gcc.git
14 years agoUse sizetype instead of unsigned_type_node.
Sebastian Pop [Mon, 8 Mar 2010 17:50:03 +0000 (17:50 +0000)]
Use sizetype instead of unsigned_type_node.

2010-03-05  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-sese-to-poly.c (add_param_constraints): Use sizetype
instead of unsigned_type_node.

From-SVN: r157290

14 years agoFix PR43065: Insert bounds on pointer type parameters.
Sebastian Pop [Mon, 8 Mar 2010 17:49:57 +0000 (17:49 +0000)]
Fix PR43065: Insert bounds on pointer type parameters.

2010-03-05  Sebastian Pop  <sebastian.pop@amd.com>
    Reza Yazdani  <reza.yazdani@amd.com>

PR middle-end/43065
* graphite-sese-to-poly.c (add_param_constraints): Insert bounds
on pointer type parameters.

* gcc.dg/graphite/run-id-4.c: New.

Co-Authored-By: Reza Yazdani <reza.yazdani@amd.com>
From-SVN: r157289

14 years agoAdd testcase from PR43065.
Sebastian Pop [Mon, 8 Mar 2010 17:49:48 +0000 (17:49 +0000)]
Add testcase from PR43065.

2010-03-04  Sebastian Pop  <sebastian.pop@amd.com>

PR middle-end/43065
* gcc.dg/graphite/run-id-3.c: New.

From-SVN: r157288

14 years agoAdd forgotten ChangeLog entries.
Tobias Grosser [Mon, 8 Mar 2010 17:49:42 +0000 (17:49 +0000)]
Add forgotten ChangeLog entries.

Eric pointed out that my ChangeLog was incorrect. As I forgot to
commit it at all, here the corrected ChangeLogs for the last commit.

gcc/
2010-03-05  Tobias Grosser  <grosser@fim.uni-passau.de>
    Ramakrishna Upadrasta <Ramakrishna.Upadrasta@inria.fr>

* graphite-clast-to-gimple.c (clast_to_gcc_expression): Also
handle conversions from pointer to integers.
(gcc_type_for_cloog_iv): Choose the smalles signed integer as an
induction variable, to be able to work with code generated by
CLooG.
* graphite-sese-to-poly.c (scop_ivs_can_be_represented): New.
(build_poly_scop): Bail out if we cannot codegen a loop.

gcc/testsuite/
2010-03-05  Tobias Grosser  <grosser@fim.uni-passau.de>

* gcc.dg/graphite/id-18.c: New.
* gcc.dg/graphite/run-id-pr42644.c: New.

libgomp/
2010-03-05  Tobias Grosser  <grosser@fim.uni-passau.de>

* testsuite/libgomp.graphite/force-parallel-1.c: Adjust.
* testsuite/libgomp.graphite/force-parallel-2.c: Adjust.

From-SVN: r157287

14 years agoFix type problems in loop ivs.
Tobias Grosser [Mon, 8 Mar 2010 17:49:36 +0000 (17:49 +0000)]
Fix type problems in loop ivs.

Fix pr42644.
Fix pr42130 (dealII).

2010-03-03  Tobias Grosser  <grosser@fim.uni-passau.de>

* gcc/graphite-clast-to-gimple.c (clast_to_gcc_expression): Also
handle conversions from pointer to integers.
(gcc_type_for_cloog_iv): Choose the smalles signed integer as an
induction variable, to be able to work with code generated by
CLooG.
* gcc/graphite-sese-to-poly.c (scop_ivs_can_be_represented): New.
(build_poly_scop): Bail out if we cannot codegen a loop.
* gcc/testsuite/gcc.dg/graphite/id-18.c: New.
* gcc/testsuite/gcc.dg/graphite/run-id-pr42644.c: New.
* libgomp/testsuite/libgomp.graphite/force-parallel-1.c: Adjust.
* libgomp/testsuite/libgomp.graphite/force-parallel-2.c: Adjust.

From-SVN: r157286

14 years agoDo not short-cut code generation with gloog_error.
Tobias Grosser [Mon, 8 Mar 2010 17:49:28 +0000 (17:49 +0000)]
Do not short-cut code generation with gloog_error.

2010-03-04  Tobias Grosser  <grosser@fim.uni-passau.de>

* graphite-clast-to-gimple.c (translate_clast): Do not short-cut
code generation with gloog_error.

From-SVN: r157285

14 years agoFold convert SSA_NAMEs in the scalar expander.
Sebastian Pop [Mon, 8 Mar 2010 17:49:22 +0000 (17:49 +0000)]
Fold convert SSA_NAMEs in the scalar expander.

2010-03-04  Sebastian Pop  <sebastian.pop@amd.com>

* sese.c (expand_scalar_variables_ssa_name): Add new argument for type.
Call fold_convert on all the returned values.
(expand_scalar_variables_expr): Pass to expand_scalar_variables_ssa_name
the type of the resulting expression.

From-SVN: r157284

14 years agographite-ppl.c (ppl_min_for_le_polyhedron): Renamed ppl_min_for_le_pointset.
Sebastian Pop [Mon, 8 Mar 2010 17:49:16 +0000 (17:49 +0000)]
graphite-ppl.c (ppl_min_for_le_polyhedron): Renamed ppl_min_for_le_pointset.

Cleanup.

2010-03-04  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-ppl.c (ppl_min_for_le_polyhedron): Renamed
ppl_min_for_le_pointset.  Use ppl_Pointset_Powerset_C_Polyhedron_minimize.
* graphite-ppl.h (ppl_min_for_le_polyhedron): Update declaration.

From-SVN: r157283

14 years agoCleanup data dep polyhedron construction.
Sebastian Pop [Mon, 8 Mar 2010 17:49:09 +0000 (17:49 +0000)]
Cleanup data dep polyhedron construction.

2010-03-04  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-dependences.c (map_into_dep_poly): Removed.
(dependence_polyhedron_1): Use combine_context_id_scat.

From-SVN: r157282

14 years agoNew function combine_context_id_scat.
Sebastian Pop [Mon, 8 Mar 2010 17:49:02 +0000 (17:49 +0000)]
New function combine_context_id_scat.

2010-03-04  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-poly.h (struct poly_scattering): Add layout documentation.
(struct poly_bb): Same.
(combine_context_id_scat): New.

From-SVN: r157281

14 years agoFix PR42326: handle default definitions.
Sebastian Pop [Mon, 8 Mar 2010 17:48:55 +0000 (17:48 +0000)]
Fix PR42326: handle default definitions.

2010-03-02  Sebastian Pop  <sebastian.pop@amd.com>

PR middle-end/42326
* sese.c (name_defined_in_loop_p): Return false for default
definitions.

* gcc.dg/graphite/pr42326.c: New.

From-SVN: r157280

14 years agographite-clast-to-gimple.c (find_cloog_iv_in_expr): Simplify and clean up the logic.
Sebastian Pop [Mon, 8 Mar 2010 17:48:48 +0000 (17:48 +0000)]
graphite-clast-to-gimple.c (find_cloog_iv_in_expr): Simplify and clean up the logic.

Cleanup.

2010-02-23  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-clast-to-gimple.c (find_cloog_iv_in_expr): Simplify
and clean up the logic.

From-SVN: r157279

14 years agoAdd constraints on the type of parameters to the scop context.
Sebastian Pop [Mon, 8 Mar 2010 17:48:42 +0000 (17:48 +0000)]
Add constraints on the type of parameters to the scop context.

2010-02-23  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-sese-to-poly.c (add_param_constraints): Enabled: remove
early return.

From-SVN: r157278

14 years agovar-tracking.c (remove_cselib_value_chains): Define only for ENABLE_CHECKING.
Jakub Jelinek [Mon, 8 Mar 2010 13:33:31 +0000 (14:33 +0100)]
var-tracking.c (remove_cselib_value_chains): Define only for ENABLE_CHECKING.

