libgo: Update to weekly.2011-12-06.

From-SVN: r182338

tree-cfg.c (replace_uses_by): Fixup TREE_CONSTANT for propagating all kinds of constants.

2011-12-14  Richard Guenther  <rguenther@suse.de>

* tree-cfg.c (replace_uses_by): Fixup TREE_CONSTANT for
propagating all kinds of constants.

From-SVN: r182337

re PR lto/51497 (The run time for the polyhedron test nf.f90 is ~10% slower with -flto after revision 182107)

2011-12-14  Richard Guenther  <rguenther@suse.de>

PR lto/51497
* lto-streamer-in.c (lto_read_body): Fixup local types
TYPE_CANONICAL and variant chain.

From-SVN: r182336

runtime: Bump memory limit in gc_test.

From-SVN: r182332

compiler: Permit omitting &T in composite literal.

From-SVN: r182331

MAINTAINERS (picochip): Changed my email address.

        * MAINTAINERS (picochip): Changed my email address.

From-SVN: r182330

tree.c (free_lang_data_in_type): Do not clear fields dependent on debuginfo level setting.

2011-12-14  Richard Guenther  <rguenther@suse.de>

* tree.c (free_lang_data_in_type): Do not clear fields
dependent on debuginfo level setting.
(find_decls_types_r): Use TYPE_STUB_DECL rathern than TREE_CHAIN.
* tree.h (TYPE_STUB_DECL): Properly check that we access
a type.

From-SVN: r182329

re PR target/50931 ([avr] Support a 24-bit scalar integer mode)

libgcc/
PR target/50931
* config/avr/t-avr (LIB1ASMSRC): Add _mulpsi3, _mulsqipsi3.
* config/avr/lib1funcs.S (__mulpsi3, __mulsqipsi3): New functions.
gcc/
PR target/50931
* config/avr/avr.md (mulpsi3): New expander.
(*umulqihipsi3, *umulhiqipsi3): New insns.
(*mulsqipsi3.libgcc, *mulpsi3.libgcc): New insns.
(mulsqipsi3, *mulpsi3): New insn-and-splits.
(ashlpsi3): Turn to expander.  Move insn code to...
(*ashlpsi3): ...this new insn.
testsuite/
PR target/50931
* gcc.target/avr/torture/int24-mul.c: New testcase.

From-SVN: r182328

tree-cfg.c (replace_uses_by): Only mark blocks altered that will make a difference.

2011-12-14  Richard Guenther  <rguenther@suse.de>

* tree-cfg.c (replace_uses_by): Only mark blocks altered
that will make a difference.  Only recompute ADDR_EXPR
invariantness if it could possibly have changed.  Do so
before folding the statement.

From-SVN: r182327

Revert the previous "Fix flags for edges from/to entry/exit basic blocks"
patch that caused regressions.

From-SVN: r182326

PR c++/51476 - ICE on PTRMEM_CST as template argument in c++11

gcc/cp/

PR c++/51476
* pt.c (convert_nontype_argument): Don't call maybe_constant_value
for PTRMEM_CST nodes.

gcc/testsuite/

PR c++/51476
* g++.dg/cpp0x/ptrmem-cst-arg1.C: New test.

From-SVN: r182324

re PR c++/51406 ([c++0x] Incorrect result of static_cast to rvalue reference to base class.)

PR c++/51406
PR c++/51161
* typeck.c (build_static_cast_1): Fix cast of lvalue to
base rvalue reference.

From-SVN: r182322

compiler: Move import of Go export data to gcc side of interface.

* go-backend.c: #include "simple-object.h" and "intl.h".
(GO_EXPORT_SEGMENT_NAME): Define if not defined.
(GO_EXPORT_SECTION_NAME): Likewise.
(go_write_export_data): Use GO_EXPORT_SECTION_NAME.
(go_read_export_data): New function.
* go-c.h (go_read_export_data): Declare.

From-SVN: r182321

re PR c++/51464 ([c++0x] ICE with invalid use of [])

/cp
2011-12-13  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/51464
* semantics.c (begin_lambda_type): Check begin_class_definition return
value for error_mark_node.
* parser.c (cp_parser_lambda_expression): Check begin_lambda_type
return value for error_mark_node.

/testsuite
2011-12-13  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/51464
* g++.dg/cpp0x/lambda/lambda-ice6.C: New.

From-SVN: r182320

Daily bump.

From-SVN: r182319

runtime: Change gc_test to test only newly allocated memory.

From-SVN: r182315

mksysinfo.sh: Ensure that IPV6 constants and types are defined.

From-SVN: r182314

syscall: Don't build wait status functions on RTEMS.

From-SVN: r182313

lto-streamer-out.c (write_symbol): Use proper 64-bit host type.

* lto-streamer-out.c (write_symbol): Use proper 64-bit host type.
* lto-cgraph.c (input_cgraph_opt_section): Use 'int' for offsets.
* lto-streamer-in.c (lto_read_body): Likewise.
(lto_input_toplevel_asms): Likewise.
* lto-section-in.c (lto_create_simple_input_block): Likewise.
* ipa-inline-analysis.c (inline_read_section): Likewise.
* ipa-prop.c (ipa_prop_read_section): Likewise.

* df.h (DF_NOTE): Fix typo in comment.
lto/
* lto.h (lto_parse_hex): Delete.
* lto.c (lto_read_decls): Use 'int' for offsets.
(lto_parse_hex): Make static and return proper 64-bit host type.
(lto_resolution_read): Use proper 64-bit host type.

From-SVN: r182310

runtime: RTEMS build fixes.

From Joel Sherrill.

From-SVN: r182309

revert an unattended change about access declaration

From-SVN: r182308

trans-mem.c (ipa_tm_create_version_alias): Set DECL_CONTEXT and DECL_LANG_SPECIFIC.

        * trans-mem.c (ipa_tm_create_version_alias): Set DECL_CONTEXT and
        DECL_LANG_SPECIFIC.

From-SVN: r182307

regmove.c (fixup_match_2): Only access call_used_regs with hard regs.

2011-12-13  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>

* regmove.c (fixup_match_2): Only access call_used_regs with hard
regs.

