summaryrefslogtreecommitdiff
path: root/gcc/ada/exp_ch5.adb
diff options
context:
space:
mode:
authorcharlet <charlet@138bc75d-0d04-0410-961f-82ee72b054a4>2004-01-05 15:20:47 +0000
committercharlet <charlet@138bc75d-0d04-0410-961f-82ee72b054a4>2004-01-05 15:20:47 +0000
commit5c61a0ffedb24ab65e306e3de2770d83fa6684cf (patch)
tree4e026dfac036b969e00a952b33c9f32df5724609 /gcc/ada/exp_ch5.adb
parent9f5f9308c29ded02d1126efe3325bb76cb38dd00 (diff)
downloadgcc-5c61a0ffedb24ab65e306e3de2770d83fa6684cf.tar.gz
2004-01-05 Robert Dewar <dewar@gnat.com>
* 1ssecsta.ads: Default_Secondary_Stack is not a constant since it may be modified by the binder generated main program if the -D switch is used. * 4onumaux.ads, 4znumaux.ads: Add Pure_Function pragmas for all imported functions (since now we expect this to be done for imported functions) * 5vtaprop.adb: Add several ??? for sections requiring more comments Minor reformatting throughout * 5zinit.adb: Minor reformatting Add 2004 to copyright date Minor changes to avoid -gnatwa warnings Correct some instances of using OR instead of OR ELSE (noted while doing reformatting) * sprint.adb: Minor updates to avoid -gnatwa warnings * s-secsta.ads, s-secsta.adb: (SS_Get_Max): New function to obtain high water mark for ss stack Default_Secondary_Stack is not a constant since it may be modified by the binder generated main program if the -D switch is used. * switch-b.adb: New -Dnnn switch for binder * switch-c.adb: Make -gnatg imply all warnings currently in -gnatwa * vms_conv.adb: Minor reformatting Add 2004 to copyright notice Add 2004 to printed copyright notice * 3vexpect.adb, 4zsytaco.adb, 3wsocthi.adb, 3zsocthi.adb, 3zsocthi.adb, 56taprop.adb, 56tpopsp.adb, 5amastop.adb, 5aml-tgt.adb, 5ataprop.adb, 5ataprop.adb, 5atpopsp.adb, 5ftaprop.adb, 5ginterr.adb, 5gmastop.adb, 5gml-tgt.adb, 5gtaprop.adb, 5hml-tgt.adb, 5hml-tgt.adb, 5hml-tgt.adb, 5htaprop.adb, 5htraceb.adb, 5itaprop.adb, 5lml-tgt.adb, 5sml-tgt.adb, 5staprop.adb, 5staprop.adb, 5stpopsp.adb, 5vasthan.adb, 5vinmaop.adb, 5vinterr.adb, 5vtaprop.adb, 5vml-tgt.adb, 5vtaprop.adb, 5wosprim.adb, 5wtaprop.adb, 5zinterr.adb, 5zintman.adb, 5zml-tgt.adb, 5ztaprop.adb, 6vcpp.adb, 6vcstrea.adb, 7staprop.adb, 7stpopsp.adb, vxaddr2line.adb, vxaddr2line.adb, xref_lib.adb, xr_tabls.adb, xr_tabls.ads, s-tasdeb.adb, s-tasdeb.adb, sem_res.ads, sem_util.adb, sem_util.adb, sem_util.ads, s-interr.adb, checks.adb, clean.adb, cstand.adb, einfo.ads, einfo.adb, exp_aggr.adb, exp_ch11.adb, exp_ch3.adb, exp_ch4.adb, exp_ch5.adb, exp_ch7.adb, exp_ch9.adb, prj-nmsc.adb, prj-pp.adb, prj-util.adb, sem_attr.adb, sem_ch10.adb, sem_ch12.adb, sem_ch4.adb, g-dirope.adb, g-dirope.ads, gnatlbr.adb, i-cstrea.adb, inline.adb, lib-xref.adb, sem_ch5.adb, sem_ch7.adb, sem_ch8.adb: Minor reformatting and code clean ups. Minor changes to prevent -gnatwa warnings * ali.adb: Minor reformatting and cleanup of code Acquire new SS indication of secondary stack use from ali files * a-numaux.ads: Add Pure_Function pragmas for all imported functions (since now we expect this to be done for imported functions) * bindgen.adb: Generate call to modify default secondary stack size if -Dnnn switch given * bindusg.adb: Add line for new -D switch * exp_aggr.adb (Type_May_Have_Bit_Aligned_Components): More appropriate replacement name for Type_May_Have_Non_Bit_Aligned_Components! Add circuitry for both records and arrays to avoid gigi processing if the type involved has non-bit-aligned components * exp_ch5.adb (Expand_Assign_Array): Avoid assumption that N_String_Literal node always references an E_String_Literal_Subtype entity. This may not be true in the future. (Possible_Bit_Aligned_Component): Move processing of Component_May_Be_Bit_Aligned from exp_ch5 to exp_util * exp_ch6.adb (Expand_Thread_Body): Pick up Default_Secondary_Stack_Size as variable so that we get value modified by possible -Dnnn binder parameter. * exp_util.adb (Component_May_Be_Bit_Aligned): New function. (Type_May_Have_Bit_Aligned_Components): New function. * exp_util.ads (Component_May_Be_Bit_Aligned): New function. (Type_May_Have_Bit_Aligned_Components): New function. * fe.h: (Set_Identifier_Casing): Fix prototype. Add declaration for Sem_Elim.Eliminate_Error_Msg. Minor reformatting. * freeze.adb (Freeze_Entity): Add RM reference to error message about importing constant atomic/volatile objects. (Freeze_Subprogram): Reset Is_Pure indication for imported subprogram unless explicit Pure_Function pragma given, to avoid insidious bug of call to non-pure imported function getting eliminated. * gnat1drv.adb, gnatbind.adb, gnatchop.adb, gnatfind.adb, gnatls.adb, gnatlink.adb, gnatmem.adb, gnatname.adb, gnatsym.adb, gnatxref.adb, gprcmd.adb, gprep.adb, make.adb: Minor reformatting Add 2004 to printed copyright notice * lib-writ.ads, lib-writ.adb: Put new SS flag in ali file if secondary stack used. * Makefile.rtl: Add entry for g-sestin.o g-sestin.ads: New file. * mdll.adb: Minor changes to avoid -gnatwa warnings * mlib-tgt.adb: Minor reformatting * opt.ads: New parameter Default_Secondary_Stack_Size (GNATBIND) New switch Sec_Stack_Used (GNAT, GNATBIND) Make Default_Secondary_Stack_Size a variable instead of a constant, so that it can be modified by the new -Dnnn bind switch. * rtsfind.adb (Load_Fail): Give full error message in configurable run-time mode if all_errors mode is set. This was not done in the case of a file not found, which was an oversight. Note if secondary stack unit is used by compiler. * sem_elab.adb (Check_A_Call): Rewrite to avoid trying to put ineffective elaborate all pragmas on non-visible packages (this happened when a renamed subprogram was called). Now the elaborate all always goes on the package containing the renaming rather than the one containing the renamed subprogram. * sem_elim.ads, sem_elim.adb (Eliminate_Error_Msg): New procedure (Process_Eliminate_Pragma): Add parameter to capture pragma location. * sem_eval.adb (Eval_String_Literal): Do not assume that string literal has an Etype that references an E_String_Literal. (Eval_String_Literal): Avoid assumption that N_String_Literal node always references an E_String_Literal_Subtype entity. This may not be true in the future. * sem_prag.adb (Process_Eliminate_Pragma): Add parameter to capture pragma location. * sem_res.adb (Resolve): Specialize msg for function name used in proc call. 2004-01-05 Ed Falis <falis@gnat.com> * g-debuti.adb: Replaced direct boolean operator with short-circuit form. 2004-01-05 Vincent Celier <celier@gnat.com> * bld.adb: Minor comment updates (Process_Declarative_Items): Correct incorrect name (Index_Name instead of Item_Name). * make.adb (Gnatmake): Special process for files to compile/check when -B is specified. Fail when there are only foreign mains in attribute Main of the project file and -B is not specified. Do not skip bind/link steps when -B is specified. * makeusg.adb: Document new switch -B * opt.ads (Build_Bind_And_Link_Full_Project): New Boolean flag * switch-m.adb: (Scan_Make_Switches): Process -B switch * vms_data.ads: Add new GNAT PRETTY qualifier /FORM_FEED_AFTER_PRAGMA_PAGE for switch -ff 2004-01-05 Richard Kenner <kenner@vlsi1.ultra.nyu.edu> * trans.c (tree_transform, case N_Free_Statement): Handle thin pointer case. * misc.c (gnat_printable_name): If VERBOSITY is 2, call Set_Identifier_Casing. * decl.c (gnat_to_gnu_entity, E_Function): Give error if return type has size that overflows. 2004-01-05 Gary Dismukes <dismukes@gnat.com> * exp_ch4.adb (Expand_Array_Comparison): Add Boolean constant to avoid -gnatwa warning on static condition. 