* var-tracking.c (remove_cselib_value_chains): Define only for
ENABLE_CHECKING.
(dataflow_set_preserve_mem_locs, dataflow_set_remove_mem_locs,
delete_slot_part, emit_notes_for_differences_1): Don't call
remove_cselib_value_chains here.
(set_slot_part, emit_notes_for_differences_2): Don't call
add_cselib_value_chains here.
(preserved_values): New vector.
(preserve_value): New function.
(add_uses, add_stores, vt_add_function_parameters): Use it
instead of cselib_preserve_value.
(changed_values_stack): New vector.
(check_changed_vars_0): New function.
(check_changed_vars_1, check_changed_vars_2): Use it.
(emit_notes_for_changes): Call set_dv_changed (*, false) on all
changed_values_stack VALUEs.
(vt_emit_notes): For all preserved_values call
add_cselib_value_chains.  If ENABLE_CHECKING call
remove_cselib_value_chains before verifying value_chains is empty.
Initialize and free changed_values_stack.
(vt_initialize): Initialize preserved_values.
(vt_finalize): Free preserved_values.

From-SVN: r157277

14 years agore PR tree-optimization/43269 (removing non dead store)
Richard Guenther [Mon, 8 Mar 2010 13:30:27 +0000 (13:30 +0000)]
re PR tree-optimization/43269 (removing non dead store)

2010-03-08  Richard Guenther  <rguenther@suse.de>

PR tree-optimization/43269
* tree-ssa-dse.c (dse_possible_dead_store_p): Fix post-dom
region detection.

* gcc.c-torture/execute/pr43269.c: New testcase.

From-SVN: r157276

14 years agoipa-prop.h (struct ipa_param_descriptor): Removed the called field.
Martin Jambor [Mon, 8 Mar 2010 11:00:07 +0000 (12:00 +0100)]
ipa-prop.h (struct ipa_param_descriptor): Removed the called field.

2010-03-08  Martin Jambor  <mjambor@suse.cz>

* ipa-prop.h (struct ipa_param_descriptor): Removed the called field.
(ipa_is_param_called): Removed.
* ipa-prop.c (ipa_note_param_call): Do not set the called flag.
(ipa_print_node_params): Do not print the called flag.
(ipa_write_node_info): Do not stream the called flag.
(ipa_read_node_info): Likewise.

From-SVN: r157273

14 years agore PR fortran/43256 ([OOP] TBP with missing optional arg)
Janus Weil [Mon, 8 Mar 2010 09:35:04 +0000 (10:35 +0100)]
re PR fortran/43256 ([OOP] TBP with missing optional arg)

2010-03-08  Janus Weil  <janus@gcc.gnu.org>

PR fortran/43256
* resolve.c (resolve_compcall): Don't set 'value.function.name' here
for TBPs, otherwise they will not be resolved properly.
(resolve_function): Use 'value.function.esym' instead of
        'value.function.name' to check if we're dealing with a TBP.
(check_class_members): Set correct type of passed object for all TBPs,
not only generic ones, except if the type is abstract.

2010-03-08  Janus Weil  <janus@gcc.gnu.org>

PR fortran/43256
* gfortran.dg/typebound_call_13.f03: New.

From-SVN: r157272

14 years agolibgfortran.h (_POSIX): Define if __MINGW32__ is defined.
Kai Tietz [Mon, 8 Mar 2010 08:01:55 +0000 (09:01 +0100)]
libgfortran.h (_POSIX): Define if __MINGW32__ is defined.

2010-03-08  Kai TIetz  <kai.tietz@onevision.com>

        PR/42950
        * libgfortran.h (_POSIX): Define if __MINGW32__ is defined.
        (gfc_printf): Define to gnu_printf for __MINGW32__ case,
        otherwise to __printf__.
        (gfc_strtof,gfc_strtod,gfc_strtold): Define for mingw case
        to POSIX compatible converter functions.
        (runtime_error): Use instead gfc_printf as formatter
        attribute name.
        (runtime_error_at): Likewise.
        (runtime_warning_at): Likewise.
        (st_printf): Likewise.
        * intrinsics/date_and_time.c (localtime_r): Undefine
        possible defined macro.
        (gmtime_r): Likewise.
        * io/read.c (convert_real): Use gfc_strtof, gfc_strtod,
        and gfc_strtold.

From-SVN: r157271

14 years agoDaily bump.
GCC Administrator [Mon, 8 Mar 2010 00:16:32 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r157269

14 years agore PR debug/43176 (var-tracking fails to notice a value change)
Jakub Jelinek [Sun, 7 Mar 2010 15:44:11 +0000 (16:44 +0100)]
re PR debug/43176 (var-tracking fails to notice a value change)

PR debug/43176
* Makefile.in (var-tracking.o): Depend on pointer-set.h.
* cselib.c (struct expand_value_data): Add dummy field.
(cselib_expand_value_rtx, cselib_expand_value_rtx_cb): Initialize
dummy to false.
(cselib_dummy_expand_value_rtx_cb): New function.
(cselib_expand_value_rtx_1): If evd->dummy is true, don't allocate
any rtl.
* cselib.h (cselib_dummy_expand_value_rtx_cb): New prototype.
* var-tracking.c: Include pointer-set.h.
(variable): Change n_var_parts to char from int.  Add
cur_loc_changed and in_changed_variables fields.
(variable_canonicalize): Remove.
(shared_var_p): New inline function.
(unshare_variable): Maintain cur_loc_changed and
in_changed_variables fields.  If var was in changed_variables,
replace it there with new_var.  Just copy cur_loc instead of
resetting it to something else.
(variable_union): Don't recompute cur_loc.  Use shared_var_p.
(dataflow_set_union): Don't call variable_canonicalize.
(loc_cmp): If both x and y are DEBUG_EXPRs, compare uids
of their DEBUG_EXPR_TREE_DECLs.
(canonicalize_loc_order_check): Verify that cur_loc is NULL
and in_changed_variables and cur_loc_changed is false.
(variable_merge_over_cur): Clear cur_loc, in_changed_variables
and cur_loc_changed.  Don't update cur_loc here.
(variable_merge_over_src): Don't call variable_canonicalize.
(dataflow_set_preserve_mem_locs): Use shared_var_p.  When
removing loc that is equal to cur_loc, clear cur_loc,
set cur_loc_changed and ensure variable_was_changed is called.
(dataflow_set_remove_mem_locs): Use shared_var_p.  Only
compare pointers in cur_loc check, if it is equal to loc,
clear cur_loc and set cur_loc_changed.  Don't recompute cur_loc here.
(variable_different_p): Remove compare_current_location argument,
don't compare cur_loc.
(dataflow_set_different_1): Adjust variable_different_p caller.
(variable_was_changed): If dv had some var in changed_variables
already, reset in_changed_variables flag for it and propagate
cur_loc_changed over to the new variable.  On empty var
always set cur_loc_changed.  Set in_changed_variables on whatever
var is added to changed_variables.
(set_slot_part): Clear cur_loc_changed and in_changed_variables.
Use shared_var_p.  When removing loc that is equal to cur_loc,
clear cur_loc and set cur_loc_changed.  If cur_loc is NULL at the
end, don't set it to something else, just call variable_was_changed.
(delete_slot_part): Use shared_var_p.  When cur_loc equals to
loc being removed, clear cur_loc and set cur_loc_changed.
Set cur_loc_changed if all locations have been removed.
(struct expand_loc_callback_data): New type.
(vt_expand_loc_callback): Add dummy mode in which no rtxes are
allocated.  Always create SUBREGs if simplify_subreg failed.
Prefer to use cur_loc, when that fails and still in
changed_variables (and seen first time) recompute it.  Set
cur_loc_changed of variables which had to change cur_loc and
compute elcd->cur_loc_changed if any of the subexpressions used
had to change cur_loc.
(vt_expand_loc): Adjust to pass arguments in
expand_loc_callback_data structure.
(vt_expand_loc_dummy): New function.
(emitted_notes): New variable.
(emit_note_insn_var_location): For VALUEs and DEBUG_EXPR_DECLs
that weren't used for any other decl in current
emit_notes_for_changes call call vt_expand_loc_dummy to update
cur_loc.  For -fno-var-tracking-assignments, set cur_loc to
first loc_chain location if NULL before.  Always use just
cur_loc instead of first loc_chain location.  When cur_loc_changed
is false, when not --enable-checking=rtl just don't emit any note.
When rtl checking, compute the note and assert it is the same
as previous note.  Clear cur_loc_changed and in_changed_variables
at the end before removing from changed_variables.
(check_changed_vars_3): New function.
(emit_notes_for_changes): Traverse changed_vars to call
check_changed_vars_3 on each changed var.
(emit_notes_for_differences_1): Clear cur_loc_changed and
in_changed_variables.  Recompute cur_loc of new_var.
(emit_notes_for_differences_2): Clear cur_loc if new variable
appears.
(vt_emit_notes): Initialize and destroy emitted_notes.