From-SVN: r182306

rwlock.cc (gtm_rwlock::write_lock_generic): Fix signed/unsigned comparison werror.

        * config/posix/rwlock.cc (gtm_rwlock::write_lock_generic): Fix
        signed/unsigned comparison werror.

From-SVN: r182302

Revert two unintended commits

From-SVN: r182301

mips-cpus.def: Add Octeon2.

2011-12-13  Andrew Pinski  <apinski@cavium.com>
            Adam Nemet  <anemet@caviumnetworks.com>

* config/mips/mips-cpus.def: Add Octeon2.
* config/mips/mips-tables.opt: Regenerate.
* config/mips/mips.md (define_attr "cpu"): Add Octeon2.
* config/mips/driver-native.c (host_detect_local_cpu): Support Octeon2 also.
* config/mips/octeon.md (octeon_arith): Add Octeon2.
(octeon_condmove): Likewise.
(octeon_load): Rename to ..
(octeon_load_o1): this.
(octeon_load_o2): New reserve.
(octeon_cop_o2): New reserve.
(octeon_store):  Match Octeon2 also.
(octeon_brj): Rename to ..
(octeon_brj_o1): this.
(octeon_brj_o2): New reserve.
(octeon_imul3): Rename to ...
(octeon_imul3_o1): this.
(octeon_imul3_o2): New reserve.
(octeon_imul): Rename to ...
(octeon_imul_o1): this.
(octeon_imul_o2): New reserve.
(octeon_mfhilo): Rename to ...
(octeon_mfhilo_o1): This.
(octeon_mfhilo_o2): New reserve.
(octeon_imadd): Rename to ...
(octeon_imadd_o1): this.
(octeon_imadd_o2): New reserve.
(octeon_idiv): Rename to ..
(octeon_idiv_o1): This.
(octeon_idiv_o2_si): New reserve.
(octeon_idiv_o2_di): Likewise.
(octeon_unknown): Match Octeon2 also.
* config/mips/mips.c (mips_rtx_cost_data): Add Octeon2 cost data.
(mips_issue_rate): Octeon2 can issue 2 at a time.
* config/mips/mips.h (TARGET_OCTEON): Match Octeon2 also.
(TARGET_OCTEON2): New define.
(TUNE_OCTEON): Match Octeon2 also.

2011-12-13  Andrew Pinski  <apinski@cavium.com>
            Adam Nemet  <anemet@caviumnetworks.com>

* gcc.target/mips/mips.exp (mips_option_groups): Fix debug.  Add
-fdump-* options.
* gcc.target/mips/octeon2-pipe-1.c: New testcase.
* gcc.target/mips/octeon-pipe-1.c: New testcase.

Co-Authored-By: Adam Nemet <anemet@caviumnetworks.com>
From-SVN: r182300

arm: Implement vec_perm and vec_perm_const for NEON.

From-SVN: r182299

Delete VEC_INTERLEAVE_*_EXPR.

* tree.def (VEC_INTERLEAVE_HIGH_EXPR, VEC_INTERLEAVE_LOW_EXPR): Remove.
* gimple-pretty-print.c (dump_binary_rhs): Don't handle
VEC_INTERLEAVE_HIGH_EXPR and VEC_INTERLEAVE_LOW_EXPR.
* expr.c (expand_expr_real_2): Likewise.
* tree-cfg.c (verify_gimple_assign_binary): Likewise.
* cfgexpand.c (expand_debug_expr): Likewise.
* tree-inline.c (estimate_operator_cost): Likewise.
* tree-pretty-print.c (dump_generic_node): Likewise.
* tree-vect-generic.c (expand_vector_operations_1): Likewise.
* fold-const.c (fold_binary_loc): Likewise.
* doc/generic.texi (VEC_INTERLEAVE_HIGH_EXPR,
VEC_INTERLEAVE_LOW_EXPR): Remove documentation.
* optabs.c (optab_for_tree_code): Don't handle
VEC_INTERLEAVE_HIGH_EXPR and VEC_INTERLEAVE_LOW_EXPR.
(expand_binop, init_optabs): Remove vec_interleave_high_optab
and vec_interleave_low_optab.
* genopinit.c (optabs): Likewise.
* optabs.h (OTI_vec_interleave_high, OTI_vec_interleave_low): Remove.
(vec_interleave_high_optab, vec_interleave_low_optab): Remove.
* doc/md.texi (vec_interleave_high, vec_interleave_low): Remove
documentation.
* tree-vect-stmts.c (gen_perm_mask): Renamed to...
(vect_gen_perm_mask): ... this.  No longer static.
(perm_mask_for_reverse, vectorizable_load): Adjust callers.
* tree-vectorizer.h (vect_gen_perm_mask): New prototype.
* tree-vect-data-refs.c (vect_strided_store_supported): Don't try
VEC_INTERLEAVE_*_EXPR, use can_vec_perm_p instead of
can_vec_perm_for_code_p.
(vect_permute_store_chain): Generate VEC_PERM_EXPR with interleaving
masks instead of VEC_INTERLEAVE_HIGH_EXPR and VEC_INTERLEAVE_LOW_EXPR.
* config/i386/i386.c (expand_vec_perm_interleave2): If
expand_vec_perm_interleave3 would handle it, return false.
(expand_vec_perm_broadcast_1): Don't use vec_interleave_*_optab.

From-SVN: r182298

sparc: Default memory model to TSO for Linux and Solaris.

        * config/sparc/tso.h: New file.
        * config.gcc (sparc-*-linux*, sparc*-*-solaris2*): Use it.
        * config/sparc/sparc.c (sparc_option_override): Honor
        SUBTARGET_DEFAULT_MEMORY_MODEL.
        * config/sparc/sparc.h (SUBTARGET_DEFAULT_MEMORY_MODEL): New.

From-SVN: r182297

libgo: Solaris compatibility patches.

From Rainer Orth.

From-SVN: r182296

libgo: Update to weekly.2011-12-02.

From-SVN: r182295

libitm: Conversion to c++11 atomics.

