Jason Merrill [Thu, 6 May 2010 19:46:14 +0000 (15:46 -0400)]
gimplify.c (gimplify_expr): Set GS_ALL_DONE when appropriate.
gcc:
* gimplify.c (gimplify_expr): Set GS_ALL_DONE when appropriate.
Don't change GS_OK to GS_ALL_DONE. Make sure that all cases set
ret appropriately.
(gimplify_compound_lval): Return GS_ALL_DONE as appropriate.
gcc/cp:
* semantics.c (simplify_aggr_init_expr): Use INIT_EXPR.
Jason Merrill [Thu, 6 May 2010 19:46:03 +0000 (15:46 -0400)]
gimplify.c (gimplify_modify_expr_rhs): Don't return GS_OK for stripping WITH_SIZE_EXPR.
* gimplify.c (gimplify_modify_expr_rhs): Don't return GS_OK for
stripping WITH_SIZE_EXPR.
(gimplify_expr) [MODIFY_EXPR]: Trust GS_OK even if the rhs didn't
change.
Michael Matz [Thu, 6 May 2010 13:54:32 +0000 (13:54 +0000)]
re PR tree-optimization/43984 (PRE misses full-redundancies, inserts into loops)
PR tree-optimization/43984
* tree-ssa-pre.c (inserted_phi_names): Remove.
(inserted_exprs): Change to bitmap.
(create_expression_by_pieces): Set bits, don't append to vector.
(insert_into_preds_of_block): Don't handle inserted_phi_names.
(eliminate): Don't look at inserted_phi_names, remove deleted
insns from inserted_exprs.
(remove_dead_inserted_code): Adjust to use bitmaps instead of
vectors.
(init_pre, fini_pre): Allocate and free bitmaps.
(execute_pre): Insert insns on edges before elimination.
re PR tree-optimization/43934 (LIM should handle PHI nodes)
2010-05-06 Richard Guenther <rguenther@suse.de>
PR tree-optimization/43934
* tree-ssa-loop-im.c (movement_possibility): Handle PHI nodes.
(stmt_cost): Likewise.
(extract_true_false_args_from_phi): New helper.
(determine_max_movement): For PHI nodes verify we can hoist them
and compute their cost.
(determine_invariantness_stmt): Handle PHI nodes.
(move_computations_stmt): Likewise. Hoist PHI nodes in
if-converted form using COND_EXPRs.
(move_computations): Return TODO_cleanup_cfg if we hoisted PHI
nodes.
(tree_ssa_lim): Likewise.
* tree-flow.h (tree_ssa_lim): Adjust prototype.
* tree-ssa-loop.c (tree_ssa_loop_im): Return todo.
re PR middle-end/43987 (type-punning causes broken binaries unless -O0 is used)
2010-05-06 Richard Guenther <rguenther@suse.de>
PR tree-optimization/43987
* tree-ssa-structalias.c (could_have_pointers): For possibly
address-taken variables force pointers to be recorded.
(create_variable_info_for_1): Likewise.
(push_fields_onto_fieldstack): Pass in wheter all fields
must have pointers.
(find_func_aliases): Query types instead of vars whether
they contain pointers where appropriate.
gcc/testsuite/ChangeLog:
PR 18624
* g++.dg/warn/Wunused-7.C: Add dg-warning.
* g++.dg/template/sfinae16.C: Likewise.
* gcc.dg/Wunused-var-1.c: Moved to...
* c-c++-common/Wunused-var-1.c: ...here. New test.
* gcc.dg/Wunused-var-2.c: Moved to...
* c-c++-common/Wunused-var-2.c: ...here. New test.
* gcc.dg/Wunused-var-3.c: Moved to...
* c-c++-common/Wunused-var-3.c: ...here. New test.
* gcc.dg/Wunused-var-4.c: Moved to...
* gcc.dg/Wunused-var-1.c: ... here.
* gcc.dg/Wunused-var-5.c: Moved to...