From-SVN: r157264

14 years agore PR middle-end/42220 (FAIL: gfortran.dg/complex_intrinsic_5.f90 -m64 -O -frename...
Bernd Schmidt [Sun, 7 Mar 2010 15:20:12 +0000 (15:20 +0000)]
re PR middle-end/42220 (FAIL: gfortran.dg/complex_intrinsic_5.f90  -m64 -O -frename-registers)

PR rtl-optimization/42220
* regrename.c (scan_rtx) <case STRICT_LOW_PART, ZERO_EXTRACT>:
Use verify_reg_tracked to determine if we should use OP_OUT rather
than OP_INOUT.
(build_def_use): If we see an in-out operand for a register that we
know nothing about, treat is an output if possible, fail the block if
not.

From-SVN: r157263

14 years agoDaily bump.
GCC Administrator [Sun, 7 Mar 2010 00:16:38 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r157262

14 years agore PR debug/42897 (yet another ice in verify_ssa)
Alexandre Oliva [Sat, 6 Mar 2010 20:30:10 +0000 (20:30 +0000)]
re PR debug/42897 (yet another ice in verify_ssa)

PR debug/42897
* gimple-iterator.c (gsi_remove): Propagate only PHI DEFs removed
permanently.

From-SVN: r157259

14 years agore PR debug/42897 (yet another ice in verify_ssa)
Alexandre Oliva [Sat, 6 Mar 2010 20:28:04 +0000 (20:28 +0000)]
re PR debug/42897 (yet another ice in verify_ssa)

PR debug/42897
* tree-vect-loop.c (vect_transform_loop): Kill out-of-loop debug
uses of relevant DEFs that are dead outside the loop too.

From-SVN: r157258

14 years agore PR debug/41371 (var-tracking is slow and memory hungry)
Alexandre Oliva [Sat, 6 Mar 2010 20:26:15 +0000 (20:26 +0000)]
re PR debug/41371 (var-tracking is slow and memory hungry)

* var-tracking.c (dataflow_set_merge): Swap src and src2.
Reverted:
2010-01-13  Jakub Jelinek  <jakub@redhat.com>
PR debug/41371
* var-tracking.c (values_to_unmark): New variable.
(find_loc_in_1pdv): Clear VALUE_RECURSED_INTO of values in
values_to_unmark vector.  Moved body to...
(find_loc_in_1pdv_1): ... this.  Don't clear VALUE_RECURSED_INTO,
instead queue it into values_to_unmark vector.
(vt_find_locations): Free values_to_unmark vector.

From-SVN: r157257

14 years agoDaily bump.
GCC Administrator [Sat, 6 Mar 2010 00:16:57 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r157252

14 years agoMakefile.in (PLUGINCC, [...]): New variables.
Eric Botcazou [Fri, 5 Mar 2010 22:52:02 +0000 (22:52 +0000)]
Makefile.in (PLUGINCC, [...]): New variables.

* Makefile.in (PLUGINCC, PLUGINCFLAGS): New variables.
(site.exp): Export them when plugins are enabled.
testsuite/
* lib/plugin-support.exp (plugin-test-execute): Use PLUGINCC in lieu
of HOSTCC and PLUGINCFLAGS in lieu of HOSTCFLAGS.

From-SVN: r157246

14 years agomangle.c (mangle_decl): Give name collision error even without ASM_OUTPUT_DEF.
Jason Merrill [Fri, 5 Mar 2010 22:44:56 +0000 (17:44 -0500)]
mangle.c (mangle_decl): Give name collision error even without ASM_OUTPUT_DEF.

* mangle.c (mangle_decl): Give name collision error even without
ASM_OUTPUT_DEF.
* g++.dg/abi/mangle40.C: Require weak and alias.

From-SVN: r157245

14 years agoFix PR42326: Handle more carefully convert expressions in chrec_fold_plus and chrec_f...
Sebastian Pop [Fri, 5 Mar 2010 20:37:44 +0000 (20:37 +0000)]
Fix PR42326: Handle more carefully convert expressions in chrec_fold_plus and chrec_fold_mult.

2010-03-05  Sebastian Pop  <sebastian.pop@amd.com>

PR middle-end/42326
* tree-chrec.c (chrec_fold_plus_1): Do not handle convert expressions
that contain scevs.
(chrec_fold_multiply): Same.

* gfortran.dg/graphite/pr42326.f90: New.
* gfortran.dg/graphite/pr42326-1.f90: New.

From-SVN: r157244

14 years agognat.exp (gnat_init): Remove GNAT_UNDER_TEST_ORIG.
Rainer Orth [Fri, 5 Mar 2010 11:39:17 +0000 (11:39 +0000)]
gnat.exp (gnat_init): Remove GNAT_UNDER_TEST_ORIG.

* lib/gnat.exp (gnat_init): Remove GNAT_UNDER_TEST_ORIG.
(gnat_target_compile): Likewise.
Reinitialize GNAT_UNDER_TEST if target changes.
Set ADA_INCLUDE_PATH, ADA_OBJECTS_PATH in environment.
(local_find_gnatmake): Pass full --GCC to gnatlink.
Remove --LINK.

From-SVN: r157240

14 years agotestsuite_container_traits.h (traits_base): Add has_throwing_erase trait.
Paolo Carlini [Fri, 5 Mar 2010 01:51:56 +0000 (01:51 +0000)]
testsuite_container_traits.h (traits_base): Add has_throwing_erase trait.

2010-03-04  Paolo Carlini  <paolo.carlini@oracle.com>

* testsuite/util/testsuite_container_traits.h (traits_base): Add
has_throwing_erase trait.
(traits<vector>, traits<deque>): Typedef the latter to true_type.
* testsuite/util/exception/safety.h (generation_prohibited):
Do not test vector::erase and deque::erase: can throw if
either copy constructor or assignment operator of value_type
throws.
* testsuite/23_containers/vector/requirements/exception/
generation_prohibited.cc: Remove xfail.
* testsuite/23_containers/deque/requirements/exception/
generation_prohibited.cc: Likewise.

* include/ext/throw_allocator.h (hash<__gnu_cxx::throw_value_limit>::
operator(), hash<__gnu_cxx::throw_value_random>::operator()): Pass
argument by const ref.

* testsuite/util/testsuite_container_traits.h (traits<map>,
traits<multimap>, traits<set>, traits<multiset>,
traits<unordered_map>, traits<unordered_multimap>,
traits<unordered_set>, traits<unordered_multiset>): Typedef
consistently has_erase and has_insert as true_type.

* testsuite/util/testsuite_container_traits.h
(traits<unordered_map>, traits<unordered_multimap>,
traits<unordered_set>, traits<unordered_multiset>): Do not wrongly
typedef has_size_type_constructor as true_type: the constructor
accepting a size_type actually gets the initial number of
buckets.

From-SVN: r157239

14 years agoDaily bump.
GCC Administrator [Fri, 5 Mar 2010 00:18:03 +0000 (00:18 +0000)]
Daily bump.

From-SVN: r157236

14 years agore PR c/43248 (ICE: tree check: expected class 'type', have 'exceptional' (error_mark...
Andrew Pinski [Thu, 4 Mar 2010 19:59:19 +0000 (11:59 -0800)]
re PR c/43248 (ICE: tree check: expected class 'type', have 'exceptional' (error_mark) in useless_type_conversion_p, at tree-ssa.c:1233)

2010-03-04  Andrew Pinski  <andrew_pinski@caviumnetworks.com>

        PR c/43248
        * c-decl.c (build_compound_literal): Return early if init is
        an error_mark_node.
2010-03-04  Andrew Pinski  <andrew_pinski@caviumnetworks.com>

        PR c/43248
        * gcc.dg/compound-literal-1.c: New testcase.

From-SVN: r157233

14 years agore PR tree-optimization/43164 (ice in completely_scalarize_record, at tree-sra.c:85)
Martin Jambor [Thu, 4 Mar 2010 18:16:32 +0000 (19:16 +0100)]
re PR tree-optimization/43164 (ice in completely_scalarize_record, at tree-sra.c:85)

2010-03-04  Martin Jambor  <mjambor@suse.cz>

PR tree-optimization/43164
PR tree-optimization/43191
* tree-sra.c (type_consists_of_records_p): Reject records with
zero-size bit-fields at the end.

* testsuite/gcc.c-torture/compile/pr43164.c: New test.
* testsuite/gcc.c-torture/compile/pr43191.c: Likewise.

From-SVN: r157232

14 years agopt.c (process_partial_specialization): Use error_n instead of error.
Marco Poletti [Thu, 4 Mar 2010 16:17:07 +0000 (16:17 +0000)]
pt.c (process_partial_specialization): Use error_n instead of error.

2010-03-04  Marco Poletti  <poletti.marco@gmail.com>

* pt.c (process_partial_specialization): Use error_n instead of
error.

From-SVN: r157229

14 years ago* Makefile.in (TAGS): Remove *.y.
Mike Stump [Thu, 4 Mar 2010 16:07:20 +0000 (16:07 +0000)]
* Makefile.in (TAGS): Remove *.y.

From-SVN: r157228

14 years agore PR fortran/43244 (Invalid statement misinterpreted as FINAL declaration)
Janus Weil [Thu, 4 Mar 2010 13:42:06 +0000 (14:42 +0100)]
re PR fortran/43244 (Invalid statement misinterpreted as FINAL declaration)

2010-03-04  Janus Weil  <janus@gcc.gnu.org>

PR fortran/43244
* decl.c (gfc_match_final_decl): Make sure variable names starting with
'final...' are not misinterpreted as FINAL statements.

2010-03-04  Janus Weil  <janus@gcc.gnu.org>

PR fortran/43244
* gfortran.dg/finalize_9.f90: New.

From-SVN: r157226

14 years agore PR rtl-optimization/40761 (IRA memory hog for insanely nested loops)
Richard Biener [Thu, 4 Mar 2010 13:25:27 +0000 (13:25 +0000)]
re PR rtl-optimization/40761 (IRA memory hog for insanely nested loops)

2010-03-04  Richard Guenther  <rguenther@suse.de>

PR tree-optimization/40761
* tree-ssa-pre.c (compute_antic): Walk reverse postorder
in reverse order.
(my_rev_post_order_compute): New function.
(init_pre): Call it.

From-SVN: r157225

14 years agoreassoc_4.f: Add --param max-completely-peel-times to dg-options for spu.
Tobias Burnus [Thu, 4 Mar 2010 12:56:06 +0000 (13:56 +0100)]
reassoc_4.f: Add --param max-completely-peel-times to dg-options for spu.

2010-03-04  Tobias Burnus  <burnus@net-b.de>
            Ken Werner <ken@linux.vnet.ibm.com>

        * gfortran.dg/reassoc_4.f: Add --param max-completely-peel-times
        to dg-options for spu.
        * gfortran.dg/vect/vect-7.f90: Add vect_intfloat_cvt to the
        dump-scan target to exclude spu.

Co-Authored-By: Ken Werner <ken@linux.vnet.ibm.com>
From-SVN: r157224

14 years agoFix ChangeLog entries tabs.
Sebastian Pop [Thu, 4 Mar 2010 12:16:54 +0000 (12:16 +0000)]
Fix ChangeLog entries tabs.

From-SVN: r157218

14 years agoFix PR43209: Do not decrease the cost of an IV candidate when the cost is infinite.
Changpeng Fang [Thu, 4 Mar 2010 12:12:50 +0000 (12:12 +0000)]
Fix PR43209: Do not decrease the cost of an IV candidate when the cost is infinite.

2010-03-03  Changpeng Fang  <changpeng.fang@amd.com>

      PR middle-end/43209
      * tree-ssa-loop-ivopts.c (determine_use_iv_cost_condition): Do not decrease
      the cost of an IV candidate when the cost is infinite.

      * gcc.dg/tree-ssa/ivopts-4.c: New.

From-SVN: r157217

14 years agodoc: normalize 3DNow! spelling and spacing.
Ralf Wildenhues [Thu, 4 Mar 2010 03:01:21 +0000 (03:01 +0000)]
doc: normalize 3DNow! spelling and spacing.

gcc/:
* doc/extend.texi (Vector Extensions, X86 Built-in Functions):
Use '3DNow!' for the extension of that name, ensure normal space
after the string.
* doc/invoke.texi (i386 and x86-64 Options): Likewise.

From-SVN: r157215

14 years agoMakefile.am: Re-organize xml sources.
Benjamin Kosnik [Thu, 4 Mar 2010 02:55:08 +0000 (02:55 +0000)]
Makefile.am: Re-organize xml sources.

2010-03-03  Benjamin Kosnik  <bkoz@redhat.com>

* doc/Makefile.am: Re-organize xml sources.
(doc-latex-doxygen): New.
(doc-pdf-doxygen): New.
(stamp-xml-doxygen, stamp-latex-doxygen): New.
* doc/Makefile.in: Regenerate.
* doc/doxygen/user.cfg.in: Add support here.
* doc/xml/manual/appendix_contributing.xml: Add specifics.

* doc/xml/manual/concurrency.xml: New.
* doc/xml/manual/atomics.xml: New.
* doc/xml/manual/spine.xml: Adjust.
* doc/xml/manual/diagnostics.xml: Adjust.

From-SVN: r157214

14 years agoDaily bump.
GCC Administrator [Thu, 4 Mar 2010 00:16:43 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r157213

14 years agore PR middle-end/32693 ([H8] : ICE: in gen_rtx_SUBREG, at emit-rtl.c:693)
Jeff Law [Wed, 3 Mar 2010 22:10:17 +0000 (15:10 -0700)]
re PR middle-end/32693 ([H8] : ICE: in gen_rtx_SUBREG, at emit-rtl.c:693)

* PR middle-end/32693
* expmed.c (store_bit_field_1): Use gen_lowpart_SUBREG rather
than gen_rtx_SUBREG.
(extract_bit_field_1): Likewise.

From-SVN: r157207

14 years agosourcebuild.texi (Test directives): Document that arguments include-opts and exclude...
Janis Johnson [Wed, 3 Mar 2010 22:05:30 +0000 (22:05 +0000)]
sourcebuild.texi (Test directives): Document that arguments include-opts and exclude-opts are now optional for...

* doc/sourcebuild.texi (Test directives): Document that arguments
include-opts and exclude-opts are now optional for dg-skip-if,
dg-xfail-if, dg-xfail-run-if, and dg-shouldfail.

* lib/target-supports-dg.exp (check-flags): Provide defaults for
include-opts and exclude-opts; skip checking the flags if arguments
are the same as the defaults.
(dg-xfail-if): Verify the number of arguments, supply defaults
for unspecified optional arguments.
(dg-skip-if, dg-xfail-run-if): Verify the number of arguments.

From-SVN: r157206

14 years agomangle19-1.C: Adjust for default -Wabi.
Jason Merrill [Wed, 3 Mar 2010 20:24:45 +0000 (15:24 -0500)]
mangle19-1.C: Adjust for default -Wabi.

* g++.dg/abi/mangle19-1.C: Adjust for default -Wabi.
* g++.dg/abi/mangle23.C: Likewise.
* g++.dg/eh/simd-2.C: Likewise.
* g++.dg/ext/attribute-test-2.C: Likewise.
* g++.dg/ext/vector14.C: Likewise.
* g++.dg/other/pr34435.C: Likewise.
* g++.dg/template/conv8.C: Likewise.
* g++.dg/template/nontype9.C: Likewise.
* g++.dg/template/qualttp17.C: Likewise.
* g++.dg/template/ref1.C: Likewise.
* g++.old-deja/g++.pt/crash68.C: Likewise.
* g++.old-deja/g++.pt/ref1.C: Likewise.
* g++.old-deja/g++.pt/ref3.C: Likewise.
* g++.old-deja/g++.pt/ref4.C: Likewise.

From-SVN: r157205

14 years agore PR c++/12909 (ambiguity in mangling vector types)
Jason Merrill [Wed, 3 Mar 2010 19:02:08 +0000 (14:02 -0500)]
re PR c++/12909 (ambiguity in mangling vector types)

PR c++/12909
* lto-streamer.h (LTO_tags): Add LTO_var_decl_alias.
* lto-streamer-in.c (lto_input_tree): Read it.
* lto-streamer-out.c (output_unreferenced_globals): Write it.

From-SVN: r157204

14 years agore PR c++/12909 (ambiguity in mangling vector types)
Jason Merrill [Wed, 3 Mar 2010 19:01:58 +0000 (14:01 -0500)]
re PR c++/12909 (ambiguity in mangling vector types)

PR c++/12909
* cgraph.h (varpool_node): Add extra_name field.
* varpool.c (varpool_extra_name_alias): New.
(varpool_assemble_decl): Emit extra name aliases.
(varpool_mark_needed_node): Look past an extra name alias.
cp/
* mangle.c (mangle_decl): Handle VAR_DECL, too.

From-SVN: r157203

14 years agore PR c++/12909 (ambiguity in mangling vector types)
Jason Merrill [Wed, 3 Mar 2010 19:01:46 +0000 (14:01 -0500)]
re PR c++/12909 (ambiguity in mangling vector types)

PR c++/12909
* method.c (make_alias_for): Handle VAR_DECL, too.
* decl2.c (vague_linkage_p): Rename from vague_linkage_fn_p.
* tree.c (no_linkage_check): Adjust.
* decl.c (maybe_commonize_var): Adjust.
* mangle.c (mangle_decl): Adjust.
* cp-tree.h: Adjust.

From-SVN: r157202

14 years agore PR c++/12909 (ambiguity in mangling vector types)
Jason Merrill [Wed, 3 Mar 2010 19:01:36 +0000 (14:01 -0500)]
re PR c++/12909 (ambiguity in mangling vector types)

PR c++/12909
* mangle.c: Include cgraph.h.
(mangle_decl): If the mangled name will change in a later
ABI version, make the later mangled name an alias.
* method.c (make_alias_for): Copy DECL_ARGUMENTS.
* Make-lang.in (mangle.o): Depend on cgraph.h.

From-SVN: r157201

14 years agoconfig.gcc (sparc64-*-solaris2*, [...]): Merge into...
Eric Botcazou [Wed, 3 Mar 2010 18:14:50 +0000 (18:14 +0000)]
config.gcc (sparc64-*-solaris2*, [...]): Merge into...

* config.gcc (sparc64-*-solaris2*, sparc-*-solaris2*): Merge into...
(sparc*-*-solaris2*): ...this.

From-SVN: r157200

14 years agore PR fortran/43243 ([4.4 Regression ?] Wrong-code due to missing array temp for...
Paul Thomas [Wed, 3 Mar 2010 17:49:53 +0000 (17:49 +0000)]
re PR fortran/43243 ([4.4 Regression ?] Wrong-code due to missing array temp for DT with pointer component)

2010-03-03  Paul Thomas  <pault@gcc.gnu.org>

PR fortran/43243
* trans-array.c (gfc_conv_array_parameter): Contiguous refs to
allocatable ultimate components do not need temporaries, whilst
ultimate pointer components do.

2010-03-03  Paul Thomas  <pault@gcc.gnu.org>

PR fortran/43243
* gfortran.dg/internal_pack_12.f90: New test.

From-SVN: r157199

14 years agoUpdate gcc.dg/pr36997.c for vector size change.
H.J. Lu [Wed, 3 Mar 2010 16:54:19 +0000 (16:54 +0000)]
Update gcc.dg/pr36997.c for vector size change.

2010-03-03  H.J. Lu  <hongjiu.lu@intel.com>

* gcc.dg/pr36997.c: Adjust error message.

From-SVN: r157198

14 years agobuiltin-unreachable.c: Don't expect stack adjustments to not be present on machines...
Mike Stump [Wed, 3 Mar 2010 16:46:27 +0000 (16:46 +0000)]
builtin-unreachable.c: Don't expect stack adjustments to not be present on machines that align the...

* gcc.target/i386/builtin-unreachable.c: Don't expect stack
adjustments to not be present on machines that align the stack to
more than 4 bytes and don't have a red zone yet as that is an
unimplemented optimization.

From-SVN: r157197

14 years agore PR fortran/43169 ([OOP] gfortran rejects pure procedure with select type construct)
Janus Weil [Wed, 3 Mar 2010 15:12:40 +0000 (16:12 +0100)]
re PR fortran/43169 ([OOP] gfortran rejects pure procedure with select type construct)

2010-03-03  Janus Weil  <janus@gcc.gnu.org>

PR fortran/43169
* resolve.c (resolve_code): Correctly set gfc_current_ns for
EXEC_SELECT_TYPE.
(gfc_impure_variable): Make it work with sub-namespaces (BLOCK etc).
(gfc_pure): Ditto.

2010-03-03  Janus Weil  <janus@gcc.gnu.org>

PR fortran/43169
* gfortran.dg/impure_assignment_3.f90: New.

From-SVN: r157196

14 years agounique_ptr.h (unique_ptr<>::get_deleter): Simplify return type.
Paolo Carlini [Wed, 3 Mar 2010 11:06:59 +0000 (11:06 +0000)]
unique_ptr.h (unique_ptr<>::get_deleter): Simplify return type.

2010-03-03  Paolo Carlini  <paolo.carlini@oracle.com>
    Jonathan Wakely  <jwakely.gcc@gmail.com>

* include/bits/unique_ptr.h (unique_ptr<>::get_deleter): Simplify
return type.
* testsuite/20_util/unique_ptr/modifiers/reset_neg.cc: Adjust
dg-error line number.
* testsuite/20_util/unique_ptr/assign/assign_neg.cc: Likewise.

Co-Authored-By: Jonathan Wakely <jwakely.gcc@gmail.com>
From-SVN: r157193

14 years agolto-elf.c (is_compatible_architecture): New static function.
Eric Botcazou [Wed, 3 Mar 2010 10:29:22 +0000 (10:29 +0000)]
lto-elf.c (is_compatible_architecture): New static function.

* lto-elf.c (is_compatible_architecture): New static function.
(DEFINE_VALIDATE_EHDR): Use it to validate the architecture.

From-SVN: r157192

14 years agore PR debug/43229 (ice expand_debug_expr, at cfgexpand.c:2957)
Jakub Jelinek [Wed, 3 Mar 2010 09:58:11 +0000 (10:58 +0100)]
re PR debug/43229 (ice expand_debug_expr, at cfgexpand.c:2957)

PR debug/43229
* cfgexpand.c (expand_debug_expr): Handle DOT_PROD_EXPR,
WIDEN_MULT_EXPR and WIDEN_SUM_EXPR.  Return NULL without
ICE for vector expressions, ADDR_SPACE_CONVERT_EXPR,
FIXED_CONVERT_EXPR, OBJ_TYPE_REF and WITH_SIZE_EXPR.

* gfortran.dg/pr43229.f90: New test.

From-SVN: r157191

14 years agore PR debug/43237 (Wrong DW_AT_upper_bound)
Jakub Jelinek [Wed, 3 Mar 2010 07:02:37 +0000 (08:02 +0100)]
re PR debug/43237 (Wrong DW_AT_upper_bound)

PR debug/43237
* dwarf2out.c (add_bound_info): If a decl bound doesn't have decl_die,
fallthrough to default handling, just with want_address 0 instead of 2.
For single element lists, add_AT_loc directly, otherwise create an
artificial variable DIE and stick location list to it.

* gcc.dg/debug/dwarf2/pr43237.c: New test.

From-SVN: r157190

14 years agore PR debug/43177 (Handle at least simple cases of reversible insns in debug info)
Jakub Jelinek [Wed, 3 Mar 2010 07:01:27 +0000 (08:01 +0100)]
re PR debug/43177 (Handle at least simple cases of reversible insns in debug info)

PR debug/43177
* var-tracking.c (loc_cmp): Don't assert VALUEs have the same mode.
(VAL_EXPR_HAS_REVERSE): Define.
(reverse_op): New function.
(add_stores): For reversible operations add an extra MO_VAL_USE.

* gcc.dg/guality/pr43177.c: New test.

From-SVN: r157189

14 years agore PR debug/43177 (Handle at least simple cases of reversible insns in debug info)
Jakub Jelinek [Wed, 3 Mar 2010 06:59:09 +0000 (07:59 +0100)]
re PR debug/43177 (Handle at least simple cases of reversible insns in debug info)

PR debug/43177
* var-tracking.c (loc_cmp): Don't assert VALUEs have the same mode.
(VAL_EXPR_HAS_REVERSE): Define.
(reverse_op): New function.
(add_stores): For reversible operations add an extra MO_VAL_USE.

* gcc.dg/guality/pr43177.c: New test.

From-SVN: r157188

14 years ago* c-pretty-print.c (pp_c_specifier_qualifier_list): Print vector size.
Jason Merrill [Wed, 3 Mar 2010 05:11:13 +0000 (00:11 -0500)]
* c-pretty-print.c (pp_c_specifier_qualifier_list): Print vector size.

From-SVN: r157187

14 years agobitset (hash): Minor tweaks.
Paolo Carlini [Wed, 3 Mar 2010 02:01:56 +0000 (02:01 +0000)]
bitset (hash): Minor tweaks.

2010-03-02  Paolo Carlini  <paolo.carlini@oracle.com>

* include/debug/bitset (hash): Minor tweaks.
* include/debug/vector (hash): Likewise.
* include/profile/bitset (hash): Likewise.
* include/profile/vector (hash): Likewise.

From-SVN: r157186

14 years agofunctional_hash.h (_Fnv_hash_base<>::hash): Change to template.
Paolo Carlini [Wed, 3 Mar 2010 01:23:07 +0000 (01:23 +0000)]
functional_hash.h (_Fnv_hash_base<>::hash): Change to template.

2010-03-02  Paolo Carlini  <paolo.carlini@oracle.com>

* include/bits/functional_hash.h (_Fnv_hash_base<>::hash): Change
to template.
* include/tr1/functional_hash.h (_Fnv_hash_base<>::hash): Likewise.
* include/bits/vector.tcc (hash): Adjust.
* include/bits/basic_string.h (hash): Likewise.
* include/std/bitset (hash): Likewise.
* src/hash-string-aux.cc (hash): Likewise.

From-SVN: r157185

14 years agoDaily bump.
GCC Administrator [Wed, 3 Mar 2010 00:17:36 +0000 (00:17 +0000)]
Daily bump.

From-SVN: r157184

14 years agoconfig.gcc (sparc-*-linux*): Do not include sparc/gas.h.
Eric Botcazou [Tue, 2 Mar 2010 23:35:12 +0000 (23:35 +0000)]
config.gcc (sparc-*-linux*): Do not include sparc/gas.h.

* config.gcc (sparc-*-linux*): Do not include sparc/gas.h.
(sparc64-*-linux*): Likewise.
(sparc64-*-solaris2*): Include assembler files before linker ones.
(sparc-*-solaris2*): Simplify and reorder to match previous case.
* config/sparc/gas.h: Delete.
* config/sparc/sol2-64.h: Add copyright notice.
* config/sparc/sol2-gas-bi.h: Likewise.
* config/sparc/sol2-gld.h: Likewise.
* config/sparc/sysv4.h (TARGET_ASM_NAMED_SECTION): Delete.
* config/sparc/sol2.h (TARGET_ASM_NAMED_SECTION): Redefine.
* config/sparc/sol2-gas.h (TARGET_ASM_NAMED_SECTION): Likewise.
* config/sparc/sparc.c (TARGET_ASM_ALIGNED_SI_OP): Never redefine.
(sparc_elf_asm_named_section): Rename into...
(sparc_solaris_elf_asm_named_section): ...this.  Always define.

From-SVN: r157181

14 years agomutex (lock_guard::lock_guard): Do not lock mutex when adopt_lock_t constructor is...
Jonathan Wakely [Tue, 2 Mar 2010 23:00:21 +0000 (23:00 +0000)]
mutex (lock_guard::lock_guard): Do not lock mutex when adopt_lock_t constructor is used.

2010-03-02  Jonathan Wakely  <jwakely.gcc@gmail.com>

* include/std/mutex (lock_guard::lock_guard): Do not lock mutex when
adopt_lock_t constructor is used.
* testsuite/30_threads/lock_guard/cons/1.cc: New.

From-SVN: r157180

14 years agostl_bvector.h (hash<vector<bool, [...]): Move definition...
Paolo Carlini [Tue, 2 Mar 2010 21:49:12 +0000 (21:49 +0000)]
stl_bvector.h (hash<vector<bool, [...]): Move definition...

2010-03-02  Paolo Carlini  <paolo.carlini@oracle.com>

* include/bits/stl_bvector.h (hash<vector<bool, _Alloc>>::
operator()(const vector<bool, _Alloc>&)): Move definition...
* include/bits/vector.tcc: ... here.

From-SVN: r157179

14 years agostl_bvector.h (hash<vector<bool, [...]): Move definition...
Paolo Carlini [Tue, 2 Mar 2010 21:48:20 +0000 (21:48 +0000)]
stl_bvector.h (hash<vector<bool, [...]): Move definition...

2010-03-02  Paolo Carlini  <paolo.carlini@oracle.com>

* include/bits/stl_bvector.h (hash<vector<bool, _Alloc>>::
operator()(const vector<bool, _Alloc>&)): Move definition...
* include/bits/vector.tcc: ... here.

From-SVN: r157178

14 years agoMakefile.am (check-DEJAGNU): Include de* in normal0 dirs.
Rainer Orth [Tue, 2 Mar 2010 21:46:23 +0000 (21:46 +0000)]
Makefile.am (check-DEJAGNU): Include de* in normal0 dirs.

* testsuite/Makefile.am (check-DEJAGNU): Include de* in normal0
dirs.
* testsuite/Makefile.in: Regenerate.

From-SVN: r157177

14 years agostl_bvector.h (hash<vector<bool, _Alloc>>): Add.
Paolo Carlini [Tue, 2 Mar 2010 21:11:19 +0000 (21:11 +0000)]
stl_bvector.h (hash<vector<bool, _Alloc>>): Add.

2010-03-02  Paolo Carlini  <paolo.carlini@oracle.com>

* include/bits/stl_bvector.h (hash<vector<bool, _Alloc>>): Add.
* include/debug/vector (hash<__debug::vector<bool, _Alloc>>):
Likewise.
* include/profile/vector (hash<__profile::vector<bool, _Alloc>>):
Likewise.
* testsuite/23_containers/vector/bool/hash/1.cc: New.

* include/std/bitset (hash<bitset<_Nb>>): Small tweaks.
(hash<bitset<0>>): Add.
* include/debug/bitset (hash<__debug::bitset<_Nb>>): Forward to
hash<bitset<_Nb>>.
* include/profile/bitset (hash<__profile::bitset<_Nb>>): Likewise.
* testsuite/23_containers/bitset/hash/1.cc: Improve.

From-SVN: r157176

14 years ago* dg-extract-results.sh: Fix result order to match sequential case.
Rainer Orth [Tue, 2 Mar 2010 20:28:55 +0000 (20:28 +0000)]
* dg-extract-results.sh: Fix result order to match sequential case.

From-SVN: r157175

14 years agoAdd PR# to my last change
Jeff Law [Tue, 2 Mar 2010 20:00:25 +0000 (13:00 -0700)]
Add PR# to my last change

From-SVN: r157173

14 years agore PR c++/41090 (Using static label reference in c++ class constructor produces wrong...
Mike Stump [Tue, 2 Mar 2010 19:40:02 +0000 (19:40 +0000)]
re PR c++/41090 (Using static label reference in c++ class constructor produces wrong code)

PR c++/41090
* g++.dg/ext/label13.C (C::C): xfail for darwin for now.

From-SVN: r157172

14 years agoalpha.c (override_options): Fix -mtune error message.
Uros Bizjak [Tue, 2 Mar 2010 19:18:40 +0000 (20:18 +0100)]
alpha.c (override_options): Fix -mtune error message.

* config/alpha/alpha.c (override_options): Fix -mtune error message.

From-SVN: r157169

14 years agoreload1.c (rtx_p, [...]): Declare.
Jeff Law [Tue, 2 Mar 2010 18:56:50 +0000 (11:56 -0700)]
reload1.c (rtx_p, [...]): Declare.

* reload1.c (rtx_p, substitute_stack): Declare.
(substitute): Record addresses of changed rtxs.
(gen_reload_chain_without_interm_reg_p): Don't use copy_rtx anymore.
Restore the original rtx when complete.
(reload): Free subsitute_stack when complete.

From-SVN: r157168

14 years agogccint.texi (menu): Add Testsuites as a chapter.
Janis Johnson [Tue, 2 Mar 2010 17:44:30 +0000 (17:44 +0000)]
gccint.texi (menu): Add Testsuites as a chapter.

* doc/gccint.texi (menu): Add Testsuites as a chapter.
* doc/sourcebuild.texi (Testsuites): Move up a level to be a
new chapter.
(Test Idioms, Test Directives, Ada Tests, C Tests, libgcj Tests,
LTO Testing, gcov Testing, profopt Testing, compat Testing,
Torture Tests): Change from subsection to section.

From-SVN: r157167

14 years agore PR libstdc++/43230 (Revision 157158 failed 30_threads/promise/members/set_value3.cc)
Jonathan Wakely [Tue, 2 Mar 2010 15:36:00 +0000 (15:36 +0000)]
re PR libstdc++/43230 (Revision 157158 failed 30_threads/promise/members/set_value3.cc)

2010-03-02  Jonathan Wakely  <jwakely.gcc@gmail.com>

PR libstdc++/43230
* testsuite/30_threads/promise/members/set_value3.cc
(tester::~tester): Remove.

From-SVN: r157166

14 years agobitset (_Base_bitset<>::_M_getdata()): Add.
Paolo Carlini [Tue, 2 Mar 2010 14:08:52 +0000 (14:08 +0000)]
bitset (_Base_bitset<>::_M_getdata()): Add.

2010-03-02  Paolo Carlini  <paolo.carlini@oracle.com>

* include/std/bitset (_Base_bitset<>::_M_getdata()): Add.
(hash<_GLIBCXX_STD_D::bitset<_Nb>>): Add, use the latter.
* include/debug/bitset (hash<std::__debug::bitset<_Nb>>): Add.
* include/profile/bitset (hash<std::__profile::bitset<_Nb>>): Likewise.
* testsuite/23_containers/bitset/hash/1.cc: New.

From-SVN: r157165

14 years agovar-tracking.c (vt_initialize): Scan insns in ebb chunks instead of bb.
Jakub Jelinek [Tue, 2 Mar 2010 12:18:39 +0000 (13:18 +0100)]
var-tracking.c (vt_initialize): Scan insns in ebb chunks instead of bb.

* var-tracking.c (vt_initialize): Scan insns in ebb chunks instead
of bb.

Co-Authored-By: Steven Bosscher <steven@gcc.gnu.org>
From-SVN: r157164

14 years agore PR fortran/43180 (Bad results without temporary copy of intent(in) argument)
Paul Thomas [Tue, 2 Mar 2010 11:58:02 +0000 (11:58 +0000)]
re PR fortran/43180 (Bad results without temporary copy of intent(in) argument)

2010-03-02  Paul Thomas  <pault@gcc.gnu.org>

PR fortran/43180
* trans-array.c (gfc_conv_array_parameter): A full array of
derived type need not be restricted to a symbol without an
array spec to use the call to gfc_conv_expr_descriptor.

PR fortran/43173
* trans-array.c (gfc_conv_array_parameter): Contiguous refs to
allocatable arrays do not need temporaries.

2010-03-02  Paul Thomas  <pault@gcc.gnu.org>

PR fortran/43180
* gfortran.dg/internal_pack_10.f90: New test.

PR fortran/43173
* gfortran.dg/internal_pack_11.f90: New test.

From-SVN: r157163

14 years agoFix PR42640: Correctly initialize the value of the new induction variable.
Sebastian Pop [Tue, 2 Mar 2010 10:22:30 +0000 (10:22 +0000)]
Fix PR42640: Correctly initialize the value of the new induction variable.

2010-03-02  Reza Yazdani  <reza.yazdani@amd.com>

PR middle-end/42640
* tree-loop-distribution.c (update_phis_for_loop_copy): Replaced
the assignment from the new induction variable to the assignment
of the value from the original loop PHI function.

* gcc.dg/tree-ssa/pr42640.c: New.

From-SVN: r157161

14 years agoSmall multilib rule fixups.
Ralf Wildenhues [Tue, 2 Mar 2010 06:09:56 +0000 (06:09 +0000)]
Small multilib rule fixups.

libgcc/:
        PR other/42980
        * Makefile.in (install): Use $(MAKE) string in rule, for
        parallel make.

libiberty/:
        * Makefile.in (all): Do not use exec.

From-SVN: r157159

14 years agore PR libstdc++/43183 (std::unique_ptr::reset() does not conform to N3035.)
Jonathan Wakely [Tue, 2 Mar 2010 00:40:28 +0000 (00:40 +0000)]
re PR libstdc++/43183 (std::unique_ptr::reset() does not conform to N3035.)

2010-03-02  Jonathan Wakely  <jwakely.gcc@gmail.com>

PR libstdc++/43183
* include/bits/unique_ptr.h (reset): Fix as per working paper.
(operator*, operator->, operator[], operator bool, release): Use
pointer's null value instead of 0.
* testsuite/20_util/unique_ptr/assign/assign_neg.cc: Adjust.
* testsuite/20_util/unique_ptr/modifiers/reset_neg.cc: Adjust.
* testsuite/20_util/unique_ptr/modifiers/43183.cc: New.

From-SVN: r157158

14 years agoDaily bump.
GCC Administrator [Tue, 2 Mar 2010 00:17:00 +0000 (00:17 +0000)]
Daily bump.

From-SVN: r157157

14 years agosourcebuild.texi (Test directives): Clarify options to dg-skip-if.
Janis Johnson [Mon, 1 Mar 2010 23:29:18 +0000 (23:29 +0000)]
sourcebuild.texi (Test directives): Clarify options to dg-skip-if.

* doc/sourcebuild.texi (Test directives): Clarify options to
dg-skip-if.

Co-Authored-By: Daniel Jacobowitz <dan@codesourcery.com>
From-SVN: r157154

14 years agoiomanip (get_money, put_money): Add in C++0x mode; tidy.
Paolo Carlini [Mon, 1 Mar 2010 19:12:39 +0000 (19:12 +0000)]
iomanip (get_money, put_money): Add in C++0x mode; tidy.

2010-03-01  Paolo Carlini  <paolo.carlini@oracle.com>

* include/std/iomanip (get_money, put_money): Add in C++0x mode; tidy.
* testsuite/27_io/manipulators/extended/get_money/char/1.cc: New.
* testsuite/27_io/manipulators/extended/get_money/wchar_t/1.cc:
Likewise.
* testsuite/27_io/manipulators/extended/put_money/char/1.cc: Likewise.
* testsuite/27_io/manipulators/extended/put_money/wchar_t/1.cc:
Likewise.

From-SVN: r157153

14 years agoatomicfwd_cxx.h: Add typedefs to group.
Benjamin Kosnik [Mon, 1 Mar 2010 19:04:23 +0000 (19:04 +0000)]
atomicfwd_cxx.h: Add typedefs to group.

2010-03-01  Benjamin Kosnik  <bkoz@redhat.com>

* include/bits/atomicfwd_cxx.h: Add typedefs to group.

From-SVN: r157152

14 years agoconfigure.ac (gcc_cv_as_cfi_directive): Disable cfi directives unless GCC and gas...
Rainer Orth [Mon, 1 Mar 2010 18:09:19 +0000 (18:09 +0000)]
configure.ac (gcc_cv_as_cfi_directive): Disable cfi directives unless GCC and gas agree on using read-only .eh_frame...

* configure.ac (gcc_cv_as_cfi_directive) [i?86-*-solaris*]:
Disable cfi directives unless GCC and gas agree on using read-only
.eh_frame sections for 64-bit.
* configure: Regenerate.

From-SVN: r157151

14 years agore PR tree-optimization/43220 (Paritially optimized __builtin_save_stack/__builtin_re...
Richard Guenther [Mon, 1 Mar 2010 16:57:02 +0000 (16:57 +0000)]
re PR tree-optimization/43220 (Paritially optimized __builtin_save_stack/__builtin_restore_stack causes segmentation fault)

2010-03-01  Richard Guenther  <rguenther@suse.de>

PR tree-optimization/43220
* tree-ssa-ccp.c (optimize_stack_restore): Do not optimize
BUILT_IN_STACK_{SAVE,RESTORE} around alloca.

* gcc.c-torture/execute/pr43220.c: New testcase.

From-SVN: r157149

14 years agore PR middle-end/41250 (hppa has DECL_VALUE_EXPR decls appearing in the function)
Richard Guenther [Mon, 1 Mar 2010 15:43:32 +0000 (15:43 +0000)]
re PR middle-end/41250 (hppa has DECL_VALUE_EXPR decls appearing in the function)

2010-03-01  Richard Guenther  <rguenther@suse.de>
Martin Jambor  <mjambor@suse.cz>

PR middle-end/41250
* gimplify.c (gimplify_body): Unset DECL_HAS_VALUE_EXPR_P on
gimplified parameters.

Co-Authored-By: Martin Jambor <mjambor@suse.cz>
From-SVN: r157148

14 years agoset LABEL_PRESERVE_P for landing pad labels
Christian Bruel [Mon, 1 Mar 2010 14:34:44 +0000 (15:34 +0100)]
set LABEL_PRESERVE_P for landing pad labels

From-SVN: r157146

14 years agore PR pch/14940 (PCH largefile test fails on various platforms)
Rainer Orth [Mon, 1 Mar 2010 14:07:12 +0000 (14:07 +0000)]
re PR pch/14940 (PCH largefile test fails on various platforms)

PR pch/14940
* gcc.dg/pch/pch.exp: Don't XFAIL largefile.c on i?86-*-solaris2.10.

From-SVN: r157145

14 years agore PR other/32499 (libstdc++ testsuite fails on platforms without ranlib)
Rainer Orth [Mon, 1 Mar 2010 13:58:48 +0000 (13:58 +0000)]
re PR other/32499 (libstdc++ testsuite fails on platforms without ranlib)

libstdc++-v3:
PR libstdc++/32499
* testsuite/Makefile.am (check-DEJAGNU
$(check_DEJAGNU_normal_targets)): Export AR, RANLIB.
* testsuite/Makefile.in: Regenerate.

toplevel:
PR libstdc++/32499
* configure.ac (RANLIB): Default to true.
(STRIP): Likewise.
(RANLIB_FOR_TARGET): Remove superfluous : argument.
* configure: Regenerate.

From-SVN: r157144

14 years agoUse SPEC_32 and SPEC_64 in ASM_SPEC.
H.J. Lu [Mon, 1 Mar 2010 13:05:44 +0000 (13:05 +0000)]
Use SPEC_32 and SPEC_64 in ASM_SPEC.

2010-03-01  H.J. Lu  <hongjiu.lu@intel.com>

* config/i386/linux64.h (ASM_SPEC): Use SPEC_32 and SPEC_64.

From-SVN: r157143

14 years agore PR tree-optimization/43213 (Worse code generated with -O2)
Richard Guenther [Mon, 1 Mar 2010 12:56:44 +0000 (12:56 +0000)]
re PR tree-optimization/43213 (Worse code generated with -O2)

2010-03-01  Richard Guenther  <rguenther@suse.de>

PR middle-end/43213
* expr.c (expand_assignment): Use the alias-oracle to tell
if the rhs aliases the result decl.

* gcc.dg/pr43213.c: New testcase.

From-SVN: r157142

14 years agore PR pch/14940 (PCH largefile test fails on various platforms)
Rainer Orth [Mon, 1 Mar 2010 12:44:33 +0000 (12:44 +0000)]
re PR pch/14940 (PCH largefile test fails on various platforms)

gcc:
PR pch/14940
* config/host-solaris.c (HOST_HOOKS_GT_PCH_GET_ADDRESS): Redefine
to sol_gt_pch_get_address.
(TRY_EMPTY_VM_SPACE): Define for all combinations of 32 and
64-bit, SPARC and x86.
(sol_gt_pch_get_address): New function.

gcc/testsuite:
PR pch/14940
* gcc.dg/pch/pch.exp: Don't XFAIL largefile.c on i?86-*-solaris2.10.

From-SVN: r157141

14 years agotarget-supports.exp (check_effective_target_ucn_nocache): New function.
Rainer Orth [Mon, 1 Mar 2010 12:40:38 +0000 (12:40 +0000)]
target-supports.exp (check_effective_target_ucn_nocache): New function.

* lib/target-supports.exp (check_effective_target_ucn_nocache):
New function.
(check_effective_target_ucn): Likewise.

* g++.dg/other/ucnid-1.C: Don't XFAIL on *-*-solaris2.*, but skip
if !ucn.
* gcc.dg/ucnid-2.c: Likewise.
* gcc.dg/ucnid-3.c: Likewise.
* gcc.dg/ucnid-4.c: Likewise.
* gcc.dg/ucnid-11.c: Likewise.
* gcc.dg/ucnid-12.c: Likewise.
* gcc.dg/ucnid-6.c: Skip if !ucn.
* gcc.dg/ucnid-9.c: Likewise.

From-SVN: r157140

14 years agopr32000-2.c: Add missing include and exclude lists to dg-skip-if.
Rainer Orth [Mon, 1 Mar 2010 12:37:30 +0000 (12:37 +0000)]
pr32000-2.c: Add missing include and exclude lists to dg-skip-if.

* gcc.target/i386/pr32000-2.c: Add missing include and exclude
lists to dg-skip-if.
* gcc.target/i386/stackalign/return-3.c: Likewise.

From-SVN: r157139