        * local_atomic: New file.
        * libitm_i.h: Include it.
        (gtm_thread::shared_state): Use atomic template.
        * beginend.cc (GTM::gtm_clock): Use atomic template.
        (global_tid): Use atomic template if 64-bit atomics available.
        (gtm_thread::gtm_thread): Update shared_state access.
        (gtm_thread::trycommit): Likewise.
        (choose_code_path): Update global_tid access.
        * method-gl.cc (gl_mg::orec): Use atomic template.  Update all users.
        * stmlock.h (GTM::gtm_clock): Use atomic template.
        (gtm_get_clock, gtm_inc_clock): Update accesses.
        * config/linux/rwlock.cc (gtm_rwlock::read_lock): Remove
        redundant __sync_synchronize after atomic shared_state access.
        * config/posix/rwlock.cc (gtm_rwlock::read_lock): Likewise.
        (gtm_rwlock::write_lock_generic): Likewise.
        (gtm_rwlock::read_unlock): Likewise.
        * config/alpha/target.h (atomic_read_barrier): Remove.
        (atomic_write_barrier): Remove.
        * config/x86/target.h (atomic_read_barrier): Remove.
        (atomic_write_barrier): Remove.

From-SVN: r182294

compiler: Fix multiple conversions to different named slices.

From-SVN: r182293

re PR c++/14258 (typename in a using declaration not supported)

gcc/testsuite/ChangeLog

2011-12-11  Fabien Chene  <fabien@gcc.gnu.org>

PR c++/14258
* g++.dg/template/using16.C: New.
* g++.dg/template/using17.C: New.

gcc/cp/ChangeLog

2011-12-11  Fabien Chene  <fabien@gcc.gnu.org>

PR c++/14258
* cp-tree.h (USING_DECL_TYPENAME_P): New macro.
* parser.c (cp_parser_nonclass_name): Handle using declarations
that refer to a dependent type.
(cp_parser_using_declaration): Set USING_DECL_TYPENAME_P to 1 if
the using declaration refers to a dependent type.

From-SVN: r182292

compiler: Permit converting between string and named []byte/[]rune.

From-SVN: r182291

trans-mem.c (struct diagnose_tm): Remove saw_unsafe.

        PR/51443
        * trans-mem.c (struct diagnose_tm): Remove saw_unsafe.
        (diagnose_tm_1): Same.
        (ipa_tm_execute): Do not test tm_may_enter_irr before we set it.
        (ipa_tm_scan_irr_function): Return gracefully when no
        DECL_STRUCT_FUNCTION.
        (ipa_tm_scan_irr_block): Believe the user on TM attributes.

From-SVN: r182290

re PR middle-end/50628 (gfortran.fortran-torture/execute/entry_4.f90 fails)

2011-12-13  Martin Jambor  <mjambor@suse.cz>

PR middle-end/50628
* tree-sra.c (propagate_subaccesses_across_link): Do not propagate
sub-accesses of scalar accesses.

From-SVN: r182289

re PR tree-optimization/51362 (ICE: SIGFPE (division by zero) in good_cloning_opportunity_p at ipa-cp.c:2401)

2011-12-13  Martin Jambor  <mjambor@suse.cz>

PR tree-optimization/51362
* ipa-cp.c (estimate_local_effects): When estimated size of a
specialized clone is zero, bump it to one.

* testsuite/gcc.dg/ipa/pr51362.c: New test.

From-SVN: r182288

re PR testsuite/51524 ([BMI2] New regression on 182266 vs 182257)

PR testsuite/51524
* gcc.target/i386/bmi2-mulx32-1.c (gen_mulx): Add attribute regparm(2).
* gcc.target/i386/bmi2-mulx32-2.c (calc_mulx_u32): Ditto.

From-SVN: r182287

re PR debug/48354 (internal compiler error: in splice_child_die, at dwarf2out.c:8064)

2011-12-13  Richard Guenther  <rguenther@suse.de>

PR lto/48354
* tree.c (find_decls_types_r): Also walk DECL_ORIGINAL_TYPE.
* tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
Stream DECL_ORIGINAL_TYPE.
* tree-streamer-out.c (write_ts_decl_non_common_tree_pointers):
Likewise.

lto/
* lto.c (lto_ft_decl_non_common): When we merged DECL_ORIGINAL_TYPE
with the type of the TYPE_DECL clear DECL_ORIGINAL_TYPE.

* g++.dg/lto/pr48354-1_0.C: New testcase.

From-SVN: r182286

life -> live; one favourite typo of mine.

From-SVN: r182285

Revert
2011-12-12  Jakub Jelinek  <jakub@redhat.com>

PR tree-optimization/51481
* gimple-fold.c (gimple_fold_call): Call
maybe_clean_or_replace_eh_stmt.  Avoid optimization if stmt has EH
edges, but gimple_fold_builtin result can't throw.

From-SVN: r182284

re PR tree-optimization/51117 (rev.181172 causes glibc build failure)

PR tree-optimization/51117
* tree-eh.c (sink_clobbers): New function.
(execute_lower_eh_dispatch): Call it for BBs ending with
internally throwing RESX.
* cfgexpand.c (add_scope_conflicts_1): Add all conflicts only
at the first real instruction.

Co-Authored-By: Michael Matz <matz@suse.de>
From-SVN: r182283

final.c (final_scan_insn): Guard the call to begin_epilogue debug hook.

2011-12-13  Tristan Gingold  <gingold@adacore.com>

* final.c (final_scan_insn): Guard the call to begin_epilogue
debug hook.

From-SVN: r182282

vmsdbgout.c (vmsdbgout_write_source_line): New function.

2011-12-13  Tristan Gingold  <gingold@adacore.com>

* vmsdbgout.c (vmsdbgout_write_source_line): New function.
(vmsdbgout_end_prologue): Call vmsdbgout_write_source_line.
(vmsdbgout_begin_epilogue): Likewise.
(vmsdbgout_end_epilogue): Likewise.
(vmsdbgout_source_line): Move code to vmsdbgout_write_source_line.

From-SVN: r182281

re PR tree-optimization/51481 (ice: dead STMT in EH table)

2011-12-13  Richard Guenther  <rguenther@suse.de>

PR middle-end/51481
* tree-cfg.c (replace_uses_by): Pass proper arguments to
maybe_clean_or_replace_eh_stmt.

