Skip to content

R
riscv-gcc-1
Commit 446ebdbb by Arnaud Charlet
Options

[multiple changes] 

2014-06-13  Robert Dewar  <dewar@adacore.com>

	* layout.adb (Layout_Type): Anonymous access types designating
	unconstrained arrays are always thin pointers.
	* sem_ch3.adb (Replace_Type): Anonymous access types designating
	unconstrained arrays are always thin pointers.

2014-06-13  Ed Schonberg  <schonberg@adacore.com>

	* sem_attr.adb (Analyze_Attribute): Use Check_Parameterless_Call
	to resolve the prefix of an attribute that is an entity name
	and may be a parameterless call.  Minor code reorganization.

From-SVN: r211618
parent c86cf714
Showing with 77 additions and 60 deletions
gcc/ada/ChangeLog
2014-06-13 Robert Dewar <dewar@adacore.com>
* layout.adb (Layout_Type): Anonymous access types designating
unconstrained arrays are always thin pointers.
* sem_ch3.adb (Replace_Type): Anonymous access types designating
unconstrained arrays are always thin pointers.
2014-06-13 Ed Schonberg <schonberg@adacore.com>
* sem_attr.adb (Analyze_Attribute): Use Check_Parameterless_Call
to resolve the prefix of an attribute that is an entity name
and may be a parameterless call. Minor code reorganization.
2014-06-13 Robert Dewar <dewar@adacore.com>
* lib.ads, lib.adb, lib-writ.adb, lib-load.adb (Is_Compiler_Unit):
Removed.
* opt.ads (Compiler_Unit): New flag.
......
gcc/ada/layout.adb
......@@ -2462,19 +2462,11 @@ package body Layout is
Set_Size_Info (E, Base_Type (E));
Set_RM_Size (E, RM_Size (Base_Type (E)));
-- Anonymous access types in subprogram specifications are always
-- thin. In the unconstrained case we always use thin pointers for
-- anonymous access types, because otherwise we get into strange
-- conformance problems between two types, one of which can see
-- that something is unconstrained and one of which cannot. The
-- object of an extended return is treated similarly.
elsif Ekind (E) = E_Anonymous_Access_Type
and then (Nkind_In (Associated_Node_For_Itype (E),
N_Function_Specification,
N_Procedure_Specification)
or else Ekind (Scope (E)) = E_Return_Statement)
then
-- Anonymous access types are always thin, because otherwise we get
-- into strange conformance problems between two types, one of which
-- can see that something is unconstrained and one of which cannot.
elsif Ekind (E) = E_Anonymous_Access_Type then
Init_Size (E, System_Address_Size);
-- For other access types, we use either address size, or, if a fat
......
gcc/ada/sem_attr.adb
......@@ -2412,57 +2412,64 @@ package body Sem_Attr is
-- Ada 2005 (AI-345): Ensure that the compiler gives exactly the current
-- output compiling in Ada 95 mode for the case of ambiguous prefixes.
if Ada_Version < Ada_2005
and then Is_Overloaded (P)
and then Aname /= Name_Access
and then Aname /= Name_Address
and then Aname /= Name_Code_Address
and then Aname /= Name_Count
and then Aname /= Name_Result
and then Aname /= Name_Unchecked_Access
then
Error_Attr ("ambiguous prefix for % attribute", P);
-- Is this comment right??? What is "the current output"??? If this
-- is only about Ada 95 mode, why no test for Ada 95 at this point???
elsif Ada_Version >= Ada_2005
and then Is_Overloaded (P)
if Is_Overloaded (P)
and then Aname /= Name_Access
and then Aname /= Name_Address
and then Aname /= Name_Code_Address
and then Aname /= Name_Result
and then Aname /= Name_Unchecked_Access
then
-- Ada 2005 (AI-345): Since protected and task types have primitive
-- entry wrappers, the attributes Count, Caller and AST_Entry require
-- a context check
