and not present. This -- means that the bounds of the aggregate come from the index -- constraint (and the length must match). if Others_Allowed then Get_Index_Bounds (Index_Constr, Aggr_Low, Aggr_High); -- If others allowed, and no others present, then the array -- should cover all index values. If it does not, we will -- get a length check warning, but there is two cases where -- an additional warning is useful: -- If we have no positional components, and the length is -- wrong (which we can tell by others being allowed with -- missing components), and the index type is an enumeration -- type, then issue appropriate warnings about these missing -- components. They are only warnings, since the aggregate -- is fine, it's just the wrong length. We skip this check -- for standard character types (since there are no literals -- and it is too much trouble to concoct them), and also if -- any of the bounds have not-known-at-compile-time values. -- Another case warranting a warning is when the length is -- right, but as above we have an index type that is an -- enumeration, and the bounds do not match. This is a -- case where dubious sliding is allowed and we generate -- a warning that the bounds do not match. if No (Expressions (N)) and then Nkind (Index) = N_Range and then Is_Enumeration_Type (Etype (Index)) and then not Is_Standard_Character_Type (Etype (Index)) and then Compile_Time_Known_Value (Aggr_Low) and then Compile_Time_Known_Value (Aggr_High) and then Compile_Time_Known_Value (Choices_Low) and then Compile_Time_Known_Value (Choices_High) then declare ALo : constant Node_Id := Expr_Value_E (Aggr_Low); AHi : constant Node_Id := Expr_Value_E (Aggr_High); CLo : constant Node_Id := Expr_Value_E (Choices_Low); CHi : constant Node_Id := Expr_Value_E (Choices_High); Ent : Entity_Id; begin -- Warning case one, missing values at start/end. Only -- do the check if the number of entries is too small. if (Enumeration_Pos (CHi) - Enumeration_Pos (CLo)) < (Enumeration_Pos (AHi) - Enumeration_Pos (ALo)) then Error_Msg_N ("missing index value(s) in array aggregate?", N); -- Output missing value(s) at start if Chars (ALo) /= Chars (CLo) then Ent := Prev (CLo); if Chars (ALo) = Chars (Ent) then Error_Msg_Name_1 := Chars (ALo); Error_Msg_N ("\ %?", N); else Error_Msg_Name_1 := Chars (ALo); Error_Msg_Name_2 := Chars (Ent); Error_Msg_N ("\ % .. %?", N); end if; end if; -- Output missing value(s) at end if Chars (AHi) /= Chars (CHi) then Ent := Next (CHi); if Chars (AHi) = Chars (Ent) then Error_Msg_Name_1 := Chars (Ent); Error_Msg_N ("\ %?", N); else Error_Msg_Name_1 := Chars (Ent); Error_Msg_Name_2 := Chars (AHi); Error_Msg_N ("\ % .. %?", N); end if; end if; -- Warning case 2, dubious sliding. The First_Subtype -- test distinguishes between a constrained type where -- sliding is not allowed (so we will get a warning -- later that Constraint_Error will be raised), and -- the unconstrained case where sliding is permitted. elsif (Enumeration_Pos (CHi) - Enumeration_Pos (CLo)) = (Enumeration_Pos (AHi) - Enumeration_Pos (ALo)) and then Chars (ALo) /= Chars (CLo) and then not Is_Constrained (First_Subtype (Etype (N))) then Error_Msg_N ("bounds of aggregate do not match target?", N); end if; end; end if; end if; -- If no others, aggregate bounds come from aggregate Aggr_Low := Choices_Low; Aggr_High := Choices_High; end if; end Step_2; -- STEP 3: Process positional components else -- STEP 3 (A): Process positional elements Expr := First (Expressions (N)); Nb_Elements := Uint_0; while Present (Expr) loop Nb_Elements := Nb_Elements + 1; -- Ada 2005 (AI-231) if Ada_Version >= Ada_05 and then Known_Null (Expr) then Check_Can_Never_Be_Null (Etype (N), Expr); end if; if not Resolve_Aggr_Expr (Expr, Single_Elmt => True) then return Failure; end if; -- Check incorrect use of dynamically tagged expression if Is_Tagged_Type (Etype (Expr)) then Check_Dynamically_Tagged_Expression (Expr => Expr, Typ => Component_Type (Etype (N)), Related_Nod => N); end if; Next (Expr); end loop; if Others_Present then Assoc := Last (Component_Associations (N)); -- Ada 2005 (AI-231) if Ada_Version >= Ada_05 and then Known_Null (Assoc) then Check_Can_Never_Be_Null (Etype (N), Expression (Assoc)); end if; -- Ada 2005 (AI-287): In case of default initialized component -- we delay the resolution to the expansion phase. if Box_Present (Assoc) then -- Ada 2005 (AI-287): In case of default initialization -- of a component the expander will generate calls to -- the corresponding initialization subprogram. null; elsif not Resolve_Aggr_Expr (Expression (Assoc), Single_Elmt => False) then return Failure; -- Check incorrect use of dynamically tagged expression. The -- expression of the others choice has not been resolved yet. -- In order to diagnose the semantic error we create a duplicate -- tree to analyze it and perform the check. else declare Save_Analysis : constant Boolean := Full_Analysis; Expr : constant Node_Id := New_Copy_Tree (Expression (Assoc)); begin Expander_Mode_Save_And_Set (False); Full_Analysis := False; Analyze (Expr); Full_Analysis := Save_Analysis; Expander_Mode_Restore; if Is_Tagged_Type (Etype (Expr)) then Check_Dynamically_Tagged_Expression (Expr => Expr, Typ => Component_Type (Etype (N)), Related_Nod => N); end if; end; end if; end if; -- STEP 3 (B): Compute the aggregate bounds if Others_Present then Get_Index_Bounds (Index_Constr, Aggr_Low, Aggr_High); else if Others_Allowed then Get_Index_Bounds (Index_Constr, Aggr_Low, Discard); else Aggr_Low := Index_Typ_Low; end if; Aggr_High := Add (Nb_Elements - 1, To => Aggr_Low); Check_Bound (Index_Base_High, Aggr_High); end if; end if; -- STEP 4: Perform static aggregate checks and save the bounds -- Check (A) Check_Bounds (Index_Typ_Low, Index_Typ_High, Aggr_Low, Aggr_High); Check_Bounds (Index_Base_Low, Index_Base_High, Aggr_Low, Aggr_High); -- Check (B) if Others_Present and then Nb_Discrete_Choices > 0 then Check_Bounds (Aggr_Low, Aggr_High, Choices_Low, Choices_High); Check_Bounds (Index_Typ_Low, Index_Typ_High, Choices_Low, Choices_High); Check_Bounds (Index_Base_Low, Index_Base_High, Choices_Low, Choices_High); -- Check (C) elsif Others_Present and then Nb_Elements > 0 then Check_Length (Aggr_Low, Aggr_High, Nb_Elements); Check_Length (Index_Typ_Low, Index_Typ_High, Nb_Elements); Check_Length (Index_Base_Low, Index_Base_High, Nb_Elements); end if; if Raises_Constraint_Error (Aggr_Low) or else Raises_Constraint_Error (Aggr_High) then Set_Raises_Constraint_Error (N); end