* c-c++-common/Wunused-var-4.c: ...here. New test.
* gcc.dg/Wunused-var-7.c: Moved to...
* c-c++-common/Wunused-var-5.c: ...here. New test.
* gcc.dg/Wunused-var-6.c: Moved to...
* gcc.dg/Wunused-var-2.c: ... here.
* c-c++-common/Wunused-var-1.c: New test.
* c-c++-common/Wunused-var-2.c: New test.
* c-c++-common/Wunused-var-3.c: New test.
* c-c++-common/Wunused-var-4.c: New test.
* c-c++-common/Wunused-var-5.c: New test.
* g++.dg/warn/Wunused-var-1.C: New test.
* g++.dg/warn/Wunused-var-2.C: New test.
* g++.dg/warn/Wunused-var-3.C: New test.
* g++.dg/warn/Wunused-var-4.C: New test.
* g++.dg/warn/Wunused-var-5.C: New test.
* g++.dg/warn/Wunused-var-6.C: New test.
* g++.dg/warn/Wunused-var-7.C: New test.
* g++.dg/warn/Wunused-var-8.C: New test.
* g++.dg/warn/Wunused-parm-1.C: New test.
* g++.dg/warn/Wunused-parm-2.C: New test.
* g++.dg/warn/Wunused-parm-3.C: New test.
Ira Rosen [Thu, 6 May 2010 06:42:51 +0000 (06:42 +0000)]
re PR middle-end/43901 (FAIL: gcc.c-torture/compile/pr42196-2.c)
PR tree-optimization/43901
* tree-vect-stmts.c (vectorizable_call): Assert that vector
type is not NULL if it's transformation phase, and return
FALSE if it's analysis.
(vectorizable_conversion, vectorizable_operation,
vectorizable_type_demotion, vectorizable_type_promotion):
Likewise.
Steven Bosscher [Wed, 5 May 2010 21:32:44 +0000 (21:32 +0000)]
stor-layout.c (pending_sizes): Change the type to VEC(tree,gc) *.
* stor-layout.c (pending_sizes): Change the type to
VEC(tree,gc) *.
(get_pending_sizes, put_pending_size, put_pending_sizes):
Update the uses of pending_sizes.
* c-decl.c (store_parm_decls): Likewise.
* c-tree.h (struct c_arg_info): Likewise.
* tree.h: Update the prototype for get_pending_sizes and
put_pending_sizes.
Paolo Carlini [Wed, 5 May 2010 19:53:11 +0000 (19:53 +0000)]
29134.cc: Adjust for profile-mode too.
2010-05-05 Paolo Carlini <paolo.carlini@oracle.com>
* testsuite/23_containers/vector/bool/capacity/29134.cc: Adjust
for profile-mode too.
* testsuite/23_containers/vector/bool/modifiers/insert/31370.cc:
Likewise.
2010-05-05 Paolo Carlini <paolo.carlini@oracle.com>
* include/profile/unordered_map (unordered_map<>::insert(_InputIter,
_InputIter), unordered_multimap<>::insert(_InputIter, _InputIter)):
Fix typo, don't call base() on __first and __last.
2010-05-05 Paolo Carlini <paolo.carlini@oracle.com>
Fix profile-mode vs -std=c++0x
* include/profile/impl/profiler_trace.h: Do not include the
whole <algorithm>, include <bits/stl_heap.h> (additionally
to <bits/stl_algobase.h>).
(__write_cost_factors): Replace std::for_each with open-coding.
(__set_cost_factors): Likewise.
(__report): Likewise; use std::make_heap / std::sort_heap
instead of std::sort.
(__report_cost_factors): Replace std::remove with open-coding.
(__warn, __cost_factor_writer, __cost_factor_setter): Remove.