2004-01-05 Doug Rupp <rupp@gnat.com> * link.c: (shared_libgnat_default) [VMS]: Change to STATIC. 2004-01-05 Arnaud Charlet <charlet@act-europe.fr> * Makefile.in: Install ali files using INSTALL_DATA_DATE to preserve all attributes, including read-only attribute. 2004-01-05 Pascal Obry <obry@gnat.com> * bindgen.adb (Gen_Object_Files_Options): Generate the new shared library naming scheme. * mlib-prj.adb (Build_Library): Generate different names for the static or dynamic version of the GNAT runtime. This is needed to support the new shared library naming scheme. (Process_Binder_File): Add detection of shared library in binder file based on the new naming scheme. * gnatlink.adb (Process_Binder_File): Properly detect the new naming scheme for the shared runtime libraries. * Makefile.in: (LIBRARY_VERSION) [VMS]: Convert all . to _ to conform to new naming scheme. (install-gnatlib): Do not create symlinks for shared libraries. (gnatlib-shared-default): Idem. (gnatlib-shared-dual-win32): New target. Not used for now as the auto-import feature does not support arrays/records. (gnatlib-shared-win32): Do not create copy for the shared libraries. (gnatlib-shared-vms): Fix shared runtime libraries names. * osint.ads, osint.adb (Shared_Lib): New routine, returns the target dependent runtime shared library name. 2004-01-05 Vasiliy Fofanov <fofanov@act-europe.fr> * osint.adb (Read_Library_Info): Remove bogus check if ALI is older than the object. 2004-01-05 Ed Schonberg <schonberg@gnat.com> * sem_ch4.adb (Analyze_Allocator): Check restriction on dynamic protected objects when allocator has a subtype indication, not a qualified expression. Note that qualified expressions may have to be checked when limited aggregates are implemented. * sem_prag.adb (Analyze_Pragma, case Import): If enclosing package is pure, emit warning. (Analyze_Pragma, case Pure_Function): If enclosing package is pure and subprogram is imported, remove warning. 2004-01-05 Geert Bosch <bosch@gnat.com> * s-poosiz.adb: Update copyright notice. (Allocate): Use Task_Lock to protect against concurrent access. (Deallocate): Likewise. 2004-01-05 Joel Brobecker <brobecker@gnat.com> * s-stalib.adb (Elab_Final_Code): Add missing year in date inside ??? comment. git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@75432 138bc75d-0d04-0410-961f-82ee72b054a4
Diffstat (limited to 'gcc/ada/exp_ch5.adb')
-rw-r--r--gcc/ada/exp_ch5.adb88
1 files changed, 20 insertions, 68 deletions
diff --git a/gcc/ada/exp_ch5.adb b/gcc/ada/exp_ch5.adb
index a257b274ce0..7c08b2ab963 100644
--- a/gcc/ada/exp_ch5.adb
+++ b/gcc/ada/exp_ch5.adb
@@ -48,6 +48,7 @@ with Sem_Res; use Sem_Res;
with Sem_Util; use Sem_Util;
with Snames; use Snames;
with Stand; use Stand;
+with Stringt; use Stringt;
with Tbuild; use Tbuild;
with Ttypes; use Ttypes;
with Uintp; use Uintp;
@@ -75,8 +76,7 @@ package body Exp_Ch5 is
L_Type : Entity_Id;
R_Type : Entity_Id;
Ndim : Pos;
- Rev : Boolean)
- return Node_Id;
+ Rev : Boolean) return Node_Id;
-- N is an assignment statement which assigns an array value. This routine
-- expands the assignment into a loop (or nested loops for the case of a
-- multi-dimensional array) to do the assignment component by component.
@@ -104,32 +104,11 @@ package body Exp_Ch5 is
function Possible_Bit_Aligned_Component (N : Node_Id) return Boolean;
-- This function is used in processing the assignment of a record or
- -- indexed component. The back end can handle such assignments fine
- -- if the objects involved are small (64-bits or less) records or
- -- scalar items (including bit-packed arrays represented with modular
- -- types) or are both aligned on a byte boundary (starting on a byte
- -- boundary, and occupying an integral number of bytes).