if; Aggr_Low := Duplicate_Subexpr (Aggr_Low); -- Do not duplicate Aggr_High if Aggr_High = Aggr_Low + Nb_Elements -- since the addition node returned by Add is not yet analyzed. Attach -- to tree and analyze first. Reset analyzed flag to insure it will get -- analyzed when it is a literal bound whose type must be properly set. if Others_Present or else Nb_Discrete_Choices > 0 then Aggr_High := Duplicate_Subexpr (Aggr_High); if Etype (Aggr_High) = Universal_Integer then Set_Analyzed (Aggr_High, False); end if; end if; Set_Aggregate_Bounds (N, Make_Range (Loc, Low_Bound => Aggr_Low, High_Bound => Aggr_High)); -- The bounds may contain expressions that must be inserted upwards. -- Attach them fully to the tree. After analysis, remove side effects -- from upper bound, if still needed. Set_Parent (Aggregate_Bounds (N), N); Analyze_And_Resolve (Aggregate_Bounds (N), Index_Typ); Check_Unset_Reference (Aggregate_Bounds (N)); if not Others_Present and then Nb_Discrete_Choices = 0 then Set_High_Bound (Aggregate_Bounds (N), Duplicate_Subexpr (High_Bound (Aggregate_Bounds (N)))); end if; return Success; end Resolve_Array_Aggregate; --------------------------------- -- Resolve_Extension_Aggregate -- --------------------------------- -- There are two cases to consider: -- a) If the ancestor part is a type mark, the components needed are -- the difference between the components of the expected type and the -- components of the given type mark. -- b) If the ancestor part is an expression, it must be unambiguous, -- and once we have its type we can also compute the needed components -- as in the previous case. In both cases, if the ancestor type is not -- the immediate ancestor, we have to build this ancestor recursively. -- In both cases discriminants of the ancestor type do not play a -- role in the resolution of the needed components, because inherited -- discriminants cannot be used in a type extension. As a result we can -- compute independently the list of components of the ancestor type and -- of the expected type. procedure Resolve_Extension_Aggregate (N : Node_Id; Typ : Entity_Id) is A : constant Node_Id := Ancestor_Part (N); A_Type : Entity_Id; I : Interp_Index; It : Interp; function Valid_Limited_Ancestor (Anc : Node_Id) return Boolean; -- If the type is limited, verify that the ancestor part is a legal -- expression (aggregate or function call, including 'Input)) that -- does not require a copy, as specified in 7.5 (2). function Valid_Ancestor_Type return Boolean; -- Verify that the type of the ancestor part is a non-private ancestor -- of the expected type, which must be a type extension. ---------------------------- -- Valid_Limited_Ancestor -- ---------------------------- function Valid_Limited_Ancestor (Anc : Node_Id) return Boolean is begin if Is_Entity_Name (Anc) and then Is_Type (Entity (Anc)) then return True; elsif Nkind_In (Anc, N_Aggregate, N_Function_Call) then return True; elsif Nkind (Anc) = N_Attribute_Reference and then Attribute_Name (Anc) = Name_Input then return True; elsif Nkind (Anc) = N_Qualified_Expression then return Valid_Limited_Ancestor (Expression (Anc)); else return False; end if; end Valid_Limited_Ancestor; ------------------------- -- Valid_Ancestor_Type -- ------------------------- function Valid_Ancestor_Type return Boolean is Imm_Type : Entity_Id; begin Imm_Type := Base_Type (Typ); while Is_Derived_Type (Imm_Type) loop if Etype (Imm_Type) = Base_Type (A_Type) then return True; -- The base type of the parent type may appear as a private -- extension if it is declared as such in a parent unit of -- the current one. For consistency of the subsequent analysis -- use the partial view for the ancestor part. elsif Is_Private_Type (Etype (Imm_Type)) and then Present (Full_View (Etype (Imm_Type))) and then Base_Type (A_Type) = Full_View (Etype (Imm_Type)) then A_Type := Etype (Imm_Type); return True; else Imm_Type := Etype (Base_Type (Imm_Type)); end if; end loop; -- If previous loop did not find a proper ancestor, report error Error_Msg_NE ("expect ancestor type of &", A, Typ); return False; end Valid_Ancestor_Type; -- Start of processing for Resolve_Extension_Aggregate begin -- Analyze the ancestor part and account for the case where it's -- a parameterless function call. Analyze (A); Check_Parameterless_Call (A); if not Is_Tagged_Type (Typ) then Error_Msg_N ("type of extension aggregate must be tagged", N); return; elsif Is_Limited_Type (Typ) then -- Ada 2005 (AI-287): Limited aggregates are allowed if Ada_Version < Ada_05 then Error_Msg_N ("aggregate type cannot be limited", N); Explain_Limited_Type (Typ, N); return; elsif Valid_Limited_Ancestor (A) then null; else Error_Msg_N ("limited ancestor part must be aggregate or function call", A); end if; elsif Is_Class_Wide_Type (Typ) then Error_Msg_N ("aggregate cannot be of a class-wide type", N); return; end if; if Is_Entity_Name (A) and then Is_Type (Entity (A)) then A_Type := Get_Full_View (Entity (A)); if Valid_Ancestor_Type then Set_Entity (A, A_Type); Set_Etype (A, A_Type); Validate_Ancestor_Part (N); Resolve_Record_Aggregate (N, Typ); end if; elsif Nkind (A) /= N_Aggregate then if Is_Overloaded (A) then A_Type := Any_Type; Get_First_Interp (A, I, It); while Present (It.Typ) loop -- Only consider limited interpretations in the Ada 2005 case if Is_Tagged_Type (It.Typ) and then (Ada_Version >= Ada_05 or else not Is_Limited_Type (It.Typ)) then if A_Type /= Any_Type then Error_Msg_N ("cannot resolve expression", A); return; else A_Type := It.Typ; end if; end if; Get_Next_Interp (I, It); end loop; if A_Type = Any_Type then if Ada_Version >= Ada_05 then Error_Msg_N ("ancestor part must be of a tagged type", A); else Error_Msg_N ("ancestor part must be of a nonlimited tagged type", A); end if; return; end if; else A_Type := Etype (A); end if; if Valid_Ancestor_Type then Resolve (A, A_Type); Check_Unset_Reference (A); Check_Non_Static_Context (A); -- The aggregate is illegal if the ancestor expression is a call -- to a function with a limited unconstrained result, unless the -- type of the aggregate is a null extension. This restriction -- was added in AI05-67 to simplify implementation. if Nkind (A) = N_Function_Call and then Is_Limited_Type (A_Type) and then not Is_Null_Extension (Typ) and then not Is_Constrained (A_Type) then Error_Msg_N ("type of limited ancestor part must be constrained", A); elsif Is_Class_Wide_Type (Etype (A)) and then Nkind (Original_Node (A)) = N_Function_Call then -- If the ancestor