2005-03-17 Thomas Quinot <quinot@adacore.com>
* exp_ch3.adb (Check_Attr): New subprogram.
(Check_Stream_Attribute): Move the code for 13.13.2(9/1) enforcement
into a new Check_Attr subprogram, in order to provide a more
explanatory error message (including the name of the missing attribute).
(Stream_Operation_OK): Renamed from Stream_Operations_OK. This
subprogram determines whether a default implementation exists for a
given stream attribute.
(Make_Predefined_Primitive_Specs, Predefined_Primitive_Bodies):
Determine whether to generate a default implementation for each stream
attribute separately, as this depends on the specific attribute.
* exp_strm.adb (Make_Field_Attribute): For the case of an illegal
limited extension where a stream attribute is missing for a limited
component (which will have been flagged in Exp_Ch3.Sem_Attr), do not
generate a bogus reference to the missing attribute to prevent
cascaded errors. Instead, generate a null statement.
* sem_attr.adb (Check_Stream_Attribute): A stream attribute is
available for a limited type if it has been specified for an ancestor
of the type.
From-SVN: r96666
2005-03-17 Javier Miranda <miranda@adacore.com>
* einfo.adb (First_Private_Entity, Set_First_Private_Entity): Addition
of one barrier to avoid wrong usage of this attribute.
* sem_ch12.adb (Formal_Entity): Fix erroneous usage of the attribute
First_Private_Entity.
* sem_ch7.adb (Install_Visible_Declarations): Add a barrier to protect
the subprogram against wrong usage.
Adapt the code to traverse the entities in the
scope of a record_type because in addition to its usage regarding
packages, this subprogram is also called by Expand_N_Freeze_Entity
to install the visible declarations of the enclosing scope of a
record_type_with_private to establish the proper visibility before
freezing the entity and related subprograms.
From-SVN: r96664
2005-03-17 Robert Dewar <dewar@adacore.com>
* checks.adb (Apply_Array_Size_Check): Completely remove this for GCC
3, since we now expect GCC 3 to do all the work.
From-SVN: r96663
2005-03-17 Javier Miranda <miranda@adacore.com>
* a-tags.ads, a-tags.adb (Get_Expanded_Name): Removed.
(Get_Inheritance_Depth): Removed.
(Set_Inheritance_Depth): Removed.
* rtsfind.ads, exp_disp.ads, exp_disp.adb: Remove support to call the
subprogram Get_Expanded_Name because it is not referenced by the
frontend.
* i-cpp.ads, i-cpp.adb (CPP_Get_Expanded_Name): Removed.
(CPP_Get_Inheritance_Depth): Removed.
(CPP_Set_Inheritance_Depth): Removed.
* tbuild.ads, tbuild.adb (Make_DT_Component): Removed.
From-SVN: r96662
2005-03-17 Eric Botcazou <ebotcazou@adacore.com>
* ada-tree.h: (DECL_RENAMING_GLOBAL_P): New predicate.
(DECL_RENAMED_OBJECT): New accessor macro.
(SET_DECL_RENAMED_OBJECT): New setter macro.
* decl.c (gnat_to_gnu_entity) <E_Variable>: Stabilize the renamed
object in all cases. Attach the renamed object to the VAR_DECL.
(gnat_to_gnu_field): Do not lift the record wrapper if the size of the
field is not prescribed.
* misc.c (gnat_handle_option): Handle -gnatO separately.
(gnat_print_decl) <VAR_DECL>: New case.
Print the DECL_RENAMED_OBJECT node.
* lang.opt: Declare separate -gnatO option.
* trans.c (tree_transform) <N_Identifier>: If the object is a renaming
pointer, replace it with the renamed object.
<N_Validate_Unchecked_Conversion>: Warn for a conversion to a fat
pointer type if the source is not a fat pointer type whose underlying
array has the same non-zero alias set as that of the destination array.
From-SVN: r96660
2005-03-17 Pascal Obry <obry@adacore.com>
* adaint.h, adaint.c (__gnat_waitpid): Moved to expect.c where it is
used.
* expect.c (__gnat_waitpid): Moved here from adaint.c.
Reimplement under Win32 using Win32 API.
(__gnat_kill) [Win32]: Properly close the process handle before leaving
this routine.
From-SVN: r96659
2005-03-18 Alexey Neyman <alex.neyman@auriga.ru>
Paolo Bonzini <gcc.gnu.org>
* calls.c (setjmp_call_p, special_function_p): Update comments
at the head of the functions.
From-SVN: r96656
* cgraph.c (cgraph_remove_node): Avoid loop in code deciding whether
function body should be released; do not proactively release function
bodies in non-unit-at-a-time mode.
From-SVN: r96654
PR tree-optimization/20474
* tree-vect-analyze.c (vect_analyze_pointer_ref_access): Check the
size_type of the relevant pointer. Check for COMPLETE_TYPE_P.
From-SVN: r96632
2005-03-16 Andrew Haley <aph@redhat.com>
* expr.c (expand_java_arrayload): Don't generate a
NullPointerException based on the type of the node.
(build_java_array_length_access): Likewise.
From-SVN: r96611
2005-03-17 Paolo Carlini <pcarlini@suse.de>
PR c++/19966
* cp-tree.h (grok_op_properties): Change return type to void.
* decl.c (grok_op_properties): Return early - don't check the
arity - in case of a static member or an operator that cannot
be non-member; tidy a bit.
From-SVN: r96609
cp:
PR c++/20186
* pt.c (contains_dependent_cast_p): Remove.
(fold_non_dependent_expr): Don't use it.
(value_dependent_expression_p): Use a switch statement.
reference_exprs can be dependent.
From-SVN: r96605
patch to fix extraneous nop at function start, bug from David Mosberger
* config/ia64/ia64.c (issue_nops_and_insn): Check first_insn attribute,
and return without creating new state if before_nops_num is nonzero.
* config/ia64/ia64.md (first_insn): New attribute.
(alloc): Set it to yes.
From-SVN: r96597
2005-03-16 Andreas Tobler <a.tobler@schweiz.ch>
* testsuite/libjava.jni/jni.exp (gcj_jni_invocation_test_one): Add
lgcc_s to the libraries to link for Darwin.
From-SVN: r96579
2005-03-16 Dale Johannesen <dalej@apple.com>
* rtlanal.c (find_first_parameter_load): Rewrite to
return actual first param load or the call if none,
instead of first in block, when not all loads exist.
From-SVN: r96578
* open.c (new_unit): set record length to max_offset rather than
using a hard-coded limit (which was too low).
* write_big_array.f90: New test.
From-SVN: r96573