From-SVN: r182280

re PR tree-optimization/51519 (ICE: in inline_small_functions, at ipa-inline.c:1410 with -O -fno-guess-branch-probability -findirect-inlining)

2011-12-13  Richard Guenther  <rguenther@suse.de>

PR tree-optimization/51519
* ipa-inline.c (edge_badness): Use edge growth in non-guessed
branch probability case as well.

* gcc.dg/pr51519.c: New testcase.

From-SVN: r182279

20111213-1_0.c: New testcase.

2011-12-13  Richard Guenther  <rguenther@suse.de>

* gcc.dg/lto/20111213-1_0.c: New testcase.

From-SVN: r182278

vect1.ad[sb]: Add more tests.

* gnat.dg/vect1.ad[sb]: Add more tests.
* gnat.dg/vect2.ad[sb]: Likewise.
* gnat.dg/vect3.ad[sb]: Likewise.
* gnat.dg/vect4.ad[sb]: Likewise.
* gnat.dg/vect5.ad[sb]: Likewise.
* gnat.dg/vect6.ad[sb]: Likewise.

From-SVN: r182276

re PR ada/49084 (bootstrap failure with Ada enabled)

PR ada/49084
* types.h (Byte): Change typedef to 'unsigned char'.
* atree.h (struct Flag_Word): Use Byte for 'convention' field.

From-SVN: r182275

SMS: Add missing free operation in mark_loop_unsched

From-SVN: r182271

Daily bump.

From-SVN: r182270

libgo: Update to weekly.2011-11-18.

From-SVN: r182266

re PR rtl-optimization/51495 (ICE: in force_nonfallthru_and_redirect, at cfgrtl.c:1171 with computed goto)

PR rtl-optimization/51495
* function.c (thread_prologue_and_epilogue_insns): Don't add
to bb_tail basic blocks that have EDGE_COMPLEX predecessor edges
from basic blocks not needing prologue.

* gcc.c-torture/compile/pr51495.c: New test.

From-SVN: r182265

re PR tree-optimization/51481 (ice: dead STMT in EH table)

PR tree-optimization/51481
* gimple-fold.c (gimple_fold_call): Call
maybe_clean_or_replace_eh_stmt.  Avoid optimization if stmt has EH
edges, but gimple_fold_builtin result can't throw.

* gcc.dg/pr51481.c: New test.

From-SVN: r182264

re PR rtl-optimization/21617 (CRC64 algorithm optimization problem on Intel 32-bit)

2011-12-12  Vladimir Makarov  <vmakarov@redhat.com>

PR rtl-optimization/21617
* ira-color.c (bucket_allocno_compare_func): Don't compare
allocno classes.  Compare number of hard registers needed.

From-SVN: r182263

re PR tree-optimization/45830 (Code+rodata increase with -ftree-switch-conversion)

PR tree-optimization/45830
* gcc.c-torture/execute/pr45830.c: Move from here to...
* gcc.dg/torture/pr45830.c: ...this new file.  Add dg-do and
dg-additional-options magic.

From-SVN: r182262

gfortran.dg/nested_modules_2.f90: Tighten test.

2011-12-12  Janne Blomqvist  <jb@gcc.gnu.org>

* gfortran.dg/nested_modules_2.f90: Tighten test.

From-SVN: r182258

re PR c++/51496 (FAIL: libgomp.c++/for-2.C  *  (test for excess errors))

PR c++/51496
* parser.c (cp_parser_omp_for_loop): When determining whether
to use cp_parser_omp_for_incr or cp_parser_expression and when
calling cp_parser_omp_for_incr, use real_decl instead of decl.

From-SVN: r182257

net/http: delete temporary files.

From-SVN: r182256

re PR tree-optimization/50569 (unaligned memory accesses generated for memcpy)

PR tree-optimization/50569
* tree-sra.c (build_ref_for_model): Replicate a chain of COMPONENT_REFs
in the expression of MODEL instead of just the last one.

From-SVN: r182252

Fix flags for edges from/to entry/exit basic blocks.

* cgraphunit.c (init_lowered_empty_function):
Fix flags for new edges.

From-SVN: r182251

revert: re PR target/51135 (SIGSEGV during exception cleanup on win32)

2011-12-12  Paolo Carlini  <paolo.carlini@oracle.com>

Revert:
2011-12-12  Kai Tietz  <ktietz@redhat.com>

PR libstdc++/51135
* libsupc++/cxxabi.h (__cxa_dtor_type): New type.
(__cxa_throw): Use it for destructor-argument.
* libsupc++/eh_throw.cc (__cxa_throw): Likewise.
* libsupc++/unwind-cxx.h (__cxa_exception): Change type of member
exceptionDestructor to __cxa_dtor_type.
* config/os/mingw32-w64/os_defines.h (_GLIBCXX_USE_THISCALL_ON_DTOR):
Define.
(__cxa_dtor_type): Declare target secific type variant.
* config/os/mingw32/os_defines.h: Likewise.

From-SVN: r182250

2011-12-12  Paolo Carlini  <paolo.carlini@oracle.com>

* Revert the last commit.

From-SVN: r182249

re PR testsuite/51511 (gcc.dg/pr45819.c fails on arm-linux-gnueabi due to PR50078+PR51466 fix)

PR testsuite/51511
* gcc.dg/pr45819.c: Add -w to dg-options.

From-SVN: r182247

re PR middle-end/51510 (gcc for arm-linux-gnueabi ICEs on gcc.c-torture/execute/20050713-1.c due to PR50074+PR51323 fix)

PR middle-end/51510
* calls.c (internal_arg_pointer_based_exp_scan): Don't use
VEC_safe_grow_cleared if idx is smaller than VEC_length.

From-SVN: r182246

re PR tree-optimization/50873 (The fix to PR50730 causes gcc.c-torture/unsorted/dilayout.c to ICE)

gcc/
PR middle-end/50873
* optabs.c (maybe_legitimize_operand_same_code): Use copy_to_mode_reg
instead of force_reg.  Do nothing if the address is already a
non-virtual pseudo register.

