2007-08-16 Robert Dewar <dewar@adacore.com>
* gnat_ugn.texi: Add note on preprocessing (output file not written)
2007-08-16 Thomas Quinot <quinot@adacore.com>
* a-tags.adb: Minor reformatting.
2007-08-16 Bob Duff <duff@adacore.com>
* sem_type.ads, sem_ch4.adb: Minor reformatting.
From-SVN: r127552
2007-08-16 Vincent Celier <celier@adacore.com>
* vms_conv.adb (Process_Argument): Ensure that project related options
are not put in the -cargs section when using GNAT COMPILE.
From-SVN: r127551
2007-08-16 Hristian Kirtchev <kirtchev@adacore.com>
Bob Duff <duff@adacore.com>
Nicolas Setton <setton@adacore.com>
* sem_res.adb (Comes_From_Predefined_Lib_Unit): New.
(Resolve): Alphabetize local variables. Add new variable From_Lib. When
the statement which is being resolved comes from a predefined library
unit, all non-predefined library interpretations are skipped.
(Resolve_Op_Concat): If string concatenation was folded in the parser,
but the "&" is user defined, give an error, because the folding would
be wrong.
* sinfo.ads, sinfo.adb (Is_Folded_In_Parser): New flag to indicate that
the parser has folded a long sequence of concatenations of string
literals.
* trans.c (Handled_Sequence_Of_Statements_to_gnu): Mark "JMPBUF_SAVE"
and "JMP_BUF" variables as artificial.
(N_String_Literal): Do not use alloca for very long string literals. Use
xmalloc/free instead. Otherwise the stack might overflow.
* utils.c (init_gigi_decls): Mark "JMPBUF_T" type as created by the
compiler.
From-SVN: r127550
2007-08-16 Ed Schonberg <schonberg@adacore.com>
* sem_disp.adb (Check_Dispatching_Operation): If the operation
implements an operation inherited from a progenitor interface, verify
that they are subtype-conformant.
From-SVN: r127549
2007-08-16 Gary Dismukes <dismukes@adacore.com>
Javier Miranda <miranda@adacore.com>
* sem_ch3.adb (OK_For_Limited_Init_In_05): Allow calls to 'Input to
initialize a limited object.
(Build_Derived_Record_Type): Add missing check of rules ARM 3.9.4
13/2 and 14/2.
Make sure Has_Complex_Representation is inherited by derived type.
From-SVN: r127547
2007-08-16 Ed Schonberg <schonberg@adacore.com>
* sem_ch12.adb (Copy_Generic_Node): A reference to a child unit of the
generic for an enclosing instance is a global reference, even though
its scope is the enclosing instance.
From-SVN: r127546
2007-08-16 Hristian Kirtchev <kirtchev@adacore.com>
* sem_ch10.adb (Has_With_Clause): If the name of the with clause
currently inspected is a selected component, retrieve the entity of
its selector.
(Install_Limited_Withed_Unit): Call Has_Limited_With_Clause starting
from the immediate ancestor of Main_Unit_Entity.
(Install_Limited_Withed_Unit): Do not install the limited view of
package P if P is reachable through an ancestor chain from package C
and C also has a with clause for P in its body.
(Has_Limited_With_Clause): New routine.
(Has_With_Clause): New routine.
From-SVN: r127545
2007-08-16 Vincent Celier <celier@adacore.com>
* prj.ads (Project_Data): Add comments.
* prj-attr.adb: New attribute Removed_Source_Dirs
* prj-nmsc.adb (Get_Directories): Take into account new attribute
Removed_Source_Dirs
* snames.ads, snames.adb: New standard name Removed_Source_Dirs
From-SVN: r127544
2007-08-16 Nicolas Roche <roche@adacore.com>
* Makefile.in (gnatlib): Propagate FORCE_DEBUG_ADAFLAGS value to sub
makefiles
* Make-lang.in: Update dependencies
From-SVN: r127543
2007-08-16 Vincent Celier <celier@adacore.com>
* make.adb (Collect_Arguments): Call Test_If_Relative_Path with
Including_Non_Switch set to False.
(Gnatmake): For the compiler, call Test_If_Relative_Path with
Including_Non_Switch set to False.
* makeutl.adb, makeutl.ads (Test_If_Relative_Path): New Boolean
parameter Including_Non_Switch, defaulted to True. When
Including_Non_Switch is False, options that are not switches and
appear as relative path are not converted to absolute paths.
From-SVN: r127542
2007-08-16 Javier Miranda <miranda@adacore.com>
* exp_ch9.adb (Build_Protected_Entry): Undo previous change because it
is not really required and can introduce regression with the debugger.
The original problem is fixed with the patch written for checks.adb.
From-SVN: r127540
2007-08-16 Javier Miranda <miranda@adacore.com>
* exp_attr.adb (Attribute_Priority): Add missing support for entries
and entry barriers.
From-SVN: r127539
2007-08-16 Gary Dismukes <dismukes@adacore.com>
Ed Schonberg <schonberg@adacore.com>
Javier Miranda <miranda@adacore.com>
* exp_aggr.adb (Build_Record_Aggr_Code): Extend the test for an
ancestor part given by an aggregate to test for an unchecked conversion,
since this can occur in some cases when the ancestor part is a function
call, and we don't want to fall into the recursive call to this
procedure in that case.
* exp_ch3.adb (Stream_Operation_OK): Revise tests for availability of
stream attributes on limited types to account for user-specified
attributes as well as whether Input (resp. Output) becomes available
due to Read (resp. Write) being available for the type. Change Boolean
variable to the more accurate name
Has_Predefined_Or_Specified_Stream_Attribute. Change convoluted
double-"not" predicate at beginning of return statement to more
understandable form.
* exp_ch5.adb (Expand_N_Extended_Return_Statement): If the extended
return has an associated N_Handled_Sequence_Of_Statements, then wrap it
in a block statement and use that as the first statement of the
expanded return rather than incorrectly using the handled sequence as
the first statement.
* exp_ch6.adb (Expand_N_Subprogram_Declaration): If this is a protected
operation, generate an explicit freeze node for it rather than
generating extra formals, to ensure that gigi has the proper order of
elaboration for anonymous subtypes in the signature of the subprograms.
(Build_In_Place_Formal): Move assertion to beginning of loop.
(Is_Build_In_Place_Function_Call): Allow for an unchecked conversion
applied to a function call (occurs for some cases of 'Input).
(Make_Build_In_Place_Call_In_*): Allow for an unchecked conversion
applied to a function call (occurs for some cases of 'Input).
* exp_strm.adb (Build_Record_Or_Elementary_Input_Function): For Ada
2005, generate an extended return statement enclosing the result object
and 'Read call.
* freeze.adb (Freeze_Record_Type): Extend the current management of
components that are access type with an allocator as default value: add
missing support to the use of qualified expressions of the
allocator (which also cause freezing of the designated type!)
(Freeze_Entity): Call Freeze_Subprogram in the case of a predefined
dispatching operation, since extra formals may be needed by calls to
build-in-place functions (such as stream 'Input).
* sem_ch6.adb (Create_Extra_Formals): Skip creation of the extra
formals for 'Constrained and accessibility level in the case of a
predefined dispatching operation.
* exp_util.adb (Insert_Actions): A protected body is a valid insertion
point, no need to find the parent node.
From-SVN: r127538
2007-08-16 Gary Dismukes <dismukes@adacore.com>
* cstand.adb (Create_Standard): Create an entity for a zero-sized type
associated with Standard_Debug_Renaming_Type, to be used as the type of
the special variables whose names provide debugger encodings for
renaming declarations.
* einfo.ads, einfo.adb (Debug_Renaming_Link): Change to return Node25.
(Set_Debug_Renaming_Link): Change to set Node25.
(Write_Field13_Name): Remove case for E_Enumeration_Literal.
(Write_Field25_Name): Add case for E_Variable to output
"Debug_Renaming_Link".
(Write_Field23_Name): Correct the output string for "Limited_View".
* exp_dbug.adb: Add with and use of Tbuild.
(Debug_Renaming_Declaration): Replace creation of an enumeration type
and literal with creation of a variable of type
Standard_Debug_Renaming_Type whose name encodes both the renamed object
and the entity of the renaming declaration.
(Qualify_Entity_Name): Add the delayed qualification of the entity name
part of the name of a variable that has a Debug_Renaming_Link.
* stand.ads (Standard_Debug_Renaming_Type): New Entity_Id denoting a
special type to be associated with variables that provide debugger
encodings for renaming declarations.
From-SVN: r127537
gcc/
PR middle-end/32897
* reload.c (find_reloads): Check that the memory returned by
find_reloads_toplev was not the result of forcing a constant
to memory.
(find_reloads_toplev): Always use simplify_gen_subreg to get
the subreg of a constant. If the result is also a constant,
but not a legitimate one, force it into the constant pool
and reload its address.
gcc/testsuite/
* gcc.dg/torture/pr32897.c: New test.
From-SVN: r127536
* tree-ssa-alias.c (compute_memory_partitions): Use
alias_bitmap_obstack to allocate bitmaps.
(reset_alias_info): Factor out of init_alias_info.
Mark all name tags not associated to an SSA name for renaming.
(init_alias_info): Call it.
(create_name_tags): Tidy. Add comments.
(dump_points_to_info_for): Do not call get_mem_sym_stats_for.
testsuite/ChangeLog
gcc.dg/tree-ssa/20070815.c: New test.
From-SVN: r127520
PR middle-end/33074
* emit-rtl.c (try_split): Use INSN_LIST instead of EXPR_LIST for
REG_LIBCALL note.
* gfortran.dg/pr33074.f90: New test.
From-SVN: r127511
/cp
2007-08-15 Paolo Carlini <pcarlini@suse.de>
PR c++/33035
* pt.c (push_template_decl_real): Depending on TYPE_P
use either TYPE_CONTEXT or DECL_CONTEXT.
/testsuite
2007-08-15 Paolo Carlini <pcarlini@suse.de>
PR c++/33035
* g++.dg/template/crash68.C: New.
From-SVN: r127508
2007-08-14 Daniel Berlin <dberlin@dberlin.org>
* tree-pass.h (PROP_pta): Removed.
(TODO_rebuild_alias): New.
(pass_may_alias): Removed.
* tree-ssa-ccp.c (execute_fold_all_builtins): Only rebuild
aliasing if we changed something.
* tree-ssa-alias.c (compute_may_aliases): Make non-static. Update
SSA internally.
(pass_may_alias): Removed.
(create_structure_vars): Return TODO_rebuild_alias.
* tree-ssa-pre.c (do_pre): Return TODO_rebuild_alias.
* tree-sra.c (tree_sra): Only rebuild aliasing if something
changed.
(tree_sra_early): We never affect aliasing right now.
* tree-flow.h (compute_may_aliases): New prototype.
* passes.c: Remove pass_may_alias from the passes.
(execute_function_todo): Support TODO_rebuild_alias.
From-SVN: r127491