part is a dispatching call, it appears -- statically to be a legal ancestor, but it yields any -- member of the class, and it is not possible to determine -- whether it is an ancestor of the extension aggregate (much -- less which ancestor). It is not possible to determine the -- required components of the extension part. -- This check implements AI-306, which in fact was motivated -- by an ACT query to the ARG after this test was added. Error_Msg_N ("ancestor part must be statically tagged", A); else Resolve_Record_Aggregate (N, Typ); end if; end if; else Error_Msg_N ("no unique type for this aggregate", A); end if; end Resolve_Extension_Aggregate; ------------------------------ -- Resolve_Record_Aggregate -- ------------------------------ procedure Resolve_Record_Aggregate (N : Node_Id; Typ : Entity_Id) is Assoc : Node_Id; -- N_Component_Association node belonging to the input aggregate N Expr : Node_Id; Positional_Expr : Node_Id; Component : Entity_Id; Component_Elmt : Elmt_Id; Components : constant Elist_Id := New_Elmt_List; -- Components is the list of the record components whose value must -- be provided in the aggregate. This list does include discriminants. New_Assoc_List : constant List_Id := New_List; New_Assoc : Node_Id; -- New_Assoc_List is the newly built list of N_Component_Association -- nodes. New_Assoc is one such N_Component_Association node in it. -- Please note that while Assoc and New_Assoc contain the same -- kind of nodes, they are used to iterate over two different -- N_Component_Association lists. Others_Etype : Entity_Id := Empty; -- This variable is used to save the Etype of the last record component -- that takes its value from the others choice. Its purpose is: -- -- (a) make sure the others choice is useful -- -- (b) make sure the type of all the components whose value is -- subsumed by the others choice are the same. -- -- This variable is updated as a side effect of function Get_Value Is_Box_Present : Boolean := False; Others_Box : Boolean := False; -- Ada 2005 (AI-287): Variables used in case of default initialization -- to provide a functionality similar to Others_Etype. Box_Present -- indicates that the component takes its default initialization; -- Others_Box indicates that at least one component takes its default -- initialization. Similar to Others_Etype, they are also updated as a -- side effect of function Get_Value. procedure Add_Association (Component : Entity_Id; Expr : Node_Id; Assoc_List : List_Id; Is_Box_Present : Boolean := False); -- Builds a new N_Component_Association node which associates -- Component to expression Expr and adds it to the association -- list being built, either New_Assoc_List, or the association -- being built for an inner aggregate. function Discr_Present (Discr : Entity_Id) return Boolean; -- If aggregate N is a regular aggregate this routine will return True. -- Otherwise, if N is an extension aggregate, Discr is a discriminant -- whose value may already have been specified by N's ancestor part, -- this routine checks whether this is indeed the case and if so -- returns False, signaling that no value for Discr should appear in the -- N's aggregate part. Also, in this case, the routine appends to -- New_Assoc_List Discr the discriminant value specified in the ancestor -- part. function Get_Value (Compon : Node_Id; From : List_Id; Consider_Others_Choice : Boolean := False) return Node_Id; -- Given a record component stored in parameter Compon, the -- following function returns its value as it appears in the list -- From, which is a list of N_Component_Association nodes. If no -- component association has a choice for the searched component, -- the value provided by the others choice is returned, if there -- is one and Consider_Others_Choice is set to true. Otherwise -- Empty is returned. If there is more than one component association -- giving a value for the searched record component, an error message -- is emitted and the first found value is returned. -- -- If Consider_Others_Choice is set and the returned expression comes -- from the others choice, then Others_Etype is set as a side effect. -- An error message is emitted if the components taking their value -- from the others choice do not have same type. procedure Resolve_Aggr_Expr (Expr : Node_Id; Component : Node_Id); -- Analyzes and resolves expression Expr against the Etype of the -- Component. This routine also applies all appropriate checks to Expr. -- It finally saves a Expr in the newly created association list that -- will be attached to the final record aggregate. Note that if the -- Parent pointer of Expr is not set then Expr was produced with a -- New_Copy_Tree or some such. --------------------- -- Add_Association -- --------------------- procedure Add_Association (Component : Entity_Id; Expr : Node_Id; Assoc_List : List_Id; Is_Box_Present : Boolean := False) is Choice_List : constant List_Id := New_List; New_Assoc : Node_Id; begin Append (New_Occurrence_Of (Component, Sloc (Expr)), Choice_List); New_Assoc := Make_Component_Association (Sloc (Expr), Choices => Choice_List, Expression => Expr, Box_Present => Is_Box_Present); Append (New_Assoc, Assoc_List); end Add_Association; ------------------- -- Discr_Present -- ------------------- function Discr_Present (Discr : Entity_Id) return Boolean is Regular_Aggr : constant Boolean := Nkind (N) /= N_Extension_Aggregate; Loc : Source_Ptr; Ancestor : Node_Id; Discr_Expr : Node_Id; Ancestor_Typ : Entity_Id; Orig_Discr : Entity_Id; D : Entity_Id; D_Val : Elmt_Id := No_Elmt; -- stop junk warning Ancestor_Is_Subtyp : Boolean; begin if Regular_Aggr then return True; end if; Ancestor := Ancestor_Part (N); Ancestor_Typ := Etype (Ancestor); Loc := Sloc (Ancestor); -- For a private type with unknown discriminants, use the underlying -- record view if it is available. if Has_Unknown_Discriminants (Ancestor_Typ) and then Present (Full_View (Ancestor_Typ)) and then Present (Underlying_Record_View (Full_View (Ancestor_Typ))) then Ancestor_Typ := Underlying_Record_View (Full_View (Ancestor_Typ)); end if; Ancestor_Is_Subtyp := Is_Entity_Name (Ancestor) and then Is_Type (Entity (Ancestor)); -- If the ancestor part has no discriminants clearly N's aggregate -- part must provide a value for Discr. if not Has_Discriminants (Ancestor_Typ) then return True; -- If the ancestor part is an unconstrained subtype mark then the -- Discr must be present in N's aggregate part. elsif Ancestor_Is_Subtyp and then not Is_Constrained (Entity (Ancestor)) then return True; end if; -- Now look to see if Discr was specified in the ancestor part if Ancestor_Is_Subtyp then D_Val := First_Elmt (Discriminant_Constraint (Entity (Ancestor))); end if; Orig_Discr := Original_Record_Component (Discr); D := First_Discriminant (Ancestor_Typ); while Present (D) loop -- If Ancestor has already specified Disc value than insert its -- value in the final aggregate. if Original_Record_Component (D) = Orig_Discr then if Ancestor_Is_Subtyp then Discr_Expr := New_Copy_Tree (Node (D_Val)); else Discr_Expr := Make_Selected_Component (Loc, Prefix => Duplicate_Subexpr (Ancestor), Selector_Name => New_Occurrence_Of (Discr, Loc)); end if; Resolve_Aggr_Expr (Discr_Expr, Discr); return False; end if; Next_Discriminant (D); if Ancestor_Is_Subtyp then Next_Elmt (D_Val); end if; end loop; return True; end Discr_Present; --------------- -- Get_Value -- --------------- function Get_Value (Compon : Node_Id; From : List_Id; Consider_Others_Choice : Boolean := False) return Node_Id is Assoc : Node_Id; Expr : Node_Id := Empty; Selector_Name : Node_Id; begin Is_Box_Present := False; if Present (From) then Assoc := First (From); else return Empty; end if; while Present (Assoc) loop Selector_Name := First (Choices (Assoc)); while Present (Selector_Name) loop if Nkind (Selector_Name) = N_Others_Choice then if Consider_Others_Choice and then No (Expr) then -- We need to duplicate the expression for each -- successive component covered by the others choice. -- This is redundant if the others_choice covers only -- one component (small optimization possible???), but -- indispensable otherwise, because each one must be -- expanded individually to preserve side-effects. -- Ada 2005 (AI-287): In case of default initialization -- of components, we duplicate the corresponding default -- expression (from the record type declaration). The -- copy must carry the sloc of the association (not the -- original expression) to prevent spurious elaboration -- checks when the default includes function calls. if Box_Present (Assoc) then Others_Box := True; Is_Box_Present := True; if Expander_Active then return New_Copy_Tree (Expression (Parent (Compon)), New_Sloc => Sloc (Assoc)); else return Expression (Parent (Compon)); end if; else if Present (Others_Etype) and then Base_Type (Others_Etype) /= Base_Type (Etype (Compon)) then Error_Msg_N ("components in OTHERS choice must " & "have same type", Selector_Name); end if; Others_Etype := Etype (Compon); if Expander_Active then return New_Copy_Tree (Expression (Assoc)); else return Expression (Assoc); end if; end if; end if; elsif Chars (Compon) = Chars (Selector_Name) then if No (Expr) then -- Ada 2005 (AI-231) if Ada_Version >= Ada_05 and then Known_Null (Expression (Assoc)) then Check_Can_Never_Be_Null (Compon, Expression (Assoc)); end if; -- We need to duplicate the expression when several -- components are grouped together with a "|" choice. -- For instance "filed1 | filed2 => Expr" -- Ada 2005 (AI-287) if Box_Present (Assoc) then Is_Box_Present := True; -- Duplicate the default expression of the component -- from the record type declaration, so a new copy -- can be attached to the association. -- Note that we always copy the default expression, -- even when the association has a single choice, in -- order to create a proper association for the -- expanded aggregate. Expr := New_Copy_Tree (Expression (Parent (Compon))); else if Present (Next (Selector_Name)) then Expr := New_Copy_Tree (Expression (Assoc)); else Expr := Expression (Assoc); end if; end if; Generate_Reference (Compon, Selector_Name, 'm'); else Error_Msg_NE ("more than one value supplied for &", Selector_Name, Compon); end if; end if; Next (Selector_Name); end loop; Next (Assoc); end loop; return Expr; end Get_Value; ----------------------- -- Resolve_Aggr_Expr -- ----------------------- procedure Resolve_Aggr_Expr (Expr : Node_Id; Component : Node_Id) is New_C : Entity_Id := Component; Expr_Type : Entity_Id := Empty; function Has_Expansion_Delayed (Expr : Node_Id) return Boolean; -- If the expression is an aggregate (possibly qualified) then its -- expansion is delayed until the enclosing aggregate is expanded -- into assignments. In that case, do not generate checks on the -- expression, because they will be generated later, and will other- -- wise force a copy (to remove side-effects) that would leave a -- dynamic-sized aggregate in the code, something that gigi cannot -- handle. Relocate : Boolean; -- Set to True if the resolved Expr node needs to be relocated -- when attached to the newly created association list. This node -- need not be relocated if its parent pointer is not set. -- In fact in this case Expr is the output of a New_Copy_Tree call. -- if Relocate is True then we have analyzed the expression node -- in the original aggregate and hence it needs to be relocated -- when moved over the new association list. function Has_Expansion_Delayed (Expr : Node_Id) return Boolean is Kind : constant Node_Kind := Nkind (Expr); begin return (Nkind_In (Kind, N_Aggregate, N_Extension_Aggregate) and then Present (Etype (Expr)) and then Is_Record_Type (Etype (Expr)) and then Expansion_Delayed (Expr)) or else (Kind = N_Qualified_Expression and then Has_Expansion_Delayed (Expression (Expr))); end Has_Expansion_Delayed; -- Start of processing for Resolve_Aggr_Expr begin -- If the type of the component is elementary or the type of the -- aggregate does not contain discriminants, use the type of the -- component to resolve Expr. if Is_Elementary_Type (Etype (Component)) or else not Has_Discriminants (Etype (N)) then Expr_Type := Etype (Component); -- Otherwise we have to pick up the new type of the component from -- the new constrained subtype of the aggregate. In fact components -- which are of a composite type might be constrained by a -- discriminant, and we want to resolve Expr against the subtype were -- all discriminant occurrences are replaced with their actual value. else New_C := First_Component (Etype (N)); while Present (New_C) loop if Chars (New_C) = Chars (Component) then Expr_Type := Etype (New_C); exit; end if; Next_Component (New_C); end loop; pragma Assert (Present (Expr_Type)); -- For each range in an array type where a discriminant has been -- replaced with the constraint, check that this range is within -- the range of the base type. This checks is done in the init -- proc for regular objects, but has to be done here for -- aggregates since no init proc is called for them. if Is_Array_Type (Expr_Type) then declare Index : Node_Id; -- Range of the current constrained index in the array Orig_Index : Node_Id := First_Index (Etype (Component)); -- Range corresponding to the range Index above in the -- original unconstrained record type. The bounds of this -- range may be governed by discriminants. Unconstr_Index : Node_Id := First_Index (Etype (Expr_Type)); -- Range corresponding to the range Index above for the -- unconstrained array type. This range is needed to apply -- range checks. begin Index := First_Index (Expr_Type); while Present (Index) loop if Depends_On_Discriminant (Orig_Index) then Apply_Range_Check (Index, Etype (Unconstr_Index)); end if; Next_Index (Index); Next_Index (Orig_Index); Next_Index (Unconstr_Index); end loop; end; end if; end if; -- If the Parent pointer of Expr is not set, Expr is an expression -- duplicated by New_Tree_Copy (this happens for record aggregates -- that look like (Field1 | Filed2 => Expr) or (others => Expr)). -- Such a duplicated expression must be attached to the tree -- before analysis and resolution to enforce the rule that a tree -- fragment should never be analyzed or resolved unless it is -- attached to the current compilation unit. if No (Parent (Expr)) then Set_Parent (Expr, N); Relocate := False; else Relocate := True; end if; Analyze_And_Resolve (Expr, Expr_Type); Check_Expr_OK_In_Limited_Aggregate (Expr); Check_Non_Static_Context (Expr); Check_Unset_Reference (Expr); -- Check wrong use of class-wide types if Is_Class_Wide_Type (Etype (Expr)) then Error_Msg_N ("dynamically tagged expression not allowed", Expr); end if; if not Has_Expansion_Delayed (Expr) then Aggregate_Constraint_Checks (Expr, Expr_Type); end if; if Raises_Constraint_Error (Expr) then Set_Raises_Constraint_Error (N); end if; -- If the expression has been marked as requiring a range check, -- then generate it here. if Do_Range_Check (Expr) then Set_Do_Range_Check (Expr, False); Generate_Range_Check (Expr, Expr_Type, CE_Range_Check_Failed); end if; if Relocate then Add_Association (New_C, Relocate_Node (Expr), New_Assoc_List); else Add_Association (New_C, Expr, New_Assoc_List); end if; end Resolve_Aggr_Expr; -- Start of processing for Resolve_Record_Aggregate begin -- We may end up calling Duplicate_Subexpr on expressions that are -- attached to New_Assoc_List. For this reason we need to attach it -- to the tree by setting its parent pointer to N. This parent point -- will change in STEP 8 below. Set_Parent (New_Assoc_List, N); -- STEP 1: abstract type and null record verification if Is_Abstract_Type (Typ) then Error_Msg_N ("type of aggregate cannot be abstract", N); end if; if No (First_Entity (Typ)) and then Null_Record_Present (N) then Set_Etype (N, Typ); return; elsif Present (First_Entity (Typ)) and then Null_Record_Present (N) and then not Is_Tagged_Type (Typ) then Error_Msg_N ("record aggregate cannot be null", N); return; -- If the type has no components, then the aggregate should either -- have "null record", or in Ada 2005 it could instead have a single -- component association given by "others => <>". For Ada 95 we flag -- an error at this point, but for Ada 2005 we proceed with checking -- the associations below, which will catch the case where it's not -- an aggregate with "others => <>". Note that the legality of a <> -- aggregate for a null record type was established by AI05-016. elsif No (First_Entity (Typ)) and then Ada_Version < Ada_05 then Error_Msg_N ("record aggregate must be null", N); return; end if; -- STEP 2: Verify aggregate structure Step_2 : declare Selector_Name : Node_Id; Bad_Aggregate : Boolean := False; begin if Present (Component_Associations (N)) then Assoc := First (Component_Associations (N)); else Assoc := Empty; end if; while Present (Assoc) loop Selector_Name := First (Choices (Assoc)); while Present (Selector_Name) loop if Nkind (Selector_Name) = N_Identifier then null; elsif Nkind (Selector_Name) = N_Others_Choice then if Selector_Name /= First (Choices (Assoc)) or else Present (Next (Selector_Name)) then Error_Msg_N ("OTHERS must appear alone in a choice list", Selector_Name); return; elsif Present (Next (Assoc)) then Error_Msg_N ("OTHERS must appear last in an aggregate", Selector_Name); return; -- (Ada2005): If this is an association with a box, -- indicate that the association need not represent -- any component. elsif Box_Present (Assoc) then Others_Box := True; end if; else Error_Msg_N ("selector name should be identifier or OTHERS", Selector_Name); Bad_Aggregate := True; end if; Next (Selector_Name); end loop; Next (Assoc); end loop; if Bad_Aggregate then return; end if; end Step_2; -- STEP 3: Find discriminant Values Step_3 : declare Discrim : Entity_Id; Missing_Discriminants : Boolean := False; begin if Present (Expressions (N)) then Positional_Expr := First (Expressions (N)); else Positional_Expr := Empty; end if; if Has_Unknown_Discriminants (Typ) and then Present (Underlying_Record_View (Typ)) then Discrim := First_Discriminant (Underlying_Record_View (Typ)); elsif Has_Discriminants (Typ) then Discrim := First_Discriminant (Typ); else Discrim := Empty; end if; -- First find the discriminant values in the positional components while Present (Discrim) and then Present (Positional_Expr) loop if Discr_Present (Discrim) then Resolve_Aggr_Expr (Positional_Expr, Discrim); -- Ada 2005 (AI-231) if Ada_Version >= Ada_05 and then Known_Null (Positional_Expr) then Check_Can_Never_Be_Null (Discrim, Positional_Expr); end if; Next (Positional_Expr); end if; if Present (Get_Value (Discrim, Component_Associations (N))) then Error_Msg_NE ("more than one value supplied for discriminant&", N, Discrim); end if; Next_Discriminant (Discrim); end loop; -- Find remaining discriminant values, if any, among named components while Present (Discrim) loop Expr := Get_Value (Discrim, Component_Associations (N), True); if not Discr_Present (Discrim) then if Present (Expr) then Error_Msg_NE ("more than one value supplied for discriminant&", N, Discrim); end if; elsif No (Expr) then Error_Msg_NE ("no value supplied for discriminant &", N, Discrim); Missing_Discriminants := True; else Resolve_Aggr_Expr (Expr, Discrim); end if; Next_Discriminant (Discrim); end loop; if Missing_Discriminants then return; end if; -- At this point and until the beginning of STEP 6, New_Assoc_List -- contains only the discriminants and their values. end Step_3; -- STEP 4: Set the Etype of the record aggregate -- ??? This code is pretty much a copy of Sem_Ch3.Build_Subtype. That -- routine should really be exported in sem_util or some such and used -- in sem_ch3 and here rather than have a copy of the code which is a -- maintenance nightmare. -- ??? Performance WARNING. The current implementation creates a new -- itype for all aggregates whose base type is discriminated. -- This means that for record aggregates nested inside an array -- aggregate we will create a new itype for each record aggregate -- if the array component type has discriminants. For large aggregates -- this may be a problem. What should be done in this case is -- to reuse itypes as much as possible. if Has_Discriminants (Typ) or else (Has_Unknown_Discriminants (Typ) and then Present (Underlying_Record_View (Typ))) then Build_Constrained_Itype : declare Loc : constant Source_Ptr := Sloc (N); Indic : Node_Id; Subtyp_Decl : Node_Id; Def_Id : Entity_Id; C : constant List_Id := New_List; begin New_Assoc := First (New_Assoc_List); while Present (New_Assoc) loop Append (Duplicate_Subexpr (Expression (New_Assoc)), To => C); Next (New_Assoc); end loop; if Has_Unknown_Discriminants (Typ) and then Present (Underlying_Record_View (Typ)) then Indic := Make_Subtype_Indication (Loc, Subtype_Mark => New_Occurrence_Of (Underlying_Record_View (Typ), Loc), Constraint => Make_Index_Or_Discriminant_Constraint (Loc, C)); else Indic := Make_Subtype_Indication (Loc, Subtype_Mark => New_Occurrence_Of (Base_Type (Typ), Loc), Constraint => Make_Index_Or_Discriminant_Constraint (Loc, C)); end if; Def_Id := Create_Itype (Ekind (Typ), N); Subtyp_Decl := Make_Subtype_Declaration (Loc, Defining_Identifier => Def_Id, Subtype_Indication => Indic); Set_Parent (Subtyp_Decl, Parent (N)); -- Itypes must be analyzed with checks off (see itypes.ads) Analyze (Subtyp_Decl, Suppress => All_Checks); Set_Etype (N, Def_Id); Check_Static_Discriminated_Subtype (Def_Id, Expression (First (New_Assoc_List))); end Build_Constrained_Itype; else Set_Etype (N, Typ); end if; -- STEP 5: Get remaining components according to discriminant values Step_5 : declare Record_Def : Node_Id; Parent_Typ : Entity_Id; Root_Typ : Entity_Id; Parent_Typ_List : Elist_Id; Parent_Elmt : Elmt_Id; Errors_Found : Boolean := False; Dnode : Node_Id; begin if Is_Derived_Type (Typ) and then Is_Tagged_Type (Typ) then Parent_Typ_List := New_Elmt_List; -- If this is an extension aggregate, the component list must -- include all components that are not in the given ancestor type. -- Otherwise, the component list must include components of all -- ancestors, starting with the root. if Nkind (N) = N_Extension_Aggregate then Root_Typ := Base_Type (Etype (Ancestor_Part (N))); else Root_Typ := Root_Type (Typ); if Nkind (Parent (Base_Type (Root_Typ))) = N_Private_Type_Declaration then Error_Msg_NE ("type of aggregate has private ancestor&!", N, Root_Typ); Error_Msg_N ("must use extension aggregate!", N); return; end if; Dnode := Declaration_Node (Base_Type (Root_Typ)); -- If we don't get a full declaration, then we have some -- error which will get signalled later so skip this part. -- Otherwise, gather components of root that apply to the -- aggregate type. We use the base type in case there is an -- applicable stored constraint that renames the discriminants -- of the root. if Nkind (Dnode) = N_Full_Type_Declaration then Record_Def := Type_Definition (Dnode); Gather_Components (Base_Type (Typ), Component_List (Record_Def), Governed_By => New_Assoc_List, Into => Components, Report_Errors => Errors_Found); end if; end if; Parent_Typ := Base_Type (Typ); while Parent_Typ /= Root_Typ loop Prepend_Elmt (Parent_Typ, To => Parent_Typ_List); Parent_Typ := Etype (Parent_Typ); if Nkind (Parent (Base_Type (Parent_Typ))) = N_Private_Type_Declaration or else Nkind (Parent (Base_Type (Parent_Typ))) = N_Private_Extension_Declaration then if Nkind (N) /= N_Extension_Aggregate then Error_Msg_NE ("type of aggregate has private ancestor&!", N, Parent_Typ); Error_Msg_N ("must use extension aggregate!", N); return; elsif Parent_Typ /= Root_Typ then Error_Msg_NE ("ancestor part of aggregate must be private type&", Ancestor_Part (N), Parent_Typ); return; end if; end if; end loop; -- Now collect components from all other ancestors, beginning -- with the current type. If the type has unknown discriminants -- use the component list of the Underlying_Record_View, which -- needs to be used for the subsequent expansion of the aggregate -- into assignments. Parent_Elmt := First_Elmt (Parent_Typ_List); while Present (Parent_Elmt) loop Parent_Typ := Node (Parent_Elmt); if Has_Unknown_Discriminants (Parent_Typ) and then Present (Underlying_Record_View (Typ)) then Parent_Typ := Underlying_Record_View (Parent_Typ); end if; Record_Def := Type_Definition (Parent (Base_Type (Parent_Typ))); Gather_Components (Empty, Component_List (Record_Extension_Part (Record_Def)), Governed_By => New_Assoc_List, Into => Components, Report_Errors => Errors_Found); Next_Elmt (Parent_Elmt); end loop; else Record_Def := Type_Definition (Parent (Base_Type (Typ))); if Null_Present (Record_Def) then null; elsif not Has_Unknown_Discriminants (Typ) then Gather_Components (Base_Type (Typ), Component_List (Record_Def), Governed_By => New_Assoc_List, Into => Components, Report_Errors => Errors_Found); else Gather_Components (Base_Type (Underlying_Record_View (Typ)), Component_List (Record_Def), Governed_By => New_Assoc_List, Into => Components, Report_Errors => Errors_Found); end if; end if; if Errors_Found then return; end if; end Step_5; -- STEP 6: Find component Values Component := Empty; Component_Elmt := First_Elmt (Components); -- First scan the remaining positional associations in the aggregate. -- Remember that at this point Positional_Expr contains the current -- positional association if any is left after looking for discriminant -- values in step 3. while Present (Positional_Expr) and then Present (Component_Elmt) loop Component := Node (Component_Elmt); Resolve_Aggr_Expr (Positional_Expr, Component); -- Ada 2005 (AI-231) if Ada_Version >= Ada_05 and then Known_Null (Positional_Expr) then Check_Can_Never_Be_Null (Component, Positional_Expr); end if; if Present (Get_Value (Component, Component_Associations (N))) then Error_Msg_NE ("more than one value supplied for Component &", N, Component); end if; Next (Positional_Expr); Next_Elmt (Component_Elmt); end loop; if Present (Positional_Expr) then Error_Msg_N ("too many components for record aggregate", Positional_Expr); end if; -- Now scan for the named arguments of the aggregate while Present (Component_Elmt) loop Component := Node (Component_Elmt); Expr := Get_Value (Component, Component_Associations (N), True); -- Note: The previous call to Get_Value sets the value of the -- variable Is_Box_Present. -- Ada 2005 (AI-287): Handle components with default initialization. -- Note: This feature was originally added to Ada 2005 for limited -- but it was finally allowed with any type. if Is_Box_Present then Check_Box_Component : declare Ctyp : constant Entity_Id := Etype (Component); begin -- If there is a default expression for the aggregate, copy -- it into a new association. -- If the component has an initialization procedure (IP) we -- pass the component to the expander, which will generate -- the call to such IP. -- If the component has discriminants, their values must -- be taken from their subtype. This is indispensable for -- constraints that are given by the current instance of an -- enclosing type, to allow the expansion of the aggregate -- to replace the reference to the current instance by the -- target object of the aggregate. if Present (Parent (Component)) and then Nkind (Parent (Component)) = N_Component_Declaration and then Present (Expression (Parent (Component))) then Expr := New_Copy_Tree (Expression (Parent (Component)), New_Sloc => Sloc (N)); Add_Association (Component => Component, Expr => Expr, Assoc_List => New_Assoc_List); Set_Has_Self_Reference (N); -- A box-defaulted access component gets the value null. Also -- included are components of private types whose underlying -- type is an access type. In either case set the type of the -- literal, for subsequent use in semantic checks. elsif Present (Underlying_Type (Ctyp)) and then Is_Access_Type (Underlying_Type (Ctyp)) then if not Is_Private_Type (Ctyp) then Expr := Make_Null (Sloc (N)); Set_Etype (Expr, Ctyp); Add_Association (Component => Component, Expr => Expr, Assoc_List => New_Assoc_List); -- If the component's type is private with an access type as -- its underlying type then we have to create an unchecked -- conversion to satisfy type checking. else declare Qual_Null : constant Node_Id := Make_Qualified_Expression (Sloc (N), Subtype_Mark => New_Occurrence_Of (Underlying_Type (Ctyp), Sloc (N)), Expression => Make_Null (Sloc (N))); Convert_Null : constant Node_Id := Unchecked_Convert_To (Ctyp, Qual_Null); begin Analyze_And_Resolve (Convert_Null, Ctyp); Add_Association (Component => Component, Expr => Convert_Null, Assoc_List => New_Assoc_List); end; end if; elsif Has_Non_Null_Base_Init_Proc (Ctyp) or else not Expander_Active then if Is_Record_Type (Ctyp) and then Has_Discriminants (Ctyp) and then not Is_Private_Type (Ctyp) then -- We build a partially initialized aggregate with the -- values of the discriminants and box initialization -- for the rest, if other components are present. -- The type of the aggregate is the known subtype of -- the component. The capture of discriminants must -- be recursive because subcomponents may be contrained -- (transitively) by discriminants of enclosing types. -- For a private type with discriminants, a call to the -- initialization procedure will be generated, and no -- subaggregate is needed. Capture_Discriminants : declare Loc : constant Source_Ptr := Sloc (N); Expr : Node_Id; procedure Add_Discriminant_Values (New_Aggr : Node_Id; Assoc_List : List_Id); -- The constraint to a component may be given by a -- discriminant of the enclosing type, in which case -- we have to retrieve its value, which is part of the -- enclosing aggregate. Assoc_List provides the -- discriminant associations of the current type or -- of some enclosing record. procedure Propagate_Discriminants (Aggr : Node_Id; Assoc_List : List_Id; Comp : Entity_Id); -- Nested components may themselves be discriminated -- types constrained by outer discriminants, whose -- values must be captured before the aggregate is -- expanded into assignments. ----------------------------- -- Add_Discriminant_Values -- ----------------------------- procedure Add_Discriminant_Values (New_Aggr : Node_Id; Assoc_List : List_Id) is Assoc : Node_Id; Discr : Entity_Id; Discr_Elmt : Elmt_Id; Discr_Val : Node_Id; Val : Entity_Id; begin Discr := First_Discriminant (Etype (New_Aggr)); Discr_Elmt := First_Elmt (Discriminant_Constraint (Etype (New_Aggr))); while Present (Discr_Elmt) loop Discr_Val := Node (Discr_Elmt); -- If the constraint is given by a discriminant -- it is a discriminant of an enclosing record, -- and its value has already been placed in the -- association list. if Is_Entity_Name (Discr_Val) and then Ekind (Entity (Discr_Val)) = E_Discriminant then Val := Entity (Discr_Val); Assoc := First (Assoc_List); while Present (Assoc) loop if Present (Entity (First (Choices (Assoc)))) and then Entity (First (Choices (Assoc))) = Val then Discr_Val := Expression (Assoc); exit; end if; Next (Assoc); end loop; end if; Add_Association (Discr, New_Copy_Tree (Discr_Val), Component_Associations (New_Aggr)); -- If the discriminant constraint is a current -- instance, mark the current aggregate so that -- the self-reference can be expanded later. if Nkind (Discr_Val) = N_Attribute_Reference and then Is_Entity_Name (Prefix (Discr_Val)) and then Is_Type (Entity (Prefix (Discr_Val))) and then Etype (N) = Entity (Prefix (Discr_Val)) then Set_Has_Self_Reference (N); end if; Next_Elmt (Discr_Elmt); Next_Discriminant (Discr); end loop; end Add_Discriminant_Values; ------------------------------ -- Propagate_Discriminants -- ------------------------------ procedure Propagate_Discriminants (Aggr : Node_Id; Assoc_List : List_Id; Comp : Entity_Id) is Inner_Comp : Entity_Id; Comp_Type : Entity_Id; Needs_Box : Boolean := False; New_Aggr : Node_Id; begin Inner_Comp := First_Component (Etype (Comp)); while Present (Inner_Comp) loop Comp_Type := Etype (Inner_Comp); if Is_Record_Type (Comp_Type) and then Has_Discriminants (Comp_Type) then New_Aggr := Make_Aggregate (Loc, New_List, New_List); Set_Etype (New_Aggr, Comp_Type); Add_Association (Inner_Comp, New_Aggr, Component_Associations (Aggr)); -- Collect discriminant values and recurse Add_Discriminant_Values (New_Aggr, Assoc_List); Propagate_Discriminants (New_Aggr, Assoc_List, Inner_Comp); else Needs_Box := True; end if; Next_Component (Inner_Comp); end loop; if Needs_Box then Append (Make_Component_Association (Loc, Choices => New_List (Make_Others_Choice (Loc)), Expression => Empty, Box_Present => True), Component_Associations (Aggr)); end if; end Propagate_Discriminants; begin Expr := Make_Aggregate (Loc, New_List, New_List); Set_Etype (Expr, Ctyp); -- If the enclosing type has discriminants, they -- have been collected in the aggregate earlier, and -- they may appear as constraints of subcomponents. -- Similarly if this component has discriminants, they -- might in turn be propagated to their components. if Has_Discriminants (Typ) then Add_Discriminant_Values (Expr, New_Assoc_List); Propagate_Discriminants (Expr, New_Assoc_List, Component); elsif Has_Discriminants (Ctyp) then Add_Discriminant_Values (Expr, Component_Associations (Expr)); Propagate_Discriminants (Expr, Component_Associations (Expr), Component); else declare Comp : Entity_Id; begin -- If the type has additional components, create -- an OTHERS box association for them. Comp := First_Component (Ctyp); while Present (Comp) loop if Ekind (Comp) = E_Component then if not Is_Record_Type (Etype (Comp)) then Append (Make_Component_Association (Loc, Choices => New_List (Make_Others_Choice (Loc)), Expression => Empty, Box_Present => True), Component_Associations (Expr)); end if; exit; end if; Next_Component (Comp); end loop; end; end if; Add_Association (Component => Component, Expr => Expr, Assoc_List => New_Assoc_List); end Capture_Discriminants; else Add_Association (Component => Component, Expr => Empty, Assoc_List => New_Assoc_List, Is_Box_Present => True); end if; -- Otherwise we only need to resolve the expression if the -- component has partially initialized values (required to -- expand the corresponding assignments and run-time checks). elsif Present (Expr) and then Is_Partially_Initialized_Type (Ctyp) then Resolve_Aggr_Expr (Expr, Component); end if; end Check_Box_Component; elsif No (Expr) then -- Ignore hidden components associated with the position of the -- interface tags: these are initialized dynamically. if not Present (Related_Type (Component)) then Error_Msg_NE ("no value supplied for component &!", N, Component); end if; else Resolve_Aggr_Expr (Expr, Component); end if; Next_Elmt (Component_Elmt); end loop; -- STEP 7: check for invalid components + check type in choice list Step_7 : declare Selectr : Node_Id; -- Selector name Typech : Entity_Id; -- Type of first component in choice list begin if Present (Component_Associations (N)) then Assoc := First (Component_Associations (N)); else Assoc := Empty; end if; Verification : while Present (Assoc) loop Selectr := First (Choices (Assoc)); Typech := Empty; if Nkind (Selectr) = N_Others_Choice then -- Ada 2005 (AI-287): others choice may have expression or box if No (Others_Etype) and then not Others_Box then Error_Msg_N ("OTHERS must represent at least one component", Selectr); end if; exit Verification; end if; while Present (Selectr) loop New_Assoc := First (New_Assoc_List); while Present (New_Assoc) loop Component := First (Choices (New_Assoc)); exit when Chars (Selectr) = Chars (Component); Next (New_Assoc); end loop; -- If no association, this is not a legal component of -- of the type in question, except if its association -- is provided with a box. if No (New_Assoc) then if Box_Present (Parent (Selectr)) then -- This may still be a bogus component with a box. Scan -- list of components to verify that a component with -- that name exists. declare C : Entity_Id; begin C := First_Component (Typ); while Present (C) loop if Chars (C) = Chars (Selectr) then -- If the context is an extension aggregate, -- the component must not be inherited from -- the ancestor part of the aggregate. if Nkind (N) /= N_Extension_Aggregate or else Scope (Original_Record_Component (C)) /= Etype (Ancestor_Part (N)) then exit; end if; end if; Next_Component (C); end loop; if No (C) then Error_Msg_Node_2 := Typ; Error_Msg_N ("& is not a component of}", Selectr); end if; end; elsif Chars (Selectr) /= Name_uTag and then Chars (Selectr) /= Name_uParent and then Chars (Selectr) /= Name_uController then if not Has_Discriminants (Typ) then Error_Msg_Node_2 := Typ; Error_Msg_N ("& is not a component of}", Selectr); else Error_Msg_N ("& is not a component of the aggregate subtype", Selectr); end if; Check_Misspelled_Component (Components, Selectr); end if; elsif No (Typech) then Typech := Base_Type (Etype (Component)); elsif Typech /= Base_Type (Etype (Component)) then if not Box_Present (Parent (Selectr)) then Error_Msg_N ("components in choice list must have same type", Selectr); end if; end if; Next (Selectr); end loop; Next (Assoc); end loop Verification; end Step_7; -- STEP 8: replace the original aggregate Step_8 : declare New_Aggregate : constant Node_Id := New_Copy (N); begin Set_Expressions (New_Aggregate, No_List); Set_Etype (New_Aggregate, Etype (N)); Set_Component_Associations (New_Aggregate, New_Assoc_List); Rewrite (N, New_Aggregate); end Step_8; end Resolve_Record_Aggregate; ----------------------------- -- Check_Can_Never_Be_Null -- ----------------------------- procedure Check_Can_Never_Be_Null (Typ : Entity_Id; Expr : Node_Id) is Comp_Typ : Entity_Id; begin pragma Assert (Ada_Version >= Ada_05 and then Present (Expr) and then Known_Null (Expr)); case Ekind (Typ) is when E_Array_Type => Comp_Typ := Component_Type (Typ); when E_Component | E_Discriminant => Comp_Typ := Etype (Typ); when others => return; end case; if Can_Never_Be_Null (Comp_Typ) then -- Here we know we have a constraint error. Note that we do not use -- Apply_Compile_Time_Constraint_Error here to the Expr, which might -- seem the more natural approach. That's because in some cases the -- components are rewritten, and the replacement would be missed. Insert_Action (Compile_Time_Constraint_Error (Expr, "(Ada 2005) null not allowed in null-excluding component?"), Make_Raise_Constraint_Error (Sloc (Expr), Reason => CE_Access_Check_Failed)); -- Set proper type for bogus component (why is this needed???) Set_Etype (Expr, Comp_Typ); Set_Analyzed (Expr); end if; end Check_Can_Never_Be_Null; --------------------- -- Sort_Case_Table -- --------------------- procedure Sort_Case_Table (Case_Table : in out Case_Table_Type) is L : constant Int := Case_Table'First; U : constant Int := Case_Table'Last; K : Int; J : Int; T : Case_Bounds; begin K := L; while K /= U loop T := Case_Table (K + 1); J := K + 1; while J /= L and then Expr_Value (Case_Table (J - 1).Choice_Lo) > Expr_Value (T.Choice_Lo) loop Case_Table (J) := Case_Table (J - 1); J := J - 1; end loop; Case_Table (J) := T; K := K + 1; end loop; end Sort_Case_Table; end Sem_Aggr;