From-SVN: r182244

[multiple changes]

2011-12-12  Robert Dewar  <dewar@adacore.com>

* s-taprop-mingw.adb: Minor reformatting.

2011-12-12  Nicolas Roche  <roche@adacore.com>

* gcc-interface/Make-lang.in: Update dependencies.
Add gnattools4 target when building tools in canadian mode

2011-12-12  Javier Miranda  <miranda@adacore.com>

* exp_disp.adb (Make_VM_TSD): Complete previous
patch.

2011-12-12  Bob Duff  <duff@adacore.com>

* s-tpobop.adb: Use named notation.
* g-bytswa.ads, s-taprop.ads: Minor comment fix.

2011-12-12  Arnaud Charlet  <charlet@adacore.com>

* make.adb: Minor comment update.

From-SVN: r182242

freeze.adb, [...]: Minor reformatting.

2011-12-12  Robert Dewar  <dewar@adacore.com>

* freeze.adb, s-taprop-mingw.adb, exp_disp.adb, sem_ch13.adb: Minor
reformatting.

From-SVN: r182241

[multiple changes]

2011-12-12  Robert Dewar  <dewar@adacore.com>

* sem_ch6.adb, gnatls.adb: Minor reformatting.

2011-12-12  Ed Schonberg  <schonberg@adacore.com>

* freeze.adb (Freeze_Entity): If a type declared in a generic
context has predicates, generate a freeze node for it.
* sem_ch13.adb (Check_Aspect_At_End_Of_Declarations): If expression
has not been analyzed yet, entity has not been frozen, so analyze
it now.

2011-12-12  Javier Miranda  <miranda@adacore.com>

* exp_disp.adb (Make_VM_TSD): Generate code
to store the value 'alignment in the TSD.

2011-12-12  Bob Duff  <duff@adacore.com>

* s-taprop-mingw.adb (Yield): Do not delay 1 millisecond in Annex D
mode.

From-SVN: r182240

exp_ch4.adb (Expand_N_Allocator): Update the master of an anonymous access-to-controlled type to utilize the...

2011-12-12  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_ch4.adb (Expand_N_Allocator): Update the master of an
anonymous access-to-controlled type to utilize the heterogeneous
master associated with the current unit.

From-SVN: r182239

Correct bug-number in ChangeLog.

        PR libstdc++/51135
        * libsupc++/cxxabi.h (__cxxabi_dtor_type): New type.
        (__cxa_throw): Use it for destructor-argument.
        * libsupc++/eh_throw.cc (__cxa_throw): Likewise.
        * libsupc++/unwind-cxx.h (__cxa_exception): Change type of member
        exceptionDestructor to __cxxabi_dtor_type.
        * config/os/mingw32-w64/os_defines.h (_GLIBCXX_USE_THISCALL_ON_DTOR):
        Define.
        (__cxa_dtor_type): Declare target secific type variant.
        * config/os/mingw32/os_defines.h: Likewise.

From-SVN: r182238

PR libstdc++/511135
* libsupc++/cxxabi.h (__cxxabi_dtor_type): New type.
(__cxa_throw): Use it for destructor-argument.
* libsupc++/eh_throw.cc (__cxa_throw): Likewise.
* libsupc++/unwind-cxx.h (__cxa_exception): Change type of member
exceptionDestructor to __cxxabi_dtor_type.
* config/os/mingw32-w64/os_defines.h (_GLIBCXX_USE_THISCALL_ON_DTOR):
Define.
(__cxa_dtor_type): Declare target secific type variant.
* config/os/mingw32/os_defines.h: Likewise.

From-SVN: r182237

[multiple changes]

2011-12-12  Ed Schonberg  <schonberg@adacore.com>

* sem_ch6.adb (Analyze_Expression_Function): If the function
is not a completion, pre-analyze the expression now to prevent
spurious visibility on later entities. The body is inserted at
the end of the current declaration list or package to prevent
early freezing, but the visibility is established at the point
of definition.

2011-12-12  Bob Duff  <duff@adacore.com>

* sem.adb, sem.ads: Add debugging routines.

2011-12-12  Tristan Gingold  <gingold@adacore.com>

* gnatls.adb: (gnatls): Also add the objects dir in search list.

From-SVN: r182235

Support noexcept-specifications for transaction statements and expressions.

gcc/cp/
* semantics.c (finish_transaction_stmt, build_transaction_expr):
Accept new noexcept parameter and handle it.
* cp-tree.h (finish_transaction_stmt, build_transaction_expr): Adapt
declarations.
* parser.c (cp_parser_exception_specification_opt): Extract
noexcept-specification parsing to ...
(cp_parser_noexcept_specification_opt): ...here.  Allow for parsing
non-constexpr noexcept arguments.
(cp_parser_transaction, cp_parser_transaction_expression): Parse
and handle noexcept-specifications.
(cp_parser_function_transaction): Adapt to finish_transaction_stmt
change.
* pt.c (tsubst_expr): Adapt to new noexcept parameters when
building transactions.

gcc/testsuite/
* g++.dg/tm/noexcept-1.C: New test.
* g++.dg/tm/noexcept-2.C: New test.
* g++.dg/tm/noexcept-3.C: New test.
* g++.dg/tm/noexcept-4.C: New test.
* g++.dg/tm/noexcept-5.C: New test.

From-SVN: r182234

Add condition to MUST_NOT_THROW_EXPR.

gcc/cp/
* cp-tree.def (MUST_NOT_THROW_EXPR): Add condition parameter.
* cp-tree.h (MUST_NOT_THROW_COND): New.
(build_must_not_throw_expr): Declare.
* dump.c (cp_dump_tree): Dump MUST_NOT_THROW_EXPR condition.
* except.c (build_must_not_throw_expr): New.
(initialize_handler_parm): Use it.
(begin_eh_spec_block, wrap_cleanups_r): Adapt to condition.
* pt.c (tsubst_expr): Handle MUST_NOT_THROW_EXPR.

From-SVN: r182233

Add default handling for outermost wrappers in voidify_wrapper_expr.

gcc/
* gimplify.c (voidify_wrapper_expr): Add default handling for
outermost wrapper.

From-SVN: r182232

Prevent sharing of commit calls among transactions.

gcc/
* tree-ssa-tail-merge.c (gimple_equal_p): Don't treat transaction
commits as equal.

gcc/testsuite/
* c-c++-common/tm/20111206.c: New test.

From-SVN: r182231

[multiple changes]

2011-12-12  Robert Dewar  <dewar@adacore.com>

* exp_atag.adb, exp_atag.ads, exp_util.adb, exp_attr.adb,
sem_ch13.adb: Minor reformatting.

2011-12-12  Gary Dismukes  <dismukes@adacore.com>

* sem_ch7.adb (Uninstall_Declarations): Don't
apply check for incomplete types used as a result type for an
access-to-function type when compiling for Ada 2012 or later.
* sem_ch6.adb (Analyze_Subprogram_Declaration):
Specialize error message for interface subprograms that are
not declared abstract nor null (functions can't be declared as
null). Also, remove "(Ada 2005)" from message.

From-SVN: r182230

[multiple changes]

2011-12-12  Robert Dewar  <dewar@adacore.com>

* sem_prag.adb (GNAT_Pragma): Check comes from source.

2011-12-12  Robert Dewar  <dewar@adacore.com>

* gnatls.adb: Minor reformatting.

2011-12-12  Javier Miranda  <miranda@adacore.com>

* a-tags.ads (Alignment): New TSD field.
(Max_Predef_Prims): Value lowered to 15 (or 9 in case of
configurable runtime) Update documentation of predefined
primitives since Alignment has been removed.
* exp_disp.ads Update documentation of slots of dispatching
primitives.
* exp_disp.adb (Default_Prim_Op_Position): Update slot
values since alignment is no longer a predefined primitive.
(Is_Predefined_Dispatch_Operation): Remove _alignment.
(Is_Predefined_Internal_Operation): Remove _alignment.
(Make_DT): Update static test on the value stored in a-tags.ads
for Max_Predef_Prims; store the value of 'alignment in the TSD.
* exp_atag.ads, exp_atag.adb (Build_Get_Alignment): New subprogram
that retrieves the alignment from the TSD
* exp_util.adb (Build_Allocated_Deallocate_Proc): For deallocation
of class-wide types obtain the value of alignment from the TSD.
* exp_attr.adb (Expand_N_Attribute_Reference): For 'alignment
applied to a class-wide type invoke Build_Get_Alignment to
generate code which retrieves the value of the alignment from
the TSD.
* rtsfind.ads (RE_Alignment): New Ada.Tags entity
* sem_ch13.adb (Analyze_Attribute_Definition_Clause): For tagged
types if the value of the alignment is bigger than the Maximum
alignment then set the value of the alignment to the Maximum
alignment and report a warning.
* exp_ch3.adb (Make_Predefined_Primitive_Specs): Do not generate
spec of _alignment.
(Predefined_Primitive_Bodies): Do not generate body of _alignment.

From-SVN: r182229

[multiple changes]

2011-12-12  Gary Dismukes  <dismukes@adacore.com>

* freeze.adb (Freeze_Expression): Allow freezing of static
scalar subtypes that are prefixes of an attribute, even if not
yet marked static. Such attributes will get marked as static
later in Eval_Attribute (as called from Resolve_Attribute).
* sem_attr.adb (Eval_Attribute): Remove wrong code that does an
early return for attribute prefixes that are unfrozen source-level
types. This code was incorrectly bypassing folding of unfrozen
static subtype attributes in default expressions (the executable
example in the now-deleted comment was in fact illegal).

2011-12-12  Robert Dewar  <dewar@adacore.com>

* a-coinve.adb, sem_res.adb, prj-nmsc.adb, a-cobove.adb, a-convec.adb,
gnatls.adb, sem_ch13.adb, prj-env.adb, prj-env.ads: Minor reformatting.

2011-12-12  Tristan Gingold  <gingold@adacore.com>

* gsocket.h: Adjust previous patch.

From-SVN: r182228

[multiple changes]

2011-12-12  Thomas Quinot  <quinot@adacore.com>

* exp_disp.adb: Minor reformatting.

2011-12-12  Tristan Gingold  <gingold@adacore.com>

* gnatls.adb (Search_RTS): New procedure.
(Scan_Ls_Arg): Move code that search the RTS.
(Gnatls): search the RTS later.
* prj-env.ads, prj-env.adb (Get_Runtime_Path): New function.

2011-12-12  Ed Falis  <falis@adacore.com>

* sysdep.c: Fix treatment of VxWorks task options so that run-times
built with __SPE__ get option VX_SPE_TASK while others get VX_FP_TASK.

2011-12-12  Bob Duff  <duff@adacore.com>

* sem_type.adb, sem_type.ads, sem_ch4.adb, treepr.adb, treepr.ads:
Minor cleanup and fiddling with debug printouts.

2011-12-12  Vincent Celier  <celier@adacore.com>

* prj-nmsc.adb (Get_Directories): For a non extending project,
always get a declared object and/or exec directory if it already
exists, even when there are no sources, but do not create them.

2011-12-12  Bob Duff  <duff@adacore.com>

* sem_res.adb (Resolve): Deal with the case where an abstract
operator is called with operands of type universal_integer.

2011-12-12  Thomas Quinot  <quinot@adacore.com>

* par_sco.adb: Minor fix to dominance marker referencing WHILE
decision.

From-SVN: r182227

[multiple changes]

2011-12-12  Tristan Gingold  <gingold@adacore.com>

* mlib-tgt-specific-xi.adb: (Get_Target_Prefix): Simplify code.

2011-12-12  Thomas Quinot  <quinot@adacore.com>

* par_sco.adb: Adjust dominant marker for branches of CASE
statements.

2011-12-12  Thomas Quinot  <quinot@adacore.com>

* gsocket.h, s-oscons-tmplt.c: Ensure we do not include any system
header file prior to redefining FD_SETSIZE.