-- The prefix must be resolvble by itself, without reference to the
-- attribute. One case that requires special handling is a prefix
-- that is a function name, where one interpretation may be a
-- parameterless call. Entry attributes are handled specially below.
if Ada_Version >= Ada_2005
and then Nam_In (Aname, Name_Count, Name_Caller, Name_AST_Entry)
if Is_Entity_Name (P)
and then not Nam_In (Aname, Name_Count, Name_Caller, Name_AST_Entry)
then
declare
Count : Natural := 0;
I : Interp_Index;
It : Interp;
Check_Parameterless_Call (P);
end if;
begin
Get_First_Interp (P, I, It);
while Present (It.Nam) loop
if Comes_From_Source (It.Nam) then
Count := Count + 1;
else
Remove_Interp (I);
end if;
if Ada_Version < Ada_2005 then
if Is_Overloaded (P) then
Get_Next_Interp (I, It);
end loop;
-- Ada 2005 (AI-345): Since protected and task types have
-- primitive entry wrappers, the attributes Count, Caller and
-- AST_Entry require a context check
if Count > 1 then
Error_Attr ("ambiguous prefix for % attribute", P);
if Nam_In (Aname, Name_Count, Name_Caller, Name_AST_Entry) then
declare
Count : Natural := 0;
I : Interp_Index;
It : Interp;
begin
Get_First_Interp (P, I, It);
while Present (It.Nam) loop
if Comes_From_Source (It.Nam) then
Count := Count + 1;
else
Remove_Interp (I);
end if;
Get_Next_Interp (I, It);
end loop;
if Count > 1 then
Error_Attr ("ambiguous prefix for % attribute", P);
else
Set_Is_Overloaded (P, False);
end if;
end;
else
Set_Is_Overloaded (P, False);
Error_Attr ("ambiguous prefix for % attribute", P);
end if;
end;
end if;
else
elsif Is_Overloaded (P) then
Error_Attr ("ambiguous prefix for % attribute", P);
end if;
end if;
......@@ -2470,8 +2477,10 @@ package body Sem_Attr is
-- In SPARK, attributes of private types are only allowed if the full
-- type declaration is visible.
if Is_Entity_Name (P)
and then Present (Entity (P)) -- needed in some cases
-- Note: the check for Present (Entity (P)) defends against some error
-- conditions where the Entity field is not set.
if Is_Entity_Name (P) and then Present (Entity (P))
and then Is_Type (Entity (P))
and then Is_Private_Type (P_Type)
and then not In_Open_Scopes (Scope (P_Type))
......
gcc/ada/sem_ch3.adb
......@@ -13562,21 +13562,23 @@ package body Sem_Ch3 is
-- interface primitives.
or else (Is_Interface (Desig_Typ)
and then not Is_Class_Wide_Type (Desig_Typ))
and then not Is_Class_Wide_Type (Desig_Typ))
then
Acc_Type := New_Copy (Etype (Id));
Set_Etype (Acc_Type, Acc_Type);
Set_Scope (Acc_Type, New_Subp);
-- Compute size of anonymous access type
-- Set size of anonymous access type. Note that anonymous
-- access to Unconstrained always uses thin pointers. This
-- avoids confusion for the case where two types that should
-- conform but end up differning, because in one case we can
-- see the unconstrained designated type, and in the other
-- case we can't see it yet (full type declaration not seen
-- yet), so we default to thin in that case anyway.
if Is_Array_Type (Desig_Typ)
and then not Is_Constrained (Desig_Typ)
then
Init_Size (Acc_Type, 2 * System_Address_Size);
else
Init_Size (Acc_Type, System_Address_Size);
end if;
Init_Size (Acc_Type, System_Address_Size);
-- Set remaining characterstics of anonymous access type
Init_Alignment (Acc_Type);
Set_Directly_Designated_Type (Acc_Type, Derived_Type);
......@@ -13585,6 +13587,7 @@ package body Sem_Ch3 is
Set_Scope (New_Id, New_Subp);
-- Create a reference to it
Build_Itype_Reference (Acc_Type, Parent (Derived_Type));
else
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment