gcc.git
12 years agoMakefile.in: Handle e500v2-wrs-vxworksae like powerpc-wrs-vxworksae.
Tristan Gingold [Tue, 6 Sep 2011 12:16:01 +0000 (12:16 +0000)]
Makefile.in: Handle e500v2-wrs-vxworksae like powerpc-wrs-vxworksae.

2011-09-06  Tristan Gingold  <gingold@adacore.com>

* gcc-interface/Makefile.in: Handle e500v2-wrs-vxworksae like
powerpc-wrs-vxworksae.

From-SVN: r178592

12 years ago[multiple changes]
Arnaud Charlet [Tue, 6 Sep 2011 12:14:30 +0000 (14:14 +0200)]
[multiple changes]

2011-09-06  Thomas Quinot  <quinot@adacore.com>

* s-os_lib.ads (Spawn): Minor documentation clarification,
Success is True for a zero exit status.

2011-09-06  Ed Schonberg  <schonberg@adacore.com>

* sem_ch5.adb: Add message for common iterator error.

2011-09-06  Ed Schonberg  <schonberg@adacore.com>

* exp_ch3.adb (Build_Initialization_Call): If the target is a
selected component discriminated by a current instance, replace
the constraint with a reference to the target object, regardless
of whether the context is an init_proc.

2011-09-06  Robert Dewar  <dewar@adacore.com>

* exp_attr.adb: Descriptor_Size is never static.

2011-09-06  Robert Dewar  <dewar@adacore.com>

* gnat_ugn.texi: Add documentation for LSLOC metric in gnatmetric

2011-09-06  Hristian Kirtchev  <kirtchev@adacore.com>

* gnat_rm.texi: Clarify that attribute Descriptor_Size is
non-static.

2011-09-06  Ed Schonberg  <schonberg@adacore.com>

* sem_res.adb (Resolve): An expression that is the body of an
expression function does not freeze.

2011-09-06  Matthew Heaney  <heaney@adacore.com>

* a-csquin.ads, a-cusyqu.adb, a-cbprqu.adb, a-cbsyqu.adb,
a-cuprqu.adb: Changed copyright notice to indicate current
year only.

2011-09-06  Vincent Celier  <celier@adacore.com>

* prj.adb: Minor spelling error fix in comment
* sem_res.adb: Minor reformatting

2011-09-06  Pascal Obry  <obry@adacore.com>

* sysdep.c (winflush_nt): Removed as not needed anymore.
(winflush_95): Likewise.
(winflush_init): Likewise.
(winflush_function): Likewise.
(getc_immediate_common): Remove call to winflush_function.

From-SVN: r178591

12 years agoexp_attr.adb (Expand_N_Attribute_Reference): Rewrite the processing for Descriptor_Size.
Hristian Kirtchev [Tue, 6 Sep 2011 12:03:30 +0000 (12:03 +0000)]
exp_attr.adb (Expand_N_Attribute_Reference): Rewrite the processing for Descriptor_Size.

2011-09-06  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_attr.adb (Expand_N_Attribute_Reference): Rewrite the
processing for Descriptor_Size.
* gnat_rm.texi: Rephrase the wording for attribute Descriptor_Size
to account for its broader usage.
* sem_attr.adb (Analyze_Attribute): Change the error detection
circuitry for Descriptor_Size as the attribute is now applicable
to all types.

From-SVN: r178590

12 years agosem_attr.adb, [...]: Minor reformatting.
Robert Dewar [Tue, 6 Sep 2011 11:59:12 +0000 (11:59 +0000)]
sem_attr.adb, [...]: Minor reformatting.

2011-09-06  Robert Dewar  <dewar@adacore.com>

* sem_attr.adb, prj-nmsc.adb, exp_aggr.adb: Minor reformatting.

From-SVN: r178589

12 years agoarm.c (arm_preferred_simd_mode): Check TARGET_NEON_VECTORIZE_DOUBLE instead of TARGET...
Ira Rosen [Tue, 6 Sep 2011 11:29:53 +0000 (11:29 +0000)]
arm.c (arm_preferred_simd_mode): Check TARGET_NEON_VECTORIZE_DOUBLE instead of TARGET_NEON_VECTORIZE_QUAD.

* config/arm/arm.c (arm_preferred_simd_mode): Check
TARGET_NEON_VECTORIZE_DOUBLE instead of
TARGET_NEON_VECTORIZE_QUAD.
(arm_autovectorize_vector_sizes): Likewise.
* config/arm/arm.opt (mvectorize-with-neon-quad): Make inverse
mask of mvectorize-with-neon-double.  Add RejectNegative.
(mvectorize-with-neon-double): New.

From-SVN: r178588

12 years ago[multiple changes]
Arnaud Charlet [Tue, 6 Sep 2011 11:03:44 +0000 (13:03 +0200)]
[multiple changes]

2011-09-06  Ed Schonberg  <schonberg@adacore.com>

* lib-xref.adb (OK_To_Set_Referenced): A reference to a formal
in a parameter association must not set the Referenced flag on
the formal.
* prj-nmsc.adb (Check_File_Naming_Schemes): Remove useless formal

2011-09-06  Hristian Kirtchev  <kirtchev@adacore.com>

* gnat_rm.texi: Add a section on attribute Descriptor_Size

From-SVN: r178587

12 years agoMakefile.in (common-tools, [...]): Speed up by using -j0.
Arnaud Charlet [Tue, 6 Sep 2011 10:55:20 +0000 (10:55 +0000)]
Makefile.in (common-tools, [...]): Speed up by using -j0.

* gcc-interface/Makefile.in (common-tools, gnatmake-re,
gnatlink-re): Speed up by using -j0.

From-SVN: r178586

12 years ago[multiple changes]
Arnaud Charlet [Tue, 6 Sep 2011 10:53:58 +0000 (12:53 +0200)]
[multiple changes]

2011-09-06  Robert Dewar  <dewar@adacore.com>

* s-tpopsp-vxworks.adb, prj-nmsc.adb: Minor reformatting.

2011-09-06  Hristian Kirtchev  <kirtchev@adacore.com>

* gcc-interface/trans.c (Attribute_to_gnu): New case for
attribute Descriptor_Size.
* exp_attr.adb (Expand_N_Attribute_Reference): Add processing
for attribute Descriptor_Size.
* exp_ch7.adb (Double_Size_Of): Removed.
(Make_Finalize_Address_Stmts): Remove the code which generates
an expression to calculate the dope vector of an unconstrained
array. Instead use attribute Descriptor_Size and leave the
calculation to the back end.
(Nearest_Multiple_Rounded_Up): Removed.
(Size_Of): Removed.
* sem_attr.adb (Analyze_Attribute): Add processing for attribute
Descriptor_Size. Currently the attribute is applicable only
to unconstrained arrays.
(Eval_Attribute): Add processing for
attribute Descriptor_Size.
* snames.ads-tmpl: Add a predefined name and an Attribute_Id
for Descriptor_Size.

2011-09-06  Ed Schonberg  <schonberg@adacore.com>

* exp_aggr.adb: Remove useless formal.

From-SVN: r178585

12 years agotree-ssa-sccvn.c (visit_use): CSE stmt pieces and simplify matching.
Richard Guenther [Tue, 6 Sep 2011 10:45:09 +0000 (10:45 +0000)]
tree-ssa-sccvn.c (visit_use): CSE stmt pieces and simplify matching.

2011-09-06  Richard Guenther  <rguenther@suse.de>

* tree-ssa-sccvn.c (visit_use): CSE stmt pieces and simplify
matching.

From-SVN: r178584

12 years ago[multiple changes]
Arnaud Charlet [Tue, 6 Sep 2011 10:43:17 +0000 (12:43 +0200)]
[multiple changes]

2011-09-06  Yannick Moy  <moy@adacore.com>

* sem_ch13.adb (Analyze_Aspect_Specifications): Call
Set_Corresponding_Aspect when creating pragma from aspect.
(Add_Predicates): Use new field Corresponding_Aspect.
* sem_prag.adb (Analyze_Pragma): Make Pname hold source aspect
name when present, for the purpose of issuing error messages;
remove local procedure Error_Pragma_Arg_Alternate_Name.
* sinfo.adb, sinfo.ads (Corresponding_Aspect): New field in
N_Pragma node.
(From_Dynamic_Predicate, From_Static_Predicate): Remove fields from
N_Pragma node.

2011-09-06  Robert Dewar  <dewar@adacore.com>

* checks.adb, s-except.ads, g-socket.adb: Minor reformatting.

2011-09-06  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_ch6.adb (Build_Heap_Allocator): Add new
local variable Desig_Typ. Code and comment reformatting. Add
machinery to ensure that the allocation uses a fat pointer when
the type of the return object is a constrained array and the
function return type is an unconstrained array.

2011-09-06  Vincent Celier  <celier@adacore.com>

* make.adb, prj-part.adb, prj-nmsc.adb: Remove unused formal
parameters in subprograms.

2011-09-06  Arnaud Charlet  <charlet@adacore.com>

* s-taprop-mingw.adb (Finalize_TCB): Fix typo.

2011-09-06  Thomas Quinot  <quinot@adacore.com>

* s-taprop-vxworks.adb, s-tpoaal.adb, s-tpopsp-vxworks.adb
(System.Tasking.Primitive_Operations.Specific.Delete): Remove
subprogram.
(System.Tasking.Primitive_Operations.Specific.Set): If argument
is null, destroy task specific data, to make API consistent with
other platforms, and thus compatible with the shared version
of s-tpoaal.adb.
(System.Tasking.Primitive_Operations.ATCB_Allocation.Free_ATCB):
Document the above assumption.

From-SVN: r178583

12 years ago[multiple changes]
Arnaud Charlet [Tue, 6 Sep 2011 10:35:25 +0000 (12:35 +0200)]
[multiple changes]

2011-09-06  Ed Schonberg  <schonberg@adacore.com>

* exp_ch6.adb (Expand_Inlined_Call): Fix use of uninitialized
variable for type of return value when return type is
unconstrained and context is an assignment.

2011-09-06  Ed Schonberg  <schonberg@adacore.com>

* sem_ch8.adb (Check_Class_Wide_Actual): Do not generate body of
class-wide operation if expansion is not enabled.

2011-09-06  Eric Botcazou  <ebotcazou@adacore.com>

* checks.adb (Apply_Scalar_Range_Check): Deal with access
type prefix.

2011-09-06  Yannick Moy  <moy@adacore.com>

* sem_ch13.adb (Analyze_Aspect_Specifications, case
Aspect_Invariant): Do not issue error at this point on illegal
pragma placement, as this is checked later on when analyzing
the corresponding pragma.
* sem_prag.adb (Error_Pragma_Arg_Alternate_Name): New procedure
similar to Error_Pragma_Arg, except the source name of the
aspect/pragma to use in warnings may be equal to parameter
Alt_Name (Analyze_Pragma, case Pragma_Invariant): refine error
message to distinguish source name of pragma/aspect, and whether
the illegality resides in the type being public, or being private
without a public declaration

2011-09-06  Thomas Quinot  <quinot@adacore.com>

* g-socket.adb (Check_For_Fd_Set): On Windows, no need for bitmap
size check (fd_set is implemented differently on that platform).

2011-09-06  Thomas Quinot  <quinot@adacore.com>

* s-taprop-vxworks.adb, s-taprop-tru64.adb, s-taprop-vms.adb,
s-tpoaal.adb, s-taprop-mingw.adb, s-taprop-linux.adb,
s-taprop-solaris.adb, s-taprop-irix.adb, s-taprop.ads,
s-taprop-hpux-dce.adb, s-taprop-dummy.adb, s-taprop-posix.adb
(ATCB_Allocation): New subpackage of
System.Tasking.Primitive_Operations, shared across all targets
with full tasking runtime.
(ATCB_Allocation.New_ATCB): Moved there (from target specific
s-taprop bodies).
(ATCB_Allocation.Free_ATCB): New subprogram. Deallocate an ATCB,
taking care of establishing a local temporary ATCB if the one
being deallocated is Self, to avoid a reference to the freed
ATCB in Abort_Undefer.

2011-09-06  Thomas Quinot  <quinot@adacore.com>

* s-tassta.adb, s-taskin.ads (Free_Task): If the task is not
terminated, mark it for deallocation upon termination.
(Terminate_Task): Call Free_Task again if the task is marked
for automatic deallocation upon termination.

From-SVN: r178582

12 years agore PR libstdc++/50257 ([C++0x] unordered_map slow initialization due to huge __prime_...
Paolo Carlini [Tue, 6 Sep 2011 10:22:21 +0000 (10:22 +0000)]
re PR libstdc++/50257 ([C++0x] unordered_map slow initialization due to huge __prime_list)

2011-09-06  Paolo Carlini  <paolo.carlini@oracle.com>

PR libstdc++/50257
* include/bits/hashtable_policy.h (_Prime_rehash_policy::
    _M_next_bkt): Optimize for small argument.

From-SVN: r178581

12 years agorecog.c (asm_labels_ok): New function.
Tom de Vries [Tue, 6 Sep 2011 10:02:41 +0000 (10:02 +0000)]
recog.c (asm_labels_ok): New function.

2011-09-06  Tom de Vries  <tom@codesourcery.com>

* recog.c (asm_labels_ok): New function.
(check_asm_operands): Use asm_labels_ok.

From-SVN: r178580

12 years agoa-cbprqu.ads, [...]: Mark all entities as Implementation_Defined
Robert Dewar [Tue, 6 Sep 2011 09:46:21 +0000 (09:46 +0000)]
a-cbprqu.ads, [...]: Mark all entities as Implementation_Defined

2011-09-06  Robert Dewar  <dewar@adacore.com>

* a-cbprqu.ads, a-cbsyqu.ads, a-cuprqu.ads, a-cusyqu.ads,
a-intnam-aix.ads, a-intnam-darwin.ads, a-intnam-dummy.ads,
a-intnam-freebsd.ads, a-intnam-hpux.ads, a-intnam-irix.ads,
a-intnam-linux.ads, a-intnam-lynxos.ads, a-intnam-mingw.ads,
a-intnam-solaris.ads, a-intnam-tru64.ads,
a-intnam-vms.ads, a-intnam-vxworks.ads, a-intnam.ads, interfac.ads,
cstand.adb, s-maccod.ads: Mark all entities as Implementation_Defined
* einfo.ads, einfo.adb (Is_Implementation_Defined): New flag
* par-prag.adb: Add dummy entry for pragma Implementation_Defined
* s-rident.ads: Add new restriction No_Implementation_Identifiers
Add new profile No_Implementation_Extensions
* sem_prag.adb: Implement pragma Implementation_Defined Implement
profile No_Implementation_Extensions
* sem_util.adb: Minor reformatting (Set_Entity_With_Style_Check):
Check violation of restriction No_Implementation_Identifiers
* snames.ads-tmpl: Add entries for pragma Implementation_Defined
Add entry for Name_No_Implementation_Extensions

2011-09-06  Robert Dewar  <dewar@adacore.com>

* impunit.ads: Minor reformatting.

From-SVN: r178579

12 years ago[multiple changes]
Arnaud Charlet [Tue, 6 Sep 2011 09:37:22 +0000 (11:37 +0200)]
[multiple changes]

2011-09-06  Robert Dewar  <dewar@adacore.com>

* ali.adb, sem_ch13.adb, lib-xref.adb: Minor reformatting.

2011-09-06  Pascal Obry  <obry@adacore.com>

* s-osinte-linux.ads, s-oscons-tmplt.c: Use oscons to define sigset_t
types.

From-SVN: r178578

12 years ago[multiple changes]
Arnaud Charlet [Tue, 6 Sep 2011 09:32:56 +0000 (11:32 +0200)]
[multiple changes]

2011-09-06  Robert Dewar  <dewar@adacore.com>

* g-socket.adb: Minor reformatting

2011-09-06  Javier Miranda  <miranda@adacore.com>

* ali.adb (Scan_ALI): Add missing support to load references of
entities imported from other languages.
* ali.ads (Xref_Record): Adding new fields to store the language and
name of an imported entity.
* lib-xref.adb (Output_Import_Export_Info): Fix typo
in comment.

2011-09-06  Ed Schonberg  <schonberg@adacore.com>

* sem_prag.adb (Analyze_Pragma, case Type_Invariant): A type invariant
is allowed on a full type declaration if it is the completion of
a private declarations.
* sem_ch13.adb (Analyze_Aspect_Specifications): An invariant
aspect is allowed on a full type declaration in the private part
of a package.

From-SVN: r178577

12 years ago[multiple changes]
Arnaud Charlet [Tue, 6 Sep 2011 09:28:33 +0000 (11:28 +0200)]
[multiple changes]

2011-09-06  Robert Dewar  <dewar@adacore.com>

* sem_ch8.ads: Minor reformatting

2011-09-06  Thomas Quinot  <quinot@adacore.com>

* s-oscons-tmplt.c, g-socket.adb (GNAT.Sockets.Clear,Set,Is_Set):
Guard against socket values that are not in [0;FD_SETSIZE[

From-SVN: r178576

12 years ago[multiple changes]
Arnaud Charlet [Tue, 6 Sep 2011 09:26:02 +0000 (11:26 +0200)]
[multiple changes]

2011-09-06  Robert Dewar  <dewar@adacore.com>

* s-osinte-linux.ads, a-iteint.ads, exp_ch6.adb, s-solita.adb: Minor
reformatting.

2011-09-06  Arnaud Charlet  <charlet@adacore.com>

* s-linux-alpha.ads: Minor reformatting
* s-oscons-tmplt.c: Fix generated comments in s-oscons template.
Use sizeof instead of corresponding C defines in s-oscons template.

From-SVN: r178575

12 years agoa-convec.ads, [...]: Minor reformatting.
Vadim Godunko [Tue, 6 Sep 2011 09:20:36 +0000 (09:20 +0000)]
a-convec.ads, [...]: Minor reformatting.

2011-09-06  Vadim Godunko  <godunko@adacore.com>

* a-convec.ads, a-iteint.ads: Minor reformatting.

From-SVN: r178574

12 years ago[multiple changes]
Arnaud Charlet [Tue, 6 Sep 2011 09:15:15 +0000 (11:15 +0200)]
[multiple changes]

2011-09-06  Vincent Celier  <celier@adacore.com>

* projects.texi: Add menus and @node lines.

2011-09-06  Ed Schonberg  <schonberg@adacore.com>

* exp_ch6.adb (Expand_Inlined_Call): Handle properly the case
where the return type is an unconstrained array and the context
is an assignment. Optimize the case when the target of the
assignment is a selected component.

2011-09-06  Arnaud Charlet  <charlet@adacore.com>

* s-solita.adb: Update comments.

2011-09-06  Pascal Obry  <obry@adacore.com>

* s-linux.ads, s-linux-alpha.ads, s-linux-hppa.ads, s-linux-mipsel.ads,
s-linux-sparc.ads: Remove hard coded and now wrong definitions.
* s-oscons-tmplt.c: Add support for generating pthread related
types size on GNU/Linux as done for Darwin.
* s-osinte-linux.ads: Use s-oscons to define the pthread types.

From-SVN: r178573

12 years ago[multiple changes]
Arnaud Charlet [Tue, 6 Sep 2011 09:02:44 +0000 (11:02 +0200)]
[multiple changes]

2011-09-06  Robert Dewar  <dewar@adacore.com>

* exp_ch6.adb: Fix minor typo.

2011-09-06  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_ch7.adb: Remove with and use clauses for Get_Targ.
(Alignment_Of): Moved to the body of Nearest_Multiple_Rounded_Up.
(Double_Size_Of): Alphabetized. Update the comment on usage.
(Make_Finalize_Address_Stmts): Update comments and reformat code.
(Nearest_Multiple_Rounded_Up): New routine.
(Size_Of): Update comment on usage. The generated expression now
accounts for alignment gaps by rounding the size of the type to the
nearest multiple rounded up of the type's alignment.

From-SVN: r178572

12 years ago[multiple changes]
Arnaud Charlet [Tue, 6 Sep 2011 08:27:42 +0000 (10:27 +0200)]
[multiple changes]

2011-09-06  Robert Dewar  <dewar@adacore.com>

* exp_ch7.adb, g-comlin.adb: Minor reformatting.

2011-09-06  Steve Baird  <baird@adacore.com>

* exp_ch4.adb (Expand_Allocator_Expression): Look through
derived subprograms in checking for presence of an
Extra_Accessibility_Of_Result formal parameter.
* exp_ch6.adb (Expand_Call): Look through derived subprograms in
checking for presence of an Extra_Accessibility_Of_Result formal
parameter.
(Expand_Call.Add_Actual_Parameter): Fix a bug in the
case where the Parameter_Associatiations attribute is already set,
but set to an empty list.
(Needs_Result_Accessibility_Level):
Unconditionally return False. This is a temporary
change, disabling the Extra_Accessibility_Of_Result
mechanism.
(Expand_Simple_Function_Return): Check for
Extra_Accessibility_Of_Result parameter's presence instead of
testing Ada_Version when generating a runtime accessibility
check which makes use of the parameter.

From-SVN: r178571

12 years ago[multiple changes]
Arnaud Charlet [Tue, 6 Sep 2011 08:07:27 +0000 (10:07 +0200)]
[multiple changes]

2011-09-06  Ed Schonberg  <schonberg@adacore.com>

* exp_ch4.adb (Expand_N_Case_Expression): Actions created for the
expression in a given case alternative must be attached to the
statement list of the ccrresponding case statement alternative
They cannot be propagated ahead of the case statement, because
the validity of the expression that generated the action may
hold only for that alternative.

2011-09-06  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_ch7.adb: Remove with and use clauses for Get_Targ.
(Alignment_Of): Remove the code for strict alignment targets.
(Double_Alignment_Of): Removed.
(Double_Size_Of): New routine.
(Make_Finalize_Address_Stmts): Change the
calculation of the dope's size.  Update relevant comments.
(Size_Of): New routine.

From-SVN: r178570

12 years ago[multiple changes]
Arnaud Charlet [Tue, 6 Sep 2011 08:03:04 +0000 (10:03 +0200)]
[multiple changes]

2011-09-06  Steve Baird  <baird@adacore.com>

* einfo.ads (Extra_Accessibility): Update associated comment to use
the term "present" correctly ("present" just means that it is not
an error to query the value of the attribute - it does not imply
that the value must be non-null).
(Extra_Constrained): Ditto.
(Is_Visible_Formal): Ditto.
(Extra_Accessibility_Of_Result) Ditto; also add Inline pragma.
(Set_Extra_Accessibility_Of_Result): Add Inline pragma.
* exp_ch4.adb (Expand_Allocator_Expression): Improve a comment.
* exp_ch6.adb (Expand_Call): The callee may require an
Extra_Accessibility_Of_Result actual parameter even if Ada_Version
< Ada_2012. This can occur if the callee is exported from a Gnat
runtimes unit. Also improve a comment.

2011-09-06  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_ch7.adb: Add with and use clauses for Get_Targ.
(Alignment_Of): Rename formal parameter Some_Typ
to Typ, update related comment. Use the allocator alignment on
strict alignment targets such as SPARC.
(Double_Alignment_Of): Rename formal parameter Some_Typ to Typ, update
related comment.

2011-09-06  Gary Dismukes  <dismukes@adacore.com>

* sem_ch9.adb (Check_Interfaces): Test
Is_Limited_Type rather than Is_Limited_Record when checking that
the partial view of a synchronized full view must be limited.

2011-09-06  Emmanuel Briot  <briot@adacore.com>

* g-comlin.adb (Free): Fix memory leak.

From-SVN: r178569

12 years agosem_util.adb, [...]: Minor reformatting and code reorganization.
Robert Dewar [Tue, 6 Sep 2011 07:56:50 +0000 (07:56 +0000)]
sem_util.adb, [...]: Minor reformatting and code reorganization.

2011-09-06  Robert Dewar  <dewar@adacore.com>

* sem_util.adb, exp_ch6.adb: Minor reformatting and code reorganization.

From-SVN: r178568

12 years agoeinfo.ads (Extra_Accessibility_Of_Result): New function...
Steve Baird [Tue, 6 Sep 2011 07:46:28 +0000 (07:46 +0000)]
einfo.ads (Extra_Accessibility_Of_Result): New function...

2011-09-06  Steve Baird  <baird@adacore.com>

* einfo.ads (Extra_Accessibility_Of_Result): New function; in the
(Ada2012) cases described in AI05-0234 where the accessibility
level of a function result is "determined by the point of
call", an implicit parameter representing that accessibility
level is passed in. Extra_Accessibilty_Of_Result yields this
additional formal parameter. Extra_Accessibility_Of_Result
is analogous to the existing Extra_Accessibility
function used in the implementation of access parameters.
(Set_Extra_Accessibility_Of_Result): New procedure; sets
Extra_Accessibility_Of_Result attribute.
* einfo.adb (Extra_Accessibility_Of_Result): New function.
(Set_Extra_Accessibility_Of_Result): New procedure.
(Write_Field19_Name): Display Extra_Accessibilty_Of_Result attribute.
* sem_util.adb (Dynamic_Accessibility_Level): Set Etype of
an accessibility level literal to Natural; introduce a nested
function, Make_Level_Literal, to do this.
* exp_ch6.ads (Needs_Result_Accessibility_Level): New function;
determines whether a given function (or access-to-function
type) needs to have an implicitly-declared accessibility-level
parameter added to its profile.
(Add_Extra_Actual_To_Call): Export an existing procedure which was
previously declared in the body of Exp_Ch6.
* exp_ch6.adb (Add_Extra_Actual_To_Call): Export declaration by moving
it to exp_ch6.ads.
(Has_Unconstrained_Access_Discriminants): New Function; a
predicate on subtype entities which returns True if the given
subtype is unconstrained and has one or more access discriminants.
(Expand_Call): When expanding a call to a function which takes an
Extra_Accessibility_Of_Result parameter, pass in the appropriate
actual parameter value. In the case of a function call which is
used to initialize an allocator, this may not be possible because
the Etype of the allocator may not have been set yet. In this
case, we defer passing in the parameter and handle it later in
Expand_Allocator_Expression.
(Expand_Simple_Function_Return): When returning from a function which
returns an unconstrained subtype having at least one access
discriminant, generate the accessibility check needed to ensure that
the function result will not outlive any objects designated by its
discriminants.
(Needs_Result_Accessibility_Level): New function; see exp_ch6.ads
description.
* exp_ch4.adb (Expand_Allocator_Expression): When a function call
is used to initialize an allocator, we may need to pass in "the
accessibility level determined by the point of call" (AI05-0234)
to the function. Expand_Call, where such actual parameters are
usually generated, is too early in this case because the Etype of
the allocator (which is used in determining the level to be passed
in) may not have been set yet when Expand_Call executes. Instead,
we generate code to pass in the appropriate actual parameter
in Expand_Allocator_Expression.
* sem_ch6.adb (Create_Extra_Formals): Create
the new Extra_Accessibility_Of_Result formal if
Needs_Result_Accessibility_Level returns True. This includes the
introduction of a nested procedure, Check_Against_Result_Level.

From-SVN: r178567

12 years agoMakefile.in (X86_TARGET_PAIRS): Remove duplicate declaration.
Arnaud Charlet [Tue, 6 Sep 2011 07:06:52 +0000 (07:06 +0000)]
Makefile.in (X86_TARGET_PAIRS): Remove duplicate declaration.

* gcc-interface/Makefile.in (X86_TARGET_PAIRS): Remove duplicate
declaration.

From-SVN: r178566

12 years agoDaily bump.
GCC Administrator [Tue, 6 Sep 2011 00:18:39 +0000 (00:18 +0000)]
Daily bump.

From-SVN: r178562

12 years agosol2-64.h: Move ...
David S. Miller [Mon, 5 Sep 2011 22:23:40 +0000 (22:23 +0000)]
sol2-64.h: Move ...

* config/sparc/sol2-64.h: Move ...
* config/sparc/default-64.h: ... to here.  Update comment.
* config.gcc: Update Solaris sparc to use default-64.h, also
prefix this header into the list on sparc64-*-linux.
* config/sparc/linux64.h (TARGET_DEFAULT): Only override if
TARGET_64BIT_DEFAULT is defined.  Remove commented out reference
to MASK_HARD_QUAD.

From-SVN: r178558

12 years agore PR target/49606 (mips64: o64 Unrecognizable insn when one noreturn function callin...
Richard Sandiford [Mon, 5 Sep 2011 19:39:27 +0000 (19:39 +0000)]
re PR target/49606 (mips64: o64 Unrecognizable insn when one noreturn function calling another noreturn function)

gcc/
PR target/49606
* config/mips/mips.h (ABI_HAS_64BIT_SYMBOLS): Check Pmode.
(PMODE_INSN): New macro.
* config/mips/mips.c (gen_load_const_gp): Use PMODE_INSN.
(mips_got_load, mips_expand_synci_loop): Likewise.
(mips_save_gp_to_cprestore_slot): Handle SImode and DImode
cprestore patterns.
(mips_emit_loadgp): Use PMODE_INSN.  Handle SImode and DImode
copygp_mips16 patterns.
(mips_expand_prologue): Handle SImode and DImode potential_cprestore
and use_cprestore patterns.
(mips_override_options): Check for incompatible -mabi and -mlong
combinations.
* config/mips/mips.md (unspec_got<mode>): Rename to...
(unspec_got_<mode>): ...this.
(copygp_mips16): Use the Pmode iterator.
(potential_cprestore, cprestore, use_cprestore): Likewise.
(clear_cache, indirect_jump): Use PMODE_INSN.
(indirect_jump<mode>): Rename to...
(indirect_jump_<mode>): ...this.
(tablejump): Use PMODE_INSN.
(tablejump<mode>): Rename to...
(tablejump_<mode>): ...this.
(exception_receiver): Handle restore_gp_si and restore_gp_di.
(restore_gp): Use the Pmode iterator.
* config/mips/mips-dsp.md (mips_lbux, mips_lhx, mips_lwx): Use
PMODE_INSN.

gcc/testsuite/
PR target/49606
* gcc.target/mips/abi-main.h: New file.
* gcc.target/mips/abi-o32-long32.c: New test.
* gcc.target/mips/abi-o32-long64.c: Likewise.
* gcc.target/mips/abi-o64-long32.c: Likewise.
* gcc.target/mips/abi-o64-long64.c: Likewise.
* gcc.target/mips/abi-n32-long32.c: Likewise.
* gcc.target/mips/abi-n32-long64.c: Likewise.
* gcc.target/mips/abi-n64-long32.c: Likewise.
* gcc.target/mips/abi-n64-long64.c: Likewise.
* gcc.target/mips/abi-o32-long32-no-shared.c: Likewise.
* gcc.target/mips/abi-o32-long64-no-shared.c: Likewise.
* gcc.target/mips/abi-o64-long32-no-shared.c: Likewise.
* gcc.target/mips/abi-o64-long64-no-shared.c: Likewise.
* gcc.target/mips/abi-n32-long32-no-shared.c: Likewise.
* gcc.target/mips/abi-n32-long64-no-shared.c: Likewise.
* gcc.target/mips/abi-n64-long32-no-shared.c: Likewise.
* gcc.target/mips/abi-n64-long64-no-shared.c: Likewise.
* gcc.target/mips/abi-o32-long32-pic.c: Likewise.
* gcc.target/mips/abi-o32-long64-pic.c: Likewise.
* gcc.target/mips/abi-o64-long32-pic.c: Likewise.
* gcc.target/mips/abi-o64-long64-pic.c: Likewise.
* gcc.target/mips/abi-n32-long32-pic.c: Likewise.
* gcc.target/mips/abi-n32-long64-pic.c: Likewise.
* gcc.target/mips/abi-n64-long32-pic.c: Likewise.
* gcc.target/mips/abi-n64-long64-pic.c: Likewise.
* gcc.target/mips/abi-eabi32-long32.c: Likewise.
* gcc.target/mips/abi-eabi32-long64.c: Likewise.
* gcc.target/mips/abi-eabi64-long32.c: Likewise.
* gcc.target/mips/abi-eabi64-long64.c: Likewise.
* gcc.target/mips/mips.exp: Make -mshared implied -mabicalls.
* gcc.target/mips/branch-2.c: Remove -mabicalls.
* gcc.target/mips/branch-3.c: Likewise.
* gcc.target/mips/branch-4.c: Likewise.
* gcc.target/mips/branch-5.c: Likewise.
* gcc.target/mips/branch-6.c: Likewise.
* gcc.target/mips/branch-7.c: Likewise.
* gcc.target/mips/branch-8.c: Likewise.
* gcc.target/mips/branch-9.c: Likewise.
* gcc.target/mips/branch-10.c: Likewise.
* gcc.target/mips/branch-11.c: Likewise.
* gcc.target/mips/branch-12.c: Likewise.
* gcc.target/mips/branch-13.c: Likewise.
* gcc.target/mips/lazy-binding-1.c: Likewise.

From-SVN: r178557

12 years agomips.c (mips_gimplify_va_arg_expr): Unshare off.
Richard Sandiford [Mon, 5 Sep 2011 19:38:20 +0000 (19:38 +0000)]
mips.c (mips_gimplify_va_arg_expr): Unshare off.

gcc/
* config/mips/mips.c (mips_gimplify_va_arg_expr): Unshare off.
Fix the type of the BIT_AND_EXPR.

From-SVN: r178556

12 years agoconfig.gcc: Recognize niagara3 and niagara4 in --with-cpu and --with-tune options.
David S. Miller [Mon, 5 Sep 2011 16:57:00 +0000 (16:57 +0000)]
config.gcc: Recognize niagara3 and niagara4 in --with-cpu and --with-tune options.

* config.gcc: Recognize niagara3 and niagara4 in --with-cpu
and --with-tune options.

From-SVN: r178555

12 years agosparc-opts.h (PROCESSOR_NIAGARA3, [...]): New.
David S. Miller [Mon, 5 Sep 2011 16:00:53 +0000 (16:00 +0000)]
sparc-opts.h (PROCESSOR_NIAGARA3, [...]): New.

* config/sparc/sparc-opts.h (PROCESSOR_NIAGARA3,
PROCESSOR_NIAGARA4): New.
* config/sparc/sparc.opt: Handle new processor types.
* config/sparc/sparc.md: Add to "cpu" attribute.
* config/sparc/sparc.h (TARGET_CPU_niagara3,
TARGET_CPU_niagara4): New, treat as niagara2.
* config/sparc/linux64.h: Handle niagara3 and niagara4
like niagara2.
* config/sparc/sol2.h: Likewise.
* config/sparc/niagara2.md: Schedule niagara3 like
niagara2.
* config/sparc/sparc.c (sparc_option_override): Add
niagara3 and niagara4 handling.
(sparc32_initialize_trampoline): Likewise.
(sparc64_initialize_trampoline): Likewise.
(sparc_use_sched_lookahead): Likewise.
(sparc_issue_rate): Likewise.
(sparc_register_move_cost): Likewise.
* config/sparc/driver-sparc.c (cpu_names): Use niagara3
and niagara4 as appropriate.
* doc/invoke.texi: Document new processor types.

From-SVN: r178554

12 years agoconfig.host: Add driver-sparc.o and sparc/x-sparc on native sparc*-*-linux* builds.
David S. Miller [Mon, 5 Sep 2011 15:58:57 +0000 (15:58 +0000)]
config.host: Add driver-sparc.o and sparc/x-sparc on native sparc*-*-linux* builds.

* config.host: Add driver-sparc.o and sparc/x-sparc on
native sparc*-*-linux* builds.
* config/sparc/driver-sparc.c: Correct Linux strings.
* config/sparc/linux.h: Add DRIVER_SELF_SPECS.
* config/sparc/linux64.h: Likewise.
* doc/invoke.texi: Document that Linux also supports
-mcpu=native and -mtune=native on sparc.

From-SVN: r178553

12 years ago[multiple changes]
Arnaud Charlet [Mon, 5 Sep 2011 14:30:30 +0000 (16:30 +0200)]
[multiple changes]

2011-09-05  Hristian Kirtchev  <kirtchev@adacore.com>

* s-finmas.adb (Set_Finalize_Address): Explain the reason
for the synchronization. Move the test for null from
s-stposu.Allocate_Any_Controlled to this routine since the check
needs to be protected too.
(Set_Heterogeneous_Finalize_Address): Explain the reason for the
synchronization code.
* s-finmas.ads (Set_Heterogeneous_Finalize_Address): Add comment
explaining the context in which this routine is used.
* s-stposu.adb (Allocate_Any_Controlled): Move the test for null
to s-finmas.Set_Finalize_Address.

2011-09-05  Ed Schonberg  <schonberg@adacore.com>

* einfo.ads: Document that itypes have no parent field.

2011-09-05  Robert Dewar  <dewar@adacore.com>

* rtsfind.adb (Check_CRT): Check for overloaded entity
* rtsfind.ads: Document that entities to be found by rtsfind
cannot be overloaded
* s-taenca.adb, s-tasren.adb, s-tpobop.adb, s-tpoben.ads, s-tpoben.adb
(Lock_Entries_With_Status): New name for Lock_Entries with two
arguments (changed to meet rtsfind no overloading rule).

From-SVN: r178551

12 years ago[multiple changes]
Arnaud Charlet [Mon, 5 Sep 2011 14:18:54 +0000 (16:18 +0200)]
[multiple changes]

2011-09-05  Hristian Kirtchev  <kirtchev@adacore.com>

* s-finmas.adb (Set_Finalize_Address (Address,
Finalize_Address_Ptr)): Renamed to Set_Heterogeneous_Finalize_Address.
(Set_Finalize_Address (in out Finalization_Master,
Finalize_Address_Ptr): Add synchronization code.
* s-finmas.ads (Set_Finalize_Address (Address,
Finalize_Address_Ptr)): Renamed to Set_Heterogeneous_Finalize_Address.
* s-stposu.adb (Allocate_Any_Controlled): Update the call to
Set_Finalize_Address for the heterogeneous case.

2011-09-05  Ed Schonberg  <schonberg@adacore.com>

* sem_ch3.adb: Undo previous change, not suitable after all.

2011-09-05  Robert Dewar  <dewar@adacore.com>

* einfo.ads: Minor comment clarification.

From-SVN: r178550

12 years ago[multiple changes]
Arnaud Charlet [Mon, 5 Sep 2011 14:15:16 +0000 (16:15 +0200)]
[multiple changes]

2011-09-05  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_ch6.adb (Move_Activation_Chain): Rewritten. The routine
no longer needs to search through the entities of the return
statement scope to find the _chain.
* sem_ch5.adb: Add with and use clauses for Exp_Ch6 and Sem_Ch6.
(Analyze_Block_Statement): Add local variable
Is_BIP_Return_Statement. Add machinery to install all entities
produced by the expansion of the return object declaration.
(Install_Return_Entities): New routine.
* sem_ch6.ads, sem_ch6.adb (Install_Entity): Moved from body to spec.

2011-09-05  Ed Schonberg  <schonberg@adacore.com>

* sem_ch10.adb (Analyze_Context): Apply simple fixup if context
of subunit is incomplete.
(Analyze_Proper_Body): If parent spec is not available, do not
attempt analysis.

From-SVN: r178549

12 years ago[multiple changes]
Arnaud Charlet [Mon, 5 Sep 2011 14:12:04 +0000 (16:12 +0200)]
[multiple changes]

2011-09-05  Ed Schonberg  <schonberg@adacore.com>

* sem_disp.adb (Find_Controlling_Arg): Add checks for
interface type conversions, that are expanded into dereferences.

2011-09-05  Hristian Kirtchev  <kirtchev@adacore.com>

* sem_ch6.adb (Analyze_Subprogram_Body_Helper):
Examine the parameter and return profile of a subprogram and swap
any incomplete types coming from a limited context with their
corresponding non-limited views.
(Exchange_Limited_Views): New routine.

2011-09-05  Ed Schonberg  <schonberg@adacore.com>

* sem_ch3.adb (Prepare_Private_Subtype_Completion): Set parent
of internal entity to the subtype declaration, so that when
entities are subsequently exchanged in a package body, the tree
remains properly formatted for ASIS.

From-SVN: r178548

12 years ago[multiple changes]
Arnaud Charlet [Mon, 5 Sep 2011 14:08:42 +0000 (16:08 +0200)]
[multiple changes]

2011-09-05  Johannes Kanig  <kanig@adacore.com>

* g-comlin.adb (Set_Usage): Additional optional argument to set help
message.
(Display_Help): display the user given help message, if available.

2011-09-05  Ed Schonberg  <schonberg@adacore.com>

* sem_ch3.adb (OK_For_Limited_Init_In_05): The expression
is legal if the original node is an identifier that is a
parameterless function call.

From-SVN: r178547

12 years agoprj-nmsc.adb: Minor reformatting.
Pascal Obry [Mon, 5 Sep 2011 14:03:26 +0000 (14:03 +0000)]
prj-nmsc.adb: Minor reformatting.

2011-09-05  Pascal Obry  <obry@adacore.com>

* prj-nmsc.adb: Minor reformatting.
Add support for standalone aggregate library.
(Check_Stand_Alone_Library): Handle standalone aggregate library.
* projects.texi: Fix documentation for aggregate library projects.

From-SVN: r178546

12 years agoipcp-3.c (mark_cell): Use mask 1 << 14 instead of 1 << 18 to avoid warning on int...
Georg-Johann Lay [Mon, 5 Sep 2011 14:00:37 +0000 (14:00 +0000)]
ipcp-3.c (mark_cell): Use mask 1 << 14 instead of 1 << 18 to avoid warning on int=16 platforms..

* gcc.dg/ipa/ipcp-3.c (mark_cell): Use mask 1 << 14 instead of 1
<< 18 to avoid warning on int=16 platforms..

From-SVN: r178545

12 years ago[multiple changes]
Arnaud Charlet [Mon, 5 Sep 2011 14:00:30 +0000 (16:00 +0200)]
[multiple changes]

2011-09-05  Robert Dewar  <dewar@adacore.com>

* sem_ch6.adb: Minor reformatting.

2011-09-05  Matthew Gingell  <gingell@adacore.com>

* s-tassta.adb (Task_Wrapper): Ensure that we don't try to write the
stack guard page on PPC Linux. This patch ensures the 64K guard page at
the bottom of the stack is not overwritten.

From-SVN: r178544

12 years ago[multiple changes]
Arnaud Charlet [Mon, 5 Sep 2011 13:58:39 +0000 (15:58 +0200)]
[multiple changes]

2011-09-05  Thomas Quinot  <quinot@adacore.com>

* exp_intr.adb, s-tasini.adb: Minor reformatting.

2011-09-05  Ed Schonberg  <schonberg@adacore.com>

* sem_ch3.adb (Access_Definition): If an access type declaration
appears in a child unit, the scope of whatever anonymous type
may be generated is the child unit itself.

2011-09-05  Ed Schonberg  <schonberg@adacore.com>

* sem_ch6.adb (Analyze_Expression_Function): Do not set
Comes_From_Source on rewritten body.
(Analyze_Subprogram_Body_Helper): Check that the original node for
the body comes from source, when determining whether expansion
of a protected operation is needed.

From-SVN: r178543

12 years ago[multiple changes]
Arnaud Charlet [Mon, 5 Sep 2011 13:53:59 +0000 (15:53 +0200)]
[multiple changes]

2011-09-05  Ed Schonberg  <schonberg@adacore.com>

* exp_aggr.adb (Replace_Type): If the target of the assignment is
a selected component and the right-hand side is a self-referential
access, the proper prefix of the rewritten attribute is a copy
of the left-hand side, not of its prefix.

2011-09-05  Robert Dewar  <dewar@adacore.com>

* sem_ch5.adb: Minor reformatting.
* sem_ch3.adb: Minor code clean up.

2011-09-05  Jose Ruiz  <ruiz@adacore.com>

* exp_ch9.adb (Make_Initialize_Protection): The fact that
restriction No_Dynamic_Attachment is in effect or not should
not influence the default priority associated to a protected
object. With this change, when the restriction is in effect (as
is the case for the Ravenscar profile) the default priority of
a protected object with an interrupt handler is in the range of
interrupt priorities (Default_Interrupt_Priority) as required
by D.3(10).

2011-09-05  Arnaud Charlet  <charlet@adacore.com>

* a-iteint.ads: Fix header.

2011-09-05  Vincent Celier  <celier@adacore.com>

* prj-conf.adb (Do_Autoconf): Make sure Obj_Dir always ends
with a directory separator to avoid output such as:
  "/path/to/objauto.cgpr" has been deleted.

From-SVN: r178542

12 years agostruct-layout-1_test.h: Decrease bitfield size to work even with -fshort-enums.
Jakub Jelinek [Mon, 5 Sep 2011 13:53:08 +0000 (15:53 +0200)]
struct-layout-1_test.h: Decrease bitfield size to work even with -fshort-enums.

* gcc.dg/compat/struct-layout-1_test.h: Decrease bitfield size
to work even with -fshort-enums.

From-SVN: r178540

12 years ago[multiple changes]
Arnaud Charlet [Mon, 5 Sep 2011 13:48:16 +0000 (15:48 +0200)]
[multiple changes]

2011-09-05  Robert Dewar  <dewar@adacore.com>

* sem_ch3.adb: Minor reformatting.

2011-09-05  Ed Schonberg  <schonberg@adacore.com>

* sem_ch5.adb: Better error message.

2011-09-05  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_aggr.adb: Add with and use clause for Exp_Ch6.
(Expand_Array_Aggregate): Detect a special case of an aggregate
which contains tasks in the context of an unexpanded return
statement of a build-in-place function.
* exp_ch6.adb: Add with and use clause for Exp_Aggr.
(Expand_N_Extended_Return_Statement): Detect a delayed aggregate
which contains tasks and expand it now that the original simple
return statement has been rewritten.
* exp_ch9.adb (Build_Activation_Chain_Entity): Code
reformatting. Do not create a chain for an extended return
statement if one is already available.
(Has_Activation_Chain): New routine.

From-SVN: r178539

12 years ago[multiple changes]
Arnaud Charlet [Mon, 5 Sep 2011 13:40:04 +0000 (15:40 +0200)]
[multiple changes]

2011-09-05  Marc Sango  <sango@adacore.com>

* sem_ch3.adb (Analyze_Object_Declaration): Remove
the wrong test and add the correct test to detect the violation
of illegal use of unconstrained string type in SPARK mode.

2011-09-05  Ed Schonberg  <schonberg@adacore.com>

* sem_ch5.adb (Analyze_Iteration_Specification): Improve error
message on an iterator over an array.

2011-09-05  Robert Dewar  <dewar@adacore.com>

* lib-xref-alfa.adb: Minor reformatting.

From-SVN: r178538

12 years ago[multiple changes]
Arnaud Charlet [Mon, 5 Sep 2011 13:19:04 +0000 (15:19 +0200)]
[multiple changes]

2011-09-05  Hristian Kirtchev  <kirtchev@adacore.com>

* einfo.adb (Write_Field28_Name): Update the choices for
Extra_Formals.
* einfo.ads: Update the use of Extra_Formals in various entities.
* exp_ch6.adb (Add_Task_Actuals_To_Build_In_Place_Call): Code
reformatting. Handle the case where the function call returns
an incomplete type coming from a limited with context. Generate
a reference to the _master when the master id is associated
with an access type.
(Is_Build_In_Place_Function_Call): Code
reformatting. The Alfa mode case must appear first since otherwise
we will carry out the function name retrieval regardless of the
compilation mode.
(Make_Build_In_Place_Call_In_Allocator): Code
reformatting. Handle the case where the function call returns
an incomplete type coming from a limited with context. Remove
the reference creation when adding the task-related actuals,
this is now done in Add_Task_Actuals_To_Build_In_Place_Call.
* exp_ch7.adb (Make_Set_Finalize_Address_Call): Remove local
variable Call. Remove the useless wrapping of Set_Finalize_Address
when the finalization master is a build-in-place extra formal. The
whole mechanism of controlled allocation in a build-in-place
context is already protected by an if statement.
* sem_aux.adb (Is_Immutably_Limited_Type): Handle the case
where the type might be related to a function which returns an
incomplete type coming from a limited with.
* sem_ch6.adb (Create_Extra_Formals): Comment
reformatting. Handle the case where the function returns an
incomplete type coming from a limited with context.

2011-09-05  Johannes Kanig  <kanig@adacore.com>

* lib-xref-alfa.adb (Is_Alfa_Reference): Improve test for constant
objects and rewrite case statement as /if/elsif/endif.

2011-09-05  Thomas Quinot  <quinot@adacore.com>

* sem_ch3.adb, sem_ch4.adb: Minor reformatting.

From-SVN: r178537

12 years agosem_ch3.adb, [...]: Minor reformatting.
Robert Dewar [Mon, 5 Sep 2011 13:12:33 +0000 (13:12 +0000)]
sem_ch3.adb, [...]: Minor reformatting.

2011-09-05  Robert Dewar  <dewar@adacore.com>

* sem_ch3.adb, sem_res.adb, par.adb, par-ch6.adb, g-comlin.adb,
exp_ch6.adb, lib-xref-alfa.adb: Minor reformatting.

From-SVN: r178536

12 years ago[multiple changes]
Arnaud Charlet [Mon, 5 Sep 2011 13:11:29 +0000 (15:11 +0200)]
[multiple changes]

2011-09-05  Gary Dismukes  <dismukes@adacore.com>

* exp_ch7.adb, exp_ch6.adb: Minor reformatting.

2011-09-05  Johannes Kanig  <kanig@adacore.com>

* lib-xref-alfa.adb: Update comments.

2011-09-05  Thomas Quinot  <quinot@adacore.com>

* sem_res.adb: Minor reformatting

From-SVN: r178535

12 years agoexp_ch6.adb (Build_In_Place_Formal): If extra formals are not present, create them...
Ed Schonberg [Mon, 5 Sep 2011 13:08:30 +0000 (13:08 +0000)]
exp_ch6.adb (Build_In_Place_Formal): If extra formals are not present, create them now.

2011-09-05  Ed Schonberg  <schonberg@adacore.com>

* exp_ch6.adb (Build_In_Place_Formal): If extra formals are not
present, create them now.  Needed in case the return type was
a limited view in the function declaration.
(Make_Build_In_Place_Call_In_Allocator): If return type contains
tasks, build the activation chain for it.  Pass a reference to
the Master_Id in call to Add_Task_Actuals_To_Build_In_Place call.
* exp_ch7.adb (Make_Set_Finalize_Address_Call): Clean up interface
with build_in_place calls.
* sem_ch9.adb (Analyze_Task_Type_Declaration): If partial view was
incomplete, inatialize its Corresponding_Record_Type component.
* sem_ch10.adb (Build_Chain): Initialize Private_Dependents field
of limited views.

From-SVN: r178534

12 years agoFix typo.
Arnaud Charlet [Mon, 5 Sep 2011 13:02:49 +0000 (15:02 +0200)]
Fix typo.

From-SVN: r178533

12 years ago[multiple changes]
Arnaud Charlet [Mon, 5 Sep 2011 13:01:17 +0000 (15:01 +0200)]
[multiple changes]

2011-09-05  Johannes Kanig  <kanig@adacore.com>

* lib-xref-alfa.adb (Is_Alfa_Reference): Filter constants from effect
information.

2011-09-05  Ed Schonberg  <schonberg@adacore.com>

* par-ch6.adb (P_Subprogram): In Ada2012 mode, if the subprogram
appears within a package specification and the token after "IS"
is not a parenthesis, assume that this is an unparenthesized
expression function, even if the token appears in a line by
itself.
* par.adb: Clarify use of Labl field of scope stack in error
recovery.

From-SVN: r178532

12 years ago[multiple changes]
Arnaud Charlet [Mon, 5 Sep 2011 12:58:26 +0000 (14:58 +0200)]
[multiple changes]

2011-09-05  Bob Duff  <duff@adacore.com>

* sem_res.adb (Resolve_Intrinsic_Operator): Use unchecked
conversions instead of normal type conversions in all cases where a
type conversion would be illegal. In particular, use unchecked
conversions when the operand types are private.

2011-09-05  Johannes Kanig  <kanig@adacore.com>

* lib-xref-alfa.adb (Is_Alfa_Reference): Never declare effects on
objects of task type or protected type.

From-SVN: r178531

12 years ago[multiple changes]
Arnaud Charlet [Mon, 5 Sep 2011 12:51:44 +0000 (14:51 +0200)]
[multiple changes]

2011-09-05  Ed Schonberg  <schonberg@adacore.com>

* sem_ch6.adb (Analyze_Expression_Function): If the expression
function comes from source, indicate that so does its rewriting,
so it is compatible with any subsequent expansion of the
subprogram body (e.g. when it is a protected operation).
* sem_ch4.adb: minor reformatting

2011-09-05  Hristian Kirtchev  <kirtchev@adacore.com>

* lib.adb (Check_Same_Extended_Unit): Comment rewriting. Use
Get_Source_Unit rather than Get_Code_Unit as instantiation unfolding
may lead to wrong ancestor package in the case of instantiated subunit
bodies. If a subunit is instantiated, follow the chain of instantiations
rather than the stub structure.

From-SVN: r178530

12 years agore PR target/50289 ([avr]: call-prologues saving/restoring global register variables)
Georg-Johann Lay [Mon, 5 Sep 2011 12:20:03 +0000 (12:20 +0000)]
re PR target/50289 ([avr]: call-prologues saving/restoring global register variables)

PR target/50289
* config/avr/avr.c (sequent_regs_live): Don't recognize sequences
that contain global register variable.

From-SVN: r178528

12 years agotree-cfg.c (replace_uses_by): Use fold_stmt, not fold_stmt_inplace.
Richard Guenther [Mon, 5 Sep 2011 10:31:04 +0000 (10:31 +0000)]
tree-cfg.c (replace_uses_by): Use fold_stmt, not fold_stmt_inplace.

2011-09-05  Richard Guenther  <rguenther@suse.de>

* tree-cfg.c (replace_uses_by): Use fold_stmt, not fold_stmt_inplace.

From-SVN: r178527

12 years agostor-layout.c (layout_type): Use size_binop for array size calculations.
Richard Guenther [Mon, 5 Sep 2011 09:03:08 +0000 (09:03 +0000)]
stor-layout.c (layout_type): Use size_binop for array size calculations.

2011-09-05  Richard Guenther  <rguenther@suse.de>

* stor-layout.c (layout_type): Use size_binop for array size
calculations.

c-family/
* c-common.c (complete_array_type): Use ssize_int (-1) instead
of integer_minus_one_node for empty array upper bounds.

From-SVN: r178526

12 years agoavr.h (progmem_section): Remove Declaration.
Georg-Johann Lay [Mon, 5 Sep 2011 08:30:17 +0000 (08:30 +0000)]
avr.h (progmem_section): Remove Declaration.

* config/avr/avr.h (progmem_section): Remove Declaration.
* config/avr/avr.c (progmem_section): Make static and rename to
progmem_swtable_section.
(avr_output_addr_vec_elt): No need to switch sections.
(avr_asm_init_sections): Use output_section_asm_op as section
callback for progmem_swtable_section.
(avr_output_progmem_section_asm_op): Remove Function.
(TARGET_ASM_FUNCTION_RODATA_SECTION): New Define.
(avr_asm_function_rodata_section): New static Function.
* config/avr/elf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Output
alignment 2**1 for jump tables.

From-SVN: r178525

12 years agoipa-inline-analysis.c (set_cond_stmt_execution_predicate): Check that parameter is...
Jan Hubicka [Mon, 5 Sep 2011 07:39:36 +0000 (09:39 +0200)]
ipa-inline-analysis.c (set_cond_stmt_execution_predicate): Check that parameter is SSA name.

* ipa-inline-analysis.c (set_cond_stmt_execution_predicate): Check that
parameter is SSA name.

From-SVN: r178524

12 years agorevert: fold-const.c (extract_muldiv_1): Remove bogus TYPE_IS_SIZETYPE special-casing.
Richard Guenther [Mon, 5 Sep 2011 07:32:37 +0000 (07:32 +0000)]
revert: fold-const.c (extract_muldiv_1): Remove bogus TYPE_IS_SIZETYPE special-casing.

2011-09-05  Richard Guenther  <rguenther@suse.de>

        Revert
        2011-08-31  Richard Guenther  <rguenther@suse.de>

* fold-const.c (extract_muldiv_1): Remove bogus TYPE_IS_SIZETYPE
special-casing.

From-SVN: r178523

12 years agore PR c++/49267 (Ambiguity with conversion functions "T&" and "T&&", initializing...
Jason Merrill [Mon, 5 Sep 2011 04:33:57 +0000 (00:33 -0400)]
re PR c++/49267 (Ambiguity with conversion functions "T&" and "T&&", initializing a "T&&")

PR c++/49267
* call.c (reference_binding): Don't set is_lvalue for an rvalue
reference rfrom.

From-SVN: r178521

12 years agore PR c++/49267 (Ambiguity with conversion functions "T&" and "T&&", initializing...
Jason Merrill [Mon, 5 Sep 2011 04:33:48 +0000 (00:33 -0400)]
re PR c++/49267 (Ambiguity with conversion functions "T&" and "T&&", initializing a "T&&")

PR c++/49267
PR c++/49458
DR 1328
* call.c (reference_binding): Set rvaluedness_matches_p properly
for reference to function conversion ops.
(compare_ics): Adjust.

From-SVN: r178520

12 years agoclass.c (trivial_default_constructor_is_constexpr): Rename from synthesized_default_c...
Jason Merrill [Mon, 5 Sep 2011 04:33:40 +0000 (00:33 -0400)]
class.c (trivial_default_constructor_is_constexpr): Rename from synthesized_default_constructor_is_constexpr.

* class.c (trivial_default_constructor_is_constexpr): Rename from
synthesized_default_constructor_is_constexpr.
(type_has_constexpr_default_constructor): Adjust.
(add_implicitly_declared_members): Call it instead.
(explain_non_literal_class): Explain about non-constexpr default ctor.
* cp-tree.h: Adjust.
* method.c (synthesized_method_walk): Adjust.
* semantics.c (explain_invalid_constexpr_fn): Handle defaulted
functions, too.

From-SVN: r178519

12 years agore PR c++/50248 ([C++0x] unnecessary instantiation of constexpr constructor)
Jason Merrill [Mon, 5 Sep 2011 04:33:08 +0000 (00:33 -0400)]
re PR c++/50248 ([C++0x] unnecessary instantiation of constexpr constructor)

PR c++/50248
Core 1358
* init.c (perform_member_init): Don't diagnose missing inits here.
(emit_mem_initializers): Or here.
* method.c (process_subob_fn): Don't instantiate constexpr ctors.
* semantics.c (cx_check_missing_mem_inits): New.
(explain_invalid_constexpr_fn): Call it.
(register_constexpr_fundef): Likewise.  Leave
DECL_DECLARED_CONSTEXPR_P set when the body is unsuitable.
(cxx_eval_call_expression): Adjust diagnostics.
(cxx_eval_constant_expression): Catch use of 'this' in a constructor.

From-SVN: r178518

12 years agoDaily bump.
GCC Administrator [Mon, 5 Sep 2011 00:18:52 +0000 (00:18 +0000)]
Daily bump.

From-SVN: r178517

12 years ago* gnat.dg/specs/debug1.ads: Tweak pattern.
Eric Botcazou [Sun, 4 Sep 2011 20:56:12 +0000 (20:56 +0000)]
* gnat.dg/specs/debug1.ads: Tweak pattern.

From-SVN: r178513

12 years agore PR libffi/49594 (bootstrap failure in libffi:darwin_closure for powerpc-darwin8)
Iain Sandoe [Sun, 4 Sep 2011 19:19:39 +0000 (19:19 +0000)]
re PR libffi/49594 (bootstrap failure in libffi:darwin_closure for powerpc-darwin8)

PR libffi/49594
* src/powerpc/darwin_closure.S (stubs): Make the stub binding
helper reference track the architecture pointer size.

From-SVN: r178512

12 years agore PR debug/49901 (gfortran.dg/debug/[pr35154-dwarf2.f/pr37738.f] failures on darwin)
Iain Sandoe [Sun, 4 Sep 2011 18:50:09 +0000 (18:50 +0000)]
re PR debug/49901 (gfortran.dg/debug/[pr35154-dwarf2.f/pr37738.f] failures on darwin)

PR debug/49901
* config/darwin.h (DEBUG_MACRO_SECTION): New macro.

From-SVN: r178510

12 years agore PR fortran/50227 ([OOP] ICE-on-valid with allocatable class variable)
Janus Weil [Sun, 4 Sep 2011 10:49:13 +0000 (12:49 +0200)]
re PR fortran/50227 ([OOP] ICE-on-valid with allocatable class variable)

2011-09-04  Janus Weil  <janus@gcc.gnu.org>

PR fortran/50227
* trans-types.c (gfc_sym_type): Check for proc_name.

2011-09-04  Janus Weil  <janus@gcc.gnu.org>

PR fortran/50227
* gfortran.dg/class_45a.f03: New.
* gfortran.dg/class_45b.f03: New.

From-SVN: r178509

12 years agore PR tree-optimization/50208 (ICE: in vinfo_for_stmt at tree-vectorizer.h:598 with...
Jakub Jelinek [Sun, 4 Sep 2011 09:22:38 +0000 (11:22 +0200)]
re PR tree-optimization/50208 (ICE: in vinfo_for_stmt at tree-vectorizer.h:598 with -ftree-vectorize -fno-tree-fre -fno-tree-pre)

        PR tree-optimization/50208
        * tree-vect-patterns.c (vect_handle_widen_mult_by_const): Add an
        argument.  Check that def_stmt is inside the loop.
        (vect_recog_widen_mult_pattern): Update calls to
        vect_handle_widen_mult_by_cons.
        (vect_operation_fits_smaller_type): Check that def_stmt is
        inside the loop.

Co-Authored-By: Ira Rosen <ira.rosen@linaro.org>
From-SVN: r178508

12 years agotree-vectorizer.c (vect_print_dump_info): Print line number when dumping to a file.
Ira Rosen [Sun, 4 Sep 2011 08:53:43 +0000 (08:53 +0000)]
tree-vectorizer.c (vect_print_dump_info): Print line number when dumping to a file.

        * tree-vectorizer.c (vect_print_dump_info): Print line
        number when dumping to a file.
        (vectorize_loops): Add new messages to dump file.

From-SVN: r178507

12 years agoDaily bump.
GCC Administrator [Sun, 4 Sep 2011 00:18:34 +0000 (00:18 +0000)]
Daily bump.

From-SVN: r178506

12 years agoipa-prop.h (ipa_jump_func_t): New typedef.
Martin Jambor [Sat, 3 Sep 2011 19:48:39 +0000 (21:48 +0200)]
ipa-prop.h (ipa_jump_func_t): New typedef.

2011-09-03  Martin Jambor  <mjambor@suse.cz>

* ipa-prop.h (ipa_jump_func_t): New typedef.
(struct ipa_edge_args): Removed field argument_count, field
jump_functions turned into a vector.
(ipa_set_cs_argument_count): Removed.
(ipa_get_cs_argument_count): Updated to work on vectors.
(ipa_get_ith_jump_func): Likewise.
* ipa-prop.c (ipa_count_arguments): Removed.
(compute_scalar_jump_functions): Use ipa_get_ith_jump_func to access
jump functions.  Update caller.
(compute_pass_through_member_ptrs): Likewise.
(compute_cst_member_ptr_arguments): Likewise.
(ipa_compute_jump_functions_for_edge): Get number of arguments from
the statement, allocate vector.
(ipa_compute_jump_functions): Do not call ipa_count_arguments.
(duplicate_ipa_jump_func_array): Removed.
(ipa_edge_duplication_hook): Use VEC_copy, do not copy argument count.
(ipa_read_node_info): Allocate vector.

From-SVN: r178502

12 years agore PR middle-end/50232 (reorg.c:3971: undefined reference to `make_return_insns')
John David Anglin [Sat, 3 Sep 2011 16:30:32 +0000 (16:30 +0000)]
re PR middle-end/50232 (reorg.c:3971: undefined reference to `make_return_insns')

PR middle-end/50232
Correct log entry for last change.

From-SVN: r178501

12 years agoPR Bug middle-end/50232
John David Anglin [Sat, 3 Sep 2011 16:21:27 +0000 (16:21 +0000)]
PR Bug middle-end/50232

PR Bug middle-end/50232
* config/pa/pa.md (return): Define "return" insn pattern.
(epilogue): Use it when no epilogue is needed.
* config/pa/pa.c (pa_can_use_return_insn): New function.
* config/pa/pa-protos.h (pa_can_use_return_insn): Declare.

From-SVN: r178500

12 years agocfgexpand.c (add_stack_var): Assert that the alignment is not zero.
Eric Botcazou [Sat, 3 Sep 2011 14:56:53 +0000 (14:56 +0000)]
cfgexpand.c (add_stack_var): Assert that the alignment is not zero.

* cfgexpand.c (add_stack_var): Assert that the alignment is not zero.
* tree-ssa-ccp.c (fold_builtin_alloca_for_var): Revert latest change.
Force at least BITS_PER_UNIT alignment on the new variable.

From-SVN: r178499

12 years agoDaily bump.
GCC Administrator [Sat, 3 Sep 2011 00:18:53 +0000 (00:18 +0000)]
Daily bump.

From-SVN: r178497

12 years agoopts.c (print_specific_help): Fix off-by-one compare in assertion check.
Gary Funck [Fri, 2 Sep 2011 21:19:17 +0000 (21:19 +0000)]
opts.c (print_specific_help): Fix off-by-one compare in assertion check.

2011-09-02  Gary Funck <gary@intrepid.com>

* opts.c (print_specific_help): Fix off-by-one compare in
assertion check.
* opts.h (CL_PARAMS, CL_WARNING, CL_OPTIMIZATION, CL_DRIVER,
CL_TARGET, CL_COMMON, CL_JOINED, CL_SEPARATE, CL_UNDOCUMENTED):
Increase by +5 to allow for more languages.
* optc-gen.awk: Generate #if that ensures that the number of
languages is within the implementation-defined limit.

From-SVN: r178491

12 years agore PR middle-end/50260 (internal compiler error: Segmentation fault at ../../gcc...
Michael Matz [Fri, 2 Sep 2011 18:31:47 +0000 (18:31 +0000)]
re PR middle-end/50260 (internal compiler error: Segmentation fault  at ../../gcc/gcc/tree-ssa-live.c:88)

PR middle-end/50260
* ipa-split.c (split_function): Call add_referenced_var.

* tree-ssa-phiopt.c (cond_store_replacement): Don't call get_var_ann.
(cond_if_else_store_replacement_1): Ditto.
* tree-ssa-pre.c (get_representative_for): Ditto.
(create_expression_by_pieces): Ditto.
(insert_into_preds_of_block): Ditto.
* tree-sra.c (create_access_replacement): Ditto.
(get_replaced_param_substitute): Ditto.

From-SVN: r178489

12 years agoc6x.md (collapse-ndfa, [...]): New automata_options.
Bernd Schmidt [Fri, 2 Sep 2011 17:53:54 +0000 (17:53 +0000)]
c6x.md (collapse-ndfa, [...]): New automata_options.

* config/c6x/c6x.md (collapse-ndfa, no-comb-vect): New
automata_options.
(d1, l1, s1, m1, d2, l2, s2, m2): Changed to define_query_cpu_unit.
(l1w, s1w, l2w, s2w): Define in the main automaton.
(fps1, fpl1, adddps1, adddpl1, fps2, fpl2, adddps2, adddpl2): New
units.
* config/c6x/c6x.c (c6x_sched_insn_info): Add unit_mask member.
(c6x_unit_names): Add the new units.
(c6x_unit_codes): New static array.
(UNIT_QID_D1, UNIT_QID_L1, UNIT_QID_S1, UNIT_QID_M1, UNIT_QID_FPS1,
UNIT_QID_FPL1, UNIT_QID_ADDDPS1, UNIT_QID_ADDDPL1,
UNIT_QID_SIDE_OFFSET): New macros.
(RESERVATION_S2): Adjust value.
(c6x_option_override): Compute c6x_unit_codes.
(assign_reservations): Take the unit_mask of the last instruction
into account.  Detect floating point reservations by looking for
the new units.  Don't assign reservations if the field is already
nonzero.
(struct c6x_sched_context): Add member prev_cycle_state_ctx.
(init_sched_state): Initialize it.
(c6x_clear_sched_context): Free it.
(insn_set_clock): Clear reservation.
(prev_cycle_state): New static variable.
(c6x_init_sched_context): Save it.
(c6x_sched_init): Allocate space for it and clear it.
(c6x_sched_dfa_pre_cycle_insn): New static function.
(c6x_dfa_new_cycle): Save state at the start of a new cycle.
(c6x_variable_issue): Only record units in the unit_mask that
were not set at the start of the cycle.
(c6x_variable_issue): Compute and store the unit_mask from the
current state.
(reorg_split_calls): Ensure the new information remains correct.
(TARGET_SCHED_DFA_NEW_CYCLE, TARGET_SCHED_CLEAR_SCHED_CONTEXT,
TARGET_SCHED_DFA_PRE_CYCLE_INSN): Define.
* config/c6x/c6x.h (CPU_UNITS_QUERY): Define.
* config/c6x/c6x-sched.md.in (fp4_ls_N__CROSS_, adddp_ls_N__CROSS_):
Add special reservations.
* config/c6x/c6x-sched.md: Regenerate.

From-SVN: r178488

12 years agotestsuite_allocator.h (tracker_allocator_counter:: allocate): Update allocation count...
François Dumont [Fri, 2 Sep 2011 15:54:16 +0000 (15:54 +0000)]
testsuite_allocator.h (tracker_allocator_counter:: allocate): Update allocation count only if allocation succeeded.

2011-09-02  François Dumont  <fdumont@gcc.gnu.org>

* testsuite/util/testsuite_allocator.h (tracker_allocator_counter::
allocate): Update allocation count only if allocation succeeded.

From-SVN: r178486

12 years agoipa-prop.h (ipa_node_params): Removed fields called_with_var_arguments and node_versi...
Martin Jambor [Fri, 2 Sep 2011 15:37:38 +0000 (17:37 +0200)]
ipa-prop.h (ipa_node_params): Removed fields called_with_var_arguments and node_versionable.

2011-09-02  Martin Jambor  <mjambor@suse.cz>

* ipa-prop.h (ipa_node_params): Removed fields
called_with_var_arguments and node_versionable.
(ipa_set_called_with_variable_arg): Removed.
(ipa_is_called_with_var_arguments): Likewise.
* ipa-cp.c (ipa_get_lattice): Fixed index check in an assert.
(determine_versionability): Do not check for type attributes and va
builtins.  Record versionability into inline summary.
(initialize_node_lattices): Do not check
ipa_is_called_with_var_arguments.
(propagate_constants_accross_call): Likewise, ignore arguments we do
not have PARM_DECLs for, set variable flag for parameters that were
not passed a value.
(create_specialized_node): Dump info that we cannot change signature.
* ipa-prop.c (ipa_compute_jump_functions): Do not care about variable
number of arguments.
(ipa_make_edge_direct_to_target): Likewise.
(ipa_update_after_lto_read): Likewise.
(ipa_node_duplication_hook): Do not copy called_with_var_arguments flag.
* tree-inline.c (copy_arguments_for_versioning): Copy PARM_DECLs if
they were remapped.

* testsuite/gcc.dg/ipa/ipcp-3.c: New test.

From-SVN: r178485

12 years agore PR tree-optimization/27460 (Does not vectorize statements with mixed type COND_EXPRs)
Richard Guenther [Fri, 2 Sep 2011 13:53:32 +0000 (13:53 +0000)]
re PR tree-optimization/27460 (Does not vectorize statements with mixed type COND_EXPRs)

2011-09-02  Richard Guenther  <rguenther@suse.de>

PR tree-optimization/27460
PR middle-end/29269
* doc/md.texi (vcond): Document.
* genopinit.c (optabs): Turn vcond{,u}_optab into a conversion
optab with two modes.
* optabs.h (enum convert_optab_index): Add COI_vcond, COI_vcondu.
(enum direct_optab_index): Remove DOI_vcond, DOI_vcondu.
(vcond_optab): Adjust.
(vcondu_optab): Likewise.
(expand_vec_cond_expr_p): Adjust prototype.
* optabs.c (get_vcond_icode): Adjust.
(expand_vec_cond_expr_p): Likewise.
(expand_vec_cond_expr): Likewise.
* tree-vect-stmts.c (vect_is_simple_cond): Return the comparison
vector type.
(vectorizable_condition): Allow differing types for comparison
and result.

* config/i386/i386.c (ix86_expand_sse_cmp): Use proper mode
for the comparison.
* config/i386/sse.md (vcond<mode>): Split to
vcond<V_256:mode><VF_256:mode>, vcond<V_128:mode><VF_128:mode>,
vcond<V_128:mode><VI124_128:mode> and
vcondu<V_128:mode><VI124_128:mode>.
(vcondv2di): Change to vcond<VI8F_128:mode>v2di.
(vconduv2di): Likewise.
* config/arm/neon.md (vcond<mode>): Change to vcond*<mode><mode>.
(vcondu<mode>): Likewise.
* config/ia64/vect.md (vcond<mode>): Likewise.
(vcondu<mode>): Likewise.
(vcondv2sf): Likewise.
* config/mips/mips-ps-3d.md (vcondv2sf): Likewise.
* config/rs6000/paired.md (vcondv2sf): Likewise.
* config/rs6000/vector.md (vcond<mode>): Likewise.
(vcondu<mode>): Likewise.
* config/spu/spu.md (vcond<mode>): Likewise.
(vcondu<mode>): Likewise.

* gcc.dg/vect/vect-cond-7.c: New testcase.

From-SVN: r178480

12 years agopretty-print.h (pp_unsigned_wide_integer): New.
Richard Guenther [Fri, 2 Sep 2011 13:52:12 +0000 (13:52 +0000)]
pretty-print.h (pp_unsigned_wide_integer): New.

2011-09-02  Richard Guenther  <rguenther@suse.de>

* pretty-print.h (pp_unsigned_wide_integer): New.
* tree-pretty-print.c (dump_generic_node): Print unsigned
host-wide-int fitting INTEGER_CSTs with pp_unsigned_wide_integer.

From-SVN: r178479

12 years agobitset: Trivial formatting fixes.
Paolo Carlini [Fri, 2 Sep 2011 13:49:10 +0000 (13:49 +0000)]
bitset: Trivial formatting fixes.

2011-09-02  Paolo Carlini  <paolo.carlini@oracle.com>

* include/std/bitset: Trivial formatting fixes.

From-SVN: r178478

12 years ago* ja.po: Update.
Joseph Myers [Fri, 2 Sep 2011 13:45:31 +0000 (14:45 +0100)]
* ja.po: Update.

From-SVN: r178476

12 years agore PR target/49987 (gcc.c-torture/compile/pr34856.c fails on powerpc-darwin9 from...
Richard Sandiford [Fri, 2 Sep 2011 13:32:10 +0000 (13:32 +0000)]
re PR target/49987 (gcc.c-torture/compile/pr34856.c fails on powerpc-darwin9 from r176228)

gcc/
PR target/49987
* config/rs6000/rs6000.c (paired_expand_vector_init): Check for
valid CONST_VECTOR operands.
(rs6000_expand_vector_init): Likewise.

From-SVN: r178474

12 years agobitset (_Base_bitset<>::_M_are_all_aux): Remove.
Paolo Carlini [Fri, 2 Sep 2011 13:30:32 +0000 (13:30 +0000)]
bitset (_Base_bitset<>::_M_are_all_aux): Remove.

2011-09-02  Paolo Carlini  <paolo.carlini@oracle.com>
    Marc Glisse  <marc.glisse@normalesup.org>

* include/std/bitset (_Base_bitset<>::_M_are_all_aux): Remove.
(_Base_bitset<>::_M_are_all): Add.
(bitset<>::all): Use the latter, improve implementation.

Co-Authored-By: Marc Glisse <marc.glisse@normalesup.org>
From-SVN: r178473

12 years agocgraph.h (cgraph_indirect_call_info): Removed field thunk_delta.
Martin Jambor [Fri, 2 Sep 2011 13:26:30 +0000 (15:26 +0200)]
cgraph.h (cgraph_indirect_call_info): Removed field thunk_delta.

2011-09-02  Martin Jambor  <mjambor@suse.cz>

* cgraph.h (cgraph_indirect_call_info): Removed field thunk_delta.
* gimple-fold.c (gimple_get_virt_method_for_binfo): Rewritten to use
BINFO_VTABLE.  Parameter delta removed, all callers updated.
* tree.c (free_lang_data_in_binfo): Clear BINFO_VIRTUALs instead
BINFO_VTABLE.
* cgraph.c (cgraph_make_edge_direct): Removed parameter delta, updated
all calls.
* cgraphunit.c (cgraph_redirect_edge_call_stmt_to_callee): Removed
handling of thunk_delta.
* ipa-cp.c (get_indirect_edge_target): Removed parameter delta.
(devirtualization_time_bonus): Do not handle thunk deltas.
(ipcp_discover_new_direct_edges): Likewise.
* ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
(try_make_edge_direct_simple_call): Likewise.
(try_make_edge_direct_virtual_call): Likewise.
* lto-cgraph.c (output_cgraph_opt_summary_p): Likewise.  Mark
parameter set as unused.
(output_edge_opt_summary): Likewise.  Mark both parameters as unused.
* lto-cgraph.c (output_cgraph_opt_summary_p): Likewise.  Mark
parameter set as unused.
(output_edge_opt_summary): Likewise.  Mark both parameters as unused.
(input_edge_opt_summary): Likewise.
* lto-streamer-out.c (lto_output_ts_binfo_tree_pointers): Do not stream
BINFO_VIRTUALS at all.
* lto-streamer-in.c (lto_input_ts_binfo_tree_pointers): Likewise.

* testsuite/g++.dg/ipa/devirt-3.C: Added a distraction method.
* testsuite/g++.dg/ipa/ivinline-7.C: Added a test for direct call
discovery, xfailed test for inlining.
* testsuite/g++.dg/ipa/ivinline-9.C: Likewise.

From-SVN: r178472

12 years agotree-ssa-ccp.c (fold_builtin_alloca_for_var): Do not fold alloca (0).
Richard Guenther [Fri, 2 Sep 2011 11:53:55 +0000 (11:53 +0000)]
tree-ssa-ccp.c (fold_builtin_alloca_for_var): Do not fold alloca (0).

2011-09-02  Richard Guenther  <rguenther@suse.de>

* tree-ssa-ccp.c (fold_builtin_alloca_for_var): Do not
fold alloca (0).
(ccp_fold_stmt): Continue replacing args when folding
alloca fails.

From-SVN: r178465

12 years agore PR libstdc++/50268 ([C++0x] bitset doesn't sanitize input)
Paolo Carlini [Fri, 2 Sep 2011 10:28:36 +0000 (10:28 +0000)]
re PR libstdc++/50268 ([C++0x] bitset doesn't sanitize input)

2011-09-02  Paolo Carlini  <paolo.carlini@oracle.com>
    Marc Glisse  <marc.glisse@normalesup.org>

PR libstdc++/50268
* include/std/bitset (struct _Sanitize_val): Add.
(bitset<>::bitset(unsigned long long)): Fix.
* testsuite/23_containers/bitset/cons/50268.cc: New.

Co-Authored-By: Marc Glisse <marc.glisse@normalesup.org>
From-SVN: r178463

12 years ago[multiple changes]
Arnaud Charlet [Fri, 2 Sep 2011 10:10:35 +0000 (12:10 +0200)]
[multiple changes]

2011-09-02  Robert Dewar  <dewar@adacore.com>

* sem_ch4.adb, sem_ch6.adb: Minor reformatting.
* gcc-interface/Make-lang.in: Update dependencies.

2011-09-02  Marc Sango  <sango@adacore.com>

* sem_ch3.adb (Analyze_Object_Declaration): Change
comment and add additional check to differentiate constant of
type string from others unconstrained type.

From-SVN: r178462

12 years agosinfo.ads, sinfo.adb: New semantic attribute Premature_Use...
Ed Schonberg [Fri, 2 Sep 2011 10:07:35 +0000 (10:07 +0000)]
sinfo.ads, sinfo.adb: New semantic attribute Premature_Use...

2011-09-02  Ed Schonberg  <schonberg@adacore.com>

* sinfo.ads, sinfo.adb: New semantic attribute Premature_Use,
present in incomplete type declarations to refine the error
message the full declaration is in the same unit.
* sem_ch4.adb (Analyze_Selected_Component): If the prefix is of
an incomplete type, set the Premature_Use for additional message.
* sem_ch3.adb (Find_Type_Name): If partial view is incomplete
and Premature_Use is set, place additional information at the
point of premature use.

From-SVN: r178461

12 years ago[multiple changes]
Arnaud Charlet [Fri, 2 Sep 2011 09:59:32 +0000 (11:59 +0200)]
[multiple changes]

2011-09-02  Bob Duff  <duff@adacore.com>

* sem_ch6.adb: (Check_Post_State): Suppress warning
"postcondition refers only to pre-state" when the expression has not
yet been analyzed, because it causes false alarms. This can happen when
the postcondition contains a quantified expression, because those are
analyzed later. This is a temporary/partial fix.
(Process_Post_Conditions): Minor: change wording of warning.

2011-09-02  Marc Sango  <sango@adacore.com>

* sem_ch3.adb (Analyze_Object_Declaration): Detect the violation of
illegal use of unconstrained string type in SPARK mode.
* sem_res.adb (Analyze_Operator_Symbol): Set the
right place where the string operand of concatenation should be
violate in SPARK mode.

From-SVN: r178460

12 years agosem_prag.adb, [...]: Minor reformatting.
Robert Dewar [Fri, 2 Sep 2011 09:57:21 +0000 (09:57 +0000)]
sem_prag.adb, [...]: Minor reformatting.

2011-09-02  Robert Dewar  <dewar@adacore.com>

* sem_prag.adb, sem_util.adb, sem_ch6.adb, prj-nmsc.adb,
exp_ch3.adb: Minor reformatting.

From-SVN: r178459

12 years ago[multiple changes]
Arnaud Charlet [Fri, 2 Sep 2011 09:52:36 +0000 (11:52 +0200)]
[multiple changes]

2011-09-02  Vincent Celier  <celier@adacore.com>

* prj-conf.adb (Add_Default_GNAT_Naming_Scheme): Declare "gcc"
as the compiler driver so Is_Compilable returns True for sources.
* prj-nmsc.adb (Override_Kind): When Kind is Sep, set the source
for the body.

2011-09-02  Ed Schonberg  <schonberg@adacore.com>

* sem_prag.adb (Analyze_PPC_In_Decl_Part): for a class-wide
condition, a reference to a controlling formal must be interpreted
as having the class-wide type (or an access to such) so that the
inherited condition can be properly applied to any overriding
operation (see ARM12 6.6.1 (7)).

2011-09-02  Tristan Gingold  <gingold@adacore.com>

* init.c (__gnat_is_vms_v7): Fix case and add prototype
for LIB$GETSYI.

2011-09-02  Javier Miranda  <miranda@adacore.com>

* exp_ch3.adb (Expand_N_Object_Declaration): Do not copy the
initializing expression of a class-wide interface object declaration
if its type is limited.

2011-09-02  Johannes Kanig  <kanig@adacore.com>

* sem_util.adb (Unique_Name): To obtain a unique name for enumeration
literals, take into account the type name; the type is *not*
the scope for an enumeration literal.

2011-09-02  Ed Schonberg  <schonberg@adacore.com>

* sem_ch6.adb (Check_Overriding_Indicator): add special check
to reject an overriding indicator on a user-defined Adjust
subprogram for a limited controlled type.

2011-09-02  Ed Schonberg  <schonberg@adacore.com>

* sem_res.adb (Resolve_Actuals): add missing call to Resolve
for an actual that is a function call returning an unconstrained
limited controlled type.

2011-09-02  Tristan Gingold  <gingold@adacore.com>

* g-socthi-vms.adb (c_sendmsg, c_recvmsg): Use unpacked msg if on vms 7

2011-09-02  Johannes Kanig  <kanig@adacore.com>

* alfa.ads (Name_Of_Heap_Variable): Change value of the HEAP variable
from "HEAP" to __HEAP Change comment that refers to that variable
* put_alfa.adb: Change comment that refers to that variable

From-SVN: r178458

12 years ago[multiple changes]
Arnaud Charlet [Fri, 2 Sep 2011 09:47:42 +0000 (11:47 +0200)]
[multiple changes]

2011-09-02  Robert Dewar  <dewar@adacore.com>

* exp_ch4.adb, exp_ch6.adb, prj-nmsc.adb: Minor reformatting.

2011-09-02  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_util.adb (Extract_Renamed_Object): Renamed to
Find_Renamed_Object. This routine has been reimplemented and now uses
tree traversal to locate a renamed object.
(Is_Aliased): Replace call to Extract_Renamed_Object with
Find_Renamed_Object.

2011-09-02  Tristan Gingold  <gingold@adacore.com>

* init.c: (__gnat_is_vms_v7): New function.

2011-09-02  Olivier Hainque  <hainque@adacore.com>

* tracebak.c (STOP_FRAME, ppc elf/vxworks case): Stop on frames
that have a misaligned backchain, necessarily bogus.

From-SVN: r178457

12 years agoexp_ch3.adb (Expand_Freeze_Class_Wide_Type): Do not create TSS routine Finalize_Addre...
Hristian Kirtchev [Fri, 2 Sep 2011 09:45:45 +0000 (09:45 +0000)]
exp_ch3.adb (Expand_Freeze_Class_Wide_Type): Do not create TSS routine Finalize_Address when compiling in Alfa mode.

2011-09-02  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_ch3.adb (Expand_Freeze_Class_Wide_Type): Do not create
TSS routine Finalize_Address when compiling in Alfa mode.
(Expand_Freeze_Record_Type): Do not create TSS routine
Finalize_Address when compiling in Alfa mode.
* exp_ch4.adb (Expand_Allocator_Expression): Do not produce a
call to Set_Finalize_Address in Alfa mode because Finalize_Address is
not built.
(Expand_N_Allocator): Do not produce a call to
Set_Finalize_Address in Alfa mode because Finalize_Address is not built.
* exp_ch6.adb (Make_Build_In_Place_Call_In_Allocator): Do not
produce a call to primitive Set_Finalize_Address in Alfa mode because
Finalize_Address is not built.
* exp_ch7.adb (Build_Finalization_Master): Do not create
finalization masters in Afa mode since they are not needed.
(Build_Finalizer): Do not create scope and library-level
finalizers in Alfa mode since they are not needed.
* exp_util.adb (Build_Allocate_Deallocate_Proc): Do not expand
"new" and "free" when applied to controlled objects in Alfa mode since
this is not needed.

From-SVN: r178456

12 years ago[multiple changes]
Arnaud Charlet [Fri, 2 Sep 2011 09:43:58 +0000 (11:43 +0200)]
[multiple changes]

2011-09-02  Vincent Celier  <celier@adacore.com>

* prj-nmsc.db: (Check_Stand_Alone_Library): For SALs, allow
only library names with the syntax of Ada identifiers, to avoid errors
when compiling the binder generated files.
* projects.texi: Document restriction on SAL library names

2011-09-02  Thomas Quinot  <quinot@adacore.com>

* a-chtgbo.adb: Minor comment fix.

From-SVN: r178455