Daniel Franke [Wed, 5 May 2010 18:53:23 +0000 (14:53 -0400)]
re PR fortran/24978 (ICE in gfc_assign_data_value_range)
gcc/fortran/:
2010-05-05 Daniel Franke <franke.daniel@gmail.com>
PR fortran/24978
* gfortran.h: Removed repeat count from constructor, removed
all usages.
* data.h (gfc_assign_data_value_range): Changed return value from
void to gfc_try.
* data.c (gfc_assign_data_value): Add location to constructor element.
(gfc_assign_data_value_range): Call gfc_assign_data_value()
for each element in range. Return early if an error was generated.
* resolve.c (check_data_variable): Stop early if range assignment
generated an error.
gcc/testsuite/:
2010-05-05 Daniel Franke <franke.daniel@gmail.com>
Alexandre Oliva [Wed, 5 May 2010 10:43:36 +0000 (10:43 +0000)]
re PR debug/43478 (Missing DW_AT_location for a variable)
PR debug/43478
* df-problems.c (struct dead_debug_use, struct dead_debug): New.
(dead_debug_init, dead_debug_finish): New functions.
(dead_debug_add, dead_debug_insert_before): Likewise.
(df_note_bb_compute): Initialize a dead_debug object, add dead
debug uses to it, insert debug bind insns before death insns,
reset debug insns that refer to pending uses at the end.
* rtl.h (make_debug_expr_from_rtl): New prototype.
* varasm.c (make_debug_expr_from_rtl): New function.
Co-Authored-By: Jakub Jelinek <jakub@redhat.com>
From-SVN: r159063
Janus Weil [Wed, 5 May 2010 07:44:33 +0000 (09:44 +0200)]
re PR fortran/43696 ([OOP] Bogus error: Passed-object dummy argument must not be POINTER)
2010-05-05 Janus Weil <janus@gcc.gnu.org>
PR fortran/43696
* resolve.c (resolve_fl_derived): Some fixes for class variables.
* symbol.c (gfc_build_class_symbol): Add separate class container for
class pointers.
Eric Botcazou [Tue, 4 May 2010 17:47:52 +0000 (17:47 +0000)]
tree.c (free_lang_data_in_one_sizepos): New inline function.
* tree.c (free_lang_data_in_one_sizepos): New inline function.
(free_lang_data_in_type): Call it on TYPE_{MIN|MAX}_VALUE of numerical
types. Call it on TYPE_SIZE and TYPE_SIZE_UNIT of all types.
(free_lang_data_in_decl): Call it on DECL_SIZE and DECL_SIZE_UNIT of
all decls. Call it on DECL_FIELD_OFFSET of fields.
(find_decls_types_r): Follow DECL_VALUE_EXPR.
(iterative_hash_expr) <PLACEHOLDER_EXPR>: New case.
Martin Jambor [Tue, 4 May 2010 17:34:01 +0000 (19:34 +0200)]
tree-sra.c (build_access_from_expr_1): The first parameter type changed to simple tree.
2010-05-04 Martin Jambor <mjambor@suse.cz>
* tree-sra.c (build_access_from_expr_1): The first parameter type
changed to simple tree.
(build_access_from_expr): Likewise, gsi parameter was eliminated.
(scan_assign_result): Renamed to assignment_mod_result, enum elements
renamed as well.
(build_accesses_from_assign): Removed all parameters except for a
simple gimple statement. Now returns a simple bool.
(scan_function): All non-analysis parts moved to separate functions
sra_modify_function_body and ipa_sra_modify_function_body. Removed all
parameters and updated both callers.
(sra_modify_expr): Removed parameter data.
(sra_modify_function_body): New function.
(perform_intra_sra): Call sra_modify_function_body to modify the
function body.
(replace_removed_params_ssa_names): Parameter data changed into
adjustments vector.
(sra_ipa_modify_expr): Likewise. Also removed unused parameter gsi and
changed the parameter dont_convert to convert with the opposite
meaning.
(sra_ipa_modify_assign): Parameter data changed into adjustments
vector, return value changed to bool.