2011-12-12  Ed Schonberg  <schonberg@adacore.com>

* sem_ch13.adb (Check_Aspect_At_End_Of_Declarations): In
a generic context the aspect expressions may not have been
preanalyzed if there was no previous freeze point, so the
expressions must be preanalyzed now, and there is no conformance
to check for visibility changes.

2011-12-12  Matthew Heaney  <heaney@adacore.com>

* a-convec.adb, a-coinve.adb, a-cobove.adb (Iterator): Use
subtype Index_Type'Base for Index component (Finalize): Remove
unnecessary access check (First, Last): Cursor return value
depends on iterator index value (Iterate): Use start position as
iterator index value (Next, Previous): Forward to corresponding
cursor-based operation.
* a-cborma.adb (Iterate): Properly initialize iterator object (with 0
as node index).

From-SVN: r182226

re PR libgcj/50053 (jc1 doesn't emit static initializer or initializer code for super class without -findirect-dispatch.)

        PR libgcj/50053
        * java/lang/natClass.cc (java::lang::Class::newInstance): Special case
        member-call for 32-bit IA native Window target.

From-SVN: r182225

[multiple changes]

2011-12-12  Robert Dewar  <dewar@adacore.com>

* par_sco.adb, scos.ads, put_scos.adb, get_scos.adb: Minor reformatting.

2011-12-12  Steve Baird  <baird@adacore.com>

* sem_ch3.adb (Check_Completion): Improve a comment.

2011-12-12  Tristan Gingold  <gingold@adacore.com>

* cstand.adb: Minor comment fix.

From-SVN: r182224

re PR debug/51262 (ICE: SIGSEGV in primary_template_instantiation_p (pt.c:2874) with -flto -g)

2011-12-12  Richard Guenther  <rguenther@suse.de>

PR lto/51262
* tree.c (cp_free_lang_data): No longer clear anonymous names.

* g++.dg/opt/pr51262.C: New testcase.

From-SVN: r182221

darwin-sections.def (zobj_const_data_section): Fix over- length section name.

gcc:

* config/darwin-sections.def (zobj_const_data_section): Fix over-
length section name.

From-SVN: r182220

Fix PR testsuite/47013

From-SVN: r182219

rs6000.c (rs6000_expand_interleave): Use BYTES_BIG_ENDIAN, not TARGET_BIG_ENDIAN.

        * config/rs6000/rs6000.c (rs6000_expand_interleave): Use
        BYTES_BIG_ENDIAN, not TARGET_BIG_ENDIAN.

From-SVN: r182218

Daily bump.

From-SVN: r182217

re PR tree-optimization/51485 (ICE in vectorizable_assignment)

PR tree-optimization/51485
* tree-vect-data-refs.c (vect_analyze_data_refs): Give up on
DRs in call stmts.

* g++.dg/vect/pr51485.cc: New test.

From-SVN: r182212

re PR fortran/50923 (No warning if function return value is not set)

2011-12-11  Tobias Burnus  <burnus@net-b.de>

        PR fortran/50923
        * trans-decl.c (generate_local_decl): Set TREE_NO_WARNING only
        if the front end has printed a warning.
        (gfc_generate_function_code): Fix unset-result warning.

2011-12-11  Tobias Burnus  <burnus@net-b.de>

        PR fortran/50923
        * gfortran.dg/warn_function_without_result_2.f90: New.

From-SVN: r182211

re PR fortran/41539 ([OOP] Calling function which takes CLASS: Rank comparison does not work)