- --
- -- However, problems arise for records larger than 64 bits, or for
- -- arrays (other than bit-packed arrays represented with a modular
- -- type) if the component starts on a non-byte boundary, or does
- -- not occupy an integral number of bytes (i.e. there are some bits
- -- possibly shared with fields at the start or beginning of the
- -- component). The back end cannot handle loading and storing such
- -- components in a single operation.
- --
- -- This function is used to detect the troublesome situation. it is
- -- conservative in the sense that it produces True unless it knows
- -- for sure that the component is safe (as outlined in the first
- -- paragraph above). The code generation for record and array
- -- assignment checks for trouble using this function, and if so
- -- the assignment is generated component-wise, which the back end
- -- is required to handle correctly.
- --
- -- Note that in GNAT 3, the back end will reject such components
- -- anyway, so the hard work in checking for this case is wasted
- -- in GNAT 3, but it's harmless, so it is easier to do it in
- -- all cases, rather than conditionalize it in GNAT 5 or beyond.
+ -- indexed component. The argument N is either the left hand or right
+ -- hand side of an assignment, and this function determines if there
+ -- is a record component reference where the record may be bit aligned
+ -- in a manner that causes trouble for the back end (see description
+ -- of Sem_Util.Component_May_Be_Bit_Aligned for further details).
------------------------------
-- Change_Of_Representation --
@@ -508,9 +487,12 @@ package body Exp_Ch5 is
-- statement, a length check has already been emitted to verify that
-- the range of the left-hand side is empty.
+ -- Note that this code is not executed if we had an assignment of
+ -- a string literal to a non-bit aligned component of a record, a
+ -- case which cannot be handled by the backend
+
elsif Nkind (Rhs) = N_String_Literal then
- if Ekind (R_Type) = E_String_Literal_Subtype
- and then String_Literal_Length (R_Type) = 0
+ if String_Length (Strval (Rhs)) = 0
and then Is_Bit_Packed_Array (L_Type)
then
Rewrite (N, Make_Null_Statement (Loc));
@@ -731,8 +713,8 @@ package body Exp_Ch5 is
elsif Restrictions (No_Implicit_Conditionals) then
declare
- T : constant Entity_Id := Make_Defining_Identifier (Loc,
- Chars => Name_T);
+ T : constant Entity_Id :=
+ Make_Defining_Identifier (Loc, Chars => Name_T);
begin
Rewrite (N,
@@ -881,8 +863,7 @@ package body Exp_Ch5 is
L_Type : Entity_Id;
R_Type : Entity_Id;
Ndim : Pos;
- Rev : Boolean)
- return Node_Id
+ Rev : Boolean) return Node_Id
is
Loc : constant Source_Ptr := Sloc (N);
@@ -2244,8 +2225,8 @@ package body Exp_Ch5 is
and then List_Length (Else_Statements (N)) = 1
then
declare
- Then_Stm : Node_Id := First (Then_Statements (N));
- Else_Stm : Node_Id := First (Else_Statements (N));
+ Then_Stm : constant Node_Id := First (Then_Statements (N));
+ Else_Stm : constant Node_Id := First (Else_Statements (N));
begin
if Nkind (Then_Stm) = N_Return_Statement
@@ -3277,39 +3258,10 @@ package body Exp_Ch5 is
-- unless it is forced to do so. In the clear means we need
-- only the recursive test on the prefix.
- if No (Component_Clause (Comp)) then
- return Possible_Bit_Aligned_Component (P);
-
- -- Otherwise we have a component clause, which means that
- -- the Esize and Normalized_First_Bit fields are set and
- -- contain static values known at compile time.
-
+ if Component_May_Be_Bit_Aligned (Comp) then
+ return True;
else
- -- If we know that we have a small (64 bits or less) record
- -- or bit-packed array, then everything is fine, since the
- -- back end can handle these cases correctly.
-
- if Esize (Comp) <= 64
- and then (Is_Record_Type (Etype (Comp))
- or else
- Is_Bit_Packed_Array (Etype (Comp)))
- then
- return False;
-
- -- Otherwise if the component is not byte aligned, we
- -- know we have the nasty unaligned case.
-
- elsif Normalized_First_Bit (Comp) /= Uint_0
- or else Esize (Comp) mod System_Storage_Unit /= Uint_0
- then
- return True;
-
- -- If we are large and byte aligned, then OK at this level
- -- but we still need to test our prefix recursively.
-
- else
- return Possible_Bit_Aligned_Component (P);
- end if;
+ return Possible_Bit_Aligned_Component (P);
end if;
end;