2011-12-11  Paul Thomas  <pault@gcc.gnu.org>
Tobias Burnus  <burnus@gcc.gnu.org>

PR fortran/41539
PR fortran/43214
PR fortran/43969
PR fortran/44568
PR fortran/46356
PR fortran/46990
PR fortran/49074
* interface.c(symbol_rank): Return the rank of the _data
component of class objects.
(compare_parameter): Also compare the derived type of the class
_data component for type mismatch.  Similarly, return 1 if the
formal and _data ranks match.
(compare_actual_formal): Do not compare storage sizes for class
expressions. It is an error if an actual class array, passed to
a formal class array is not full.
* trans-expr.c (gfc_class_data_get, gfc_class_vptr_get,
gfc_vtable_field_get, gfc_vtable_hash_get, gfc_vtable_size_get,
gfc_vtable_extends_get, gfc_vtable_def_init_get,
gfc_vtable_copy_get): New functions for class API.
(gfc_conv_derived_to_class): For an array reference in an
elemental procedure call retain the ss to provide the
scalarized array reference. Moved in file.
(gfc_conv_class_to_class): New function.
        (gfc_conv_subref_array_arg): Use the type of the
class _data component as a basetype.
(gfc_conv_procedure_call): Ensure that class array expressions
have both the _data reference and an array reference. Use
gfc_conv_class_to_class to handle class arrays for elemental
functions in scalarized loops, class array elements and full
class arrays. Use a call to gfc_conv_subref_array_arg in order
that the copy-in/copy-out for passing class arrays to derived
type arrays occurs correctly.
(gfc_conv_expr): If it is missing, add the _data component
between a class object or component and an array reference.
(gfc_trans_class_array_init_assign): New function.
(gfc_trans_class_init_assign): Call it for array expressions.
* trans-array.c (gfc_add_loop_ss_code): Do not use a temp for
class scalars since their size will depend on the dynamic type.
(build_class_array_ref): New function.
(gfc_conv_scalarized_array_ref): Call build_class_array_ref.
(gfc_array_init_size): Add extra argument, expr3, that represents
the SOURCE argument. If present,use this for the element size.
(gfc_array_allocate): Also add argument expr3 and use it when
calling gfc_array_init_size.
(structure_alloc_comps): Enable class arrays.
* class.c (gfc_add_component_ref): Carry over the derived type
of the _data component.
(gfc_add_class_array_ref): New function.
(class_array_ref_detected): New static function.
(gfc_is_class_array_ref): New function that calls previous.
(gfc_is_class_scalar_expr): New function.
(gfc_build_class_symbol): Throw not implemented error for
assumed size class arrays.  Remove error that prevents
CLASS arrays.
(gfc_build_class_symbol): Prevent pointer/allocatable conflict.
Also unset codimension.
(gfc_find_derived_vtab): Make 'copy' elemental and set the
intent of the arguments accordingly.:
* trans-array.h : Update prototype for gfc_array_allocate.
* array.c (gfc_array_dimen_size): Return failure if class expr.
(gfc_array_size): Likewise.
* gfortran.h : New prototypes for gfc_add_class_array_ref,
gfc_is_class_array_ref and gfc_is_class_scalar_expr.
* trans-stmt.c (trans_associate_var): Exclude class targets
from test. Move the allocation of the _vptr to an earlier time
for class objects.
(trans_associate_var): Assign the descriptor directly for class
arrays.
(gfc_trans_allocate): Add expr3 to gfc_array_allocate arguments.
Convert array element references into sections. Do not invoke
gfc_conv_procedure_call, use gfc_trans_call instead.
* expr.c (gfc_get_corank): Fix for BT_CLASS.
(gfc_is_simply_contiguous): Exclude class from test.
* trans.c (gfc_build_array_ref): Include class array refs.
* trans.h : Include prototypes for class API functions that are
new in trans-expr. Define GFC_DECL_CLASS(node).
* resolve.c (check_typebound_baseobject ): Remove error for
non-scalar base object.
(resolve_allocate_expr): Ensure that class _data component is
present. If array, call gfc_expr_to_intialize.
(resolve_select): Remove scalar error for SELECT statement as a
temporary measure.
(resolve_assoc_var): Update 'target' (aka 'selector') as
needed. Ensure that the target expression has the right rank.
(resolve_select_type): Ensure that target expressions have a
valid locus.
(resolve_allocate_expr, resolve_fl_derived0): Fix for BT_CLASS.
* trans-decl.c (gfc_get_symbol_decl): Set GFC_DECL_CLASS, where
appropriate.
(gfc_trans_deferred_vars): Get class arrays right.
* match.c(select_type_set_tmp): Add array spec to temporary.
(gfc_match_select_type): Allow class arrays.
* check.c (array_check): Ensure that class arrays have refs.
(dim_corank_check, dim_rank_check): Retrun success if class.
* primary.c (gfc_match_varspec): Fix for class arrays and
co-arrays. Make sure that class _data is present.
(gfc_match_rvalue): Handle class arrays.
*trans-intrinsic.c (gfc_conv_intrinsic_size): Add class array
reference.
(gfc_conv_allocated): Add _data component to class expressions.
(gfc_add_intrinsic_ss_code): ditto.
* simplify.c (simplify_cobound): Fix for BT_CLASS.
(simplify_bound): Return NULL for class arrays.
(simplify_cobound): Obtain correct array_spec. Use cotype as
appropriate. Use arrayspec for bounds.

2011-12-11  Paul Thomas  <pault@gcc.gnu.org>
Tobias Burnus  <burnus@gcc.gnu.org>

PR fortran/41539
PR fortran/43214
PR fortran/43969
PR fortran/44568
PR fortran/46356
PR fortran/46990
PR fortran/49074
* gfortran.dg/class_array_1.f03: New.
* gfortran.dg/class_array_2.f03: New.
* gfortran.dg/class_array_3.f03: New.
* gfortran.dg/class_array_4.f03: New.
* gfortran.dg/class_array_5.f03: New.
* gfortran.dg/class_array_6.f03: New.
* gfortran.dg/class_array_7.f03: New.
* gfortran.dg/class_array_8.f03: New.
* gfortran.dg/coarray_poly_1.f90: New.
* gfortran.dg/coarray_poly_2.f90: New.
* gfortran.dg/coarray/poly_run_1.f90: New.
* gfortran.dg/coarray/poly_run_2.f90: New.
* gfortran.dg/class_to_type_1.f03: New.
* gfortran.dg/type_to_class_1.f03: New.
* gfortran.dg/typebound_assignment_3.f03: Remove the error.
* gfortran.dg/auto_dealloc_2.f90: Occurences of __builtin_free
now 2.
* gfortran.dg/class_19.f03: Occurences of __builtin_free now 8.

Co-Authored-By: Tobias Burnus <burnus@gcc.gnu.org>
From-SVN: r182210

re PR fortran/50690 (ICE with front end optimization and OMP workshare)

2011-12-11  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/50690
* frontend-passes.c (in_omp_workshare):  New variable.
(cfe_expr_0):  Don't eliminiate common function if it would put
the variable immediately into a WORKSHARE construct.
(optimize_namespace):  Set in_omp_workshare.
(gfc_code_walker):  Keep track of OMP PARALLEL and OMP WORKSHARE
constructs.

2011-12-11  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/50690
* gfortran.dg/gomp/workshare2.f90:  New test.
* gfortran.dg/gomp/workshare3.f90:  New test.

From-SVN: r182208

abi.xml: Replace gcc-x.y.z with GCC x.y.z or x.y...

* doc/xml/manual/abi.xml: Replace gcc-x.y.z with GCC x.y.z or x.y,
remove excessive duplication of version information.
* doc/html/*: Regenerate.

From-SVN: r182206

trans-mem.c (ipa_tm_transform_calls_redirect): Do not remove edge.

        * trans-mem.c (ipa_tm_transform_calls_redirect): Do not remove
        edge.

From-SVN: r182205

* config/sparc/sol2-unwind.h: Use #ifdef directive consistently.

From-SVN: r182204

utils2.c (gnat_build_constructor): Test the TREE_STATIC flag of elements to compute that of the constructor.

* gcc-interface/utils2.c (gnat_build_constructor): Test the TREE_STATIC
flag of elements to compute that of the constructor.

From-SVN: r182203

decl.c (gnat_to_gnu_param): Set the restrict qualifier on references built for parameters which aren't...

* gcc-interface/decl.c (gnat_to_gnu_param): Set the restrict qualifier
on references built for parameters which aren't specifically by-ref.

From-SVN: r182202

decl.c (gnat_to_gnu_entity): If there is an alignment set on a renaming...

* gcc-interface/decl.c (gnat_to_gnu_entity) <object>: If there is an
alignment set on a renaming, assert that the renamed object is aligned
enough as to make it possible to honor it.

From-SVN: r182201