(ipa_sra_modify_function_body): New function.
(sra_ipa_reset_debug_stmts): Updated a comment.
(modify_function): Use ipa_sra_modify_function_body to modify function
body.
H.J. Lu [Tue, 4 May 2010 15:47:25 +0000 (15:47 +0000)]
Handle the same VALUE for true dependence.
gcc/
2010-05-04 H.J. Lu <hongjiu.lu@intel.com>
PR middle-end/43671
* alias.c (true_dependence): Handle the same VALUE in x and mem.
(canon_true_dependence): Likewise.
(write_dependence_p): Likewise.
Jan Hubicka [Tue, 4 May 2010 15:42:15 +0000 (17:42 +0200)]
Makefile.in (cgraphbuild.o): Add dependency on ipa-utils.h
* Makefile.in (cgraphbuild.o): Add dependency on ipa-utils.h
* cgraphbuild.c: Include ipa-utils.h
(record_reference_ctx): New struct.
(record_reference): Simplify to work on initializers; not statements.
(mark_address, mark_load, mark_store): New.
(build_cgraph_edges): Simplify using walk_stmt_load_store_addr_ops;
walk PHI nodes too.
(record_references_in_initializer): Update use of record_reference.
(rebuild_cgraph_edges): Simplify using walk_stmt_load_store_addr_ops;
walk PHI nodes too.
Jan Hubicka [Tue, 4 May 2010 15:17:10 +0000 (17:17 +0200)]
lto-symtab.c (lto_cgraph_replace_node): Do not remove edges; node will be removed anyway.
* lto-symtab.c (lto_cgraph_replace_node): Do not remove edges;
node will be removed anyway.
(lto_varpool_replace_node): Allow also unanalyzed nodes;
relink aliases of node into prevailing node.
* varpool.c (varpool_remove_node): Remove aliases properly;
when removing node, remove all its aliases too; remove DECL_INITIAL
of removed node; ggc_free the varpool node.
re PR tree-optimization/43879 (-fipa-pta causes various miscompilations)
2010-05-04 Richard Guenther <rguenther@suse.de>
PR tree-optimization/43879
* tree-ssa-structalias.c (alias_get_name): Use
DECL_ASSEMBLER_NAME if available.
(create_function_info_for): Return the varinfo node.
(ipa_pta_execute): Associate same-body aliases and extra names
with their origin nodes varinfo. Dump DECL_ASSEMBLER_NAME.
* g++.dg/torture/pr43879-1_0.C: New testcase.
* g++.dg/torture/pr43879-1_1.C: Likewise.
Jack Howarth [Mon, 3 May 2010 22:37:50 +0000 (22:37 +0000)]
re PR java/43839 (libjava incorrectly uses -liconv in testsuite, jni.exp when using --with-libiconv-prefix)
PR 43839
* testsuite/Makefile.am: Override automake for site.exp creation
and add entry to set libiconv.
* testsuite/Makefile.in: Regenerate.
* testsuite/libjava.jni/jni.exp (gcj_jni_get_cxxflags_invocation):
Add new global variable libiconv to handle alternative libiconv
locations.
Jason Merrill [Mon, 3 May 2010 21:16:40 +0000 (17:16 -0400)]
re PR c++/43680 ([DR 1022] G++ is too aggressive in optimizing away bounds checking with enums)
PR c++/43680
gcc:
* c.opt (-fstrict-enums): New.
* doc/invoke.texi (C++ Dialect Options): Document -fstrict-enums.
gcc/cp:
* decl.c (finish_enum): Use the TYPE_MIN_VALUE and TYPE_MAX_VALUE
from the selected underlying type unless -fstrict-enums. Set
ENUM_UNDERLYING_TYPE to have the restricted range.
* cvt.c (type_promotes_to): Use ENUM_UNDERLYING_TYPE.
* class.c (check_bitfield_decl): Likewise.
Steven Bosscher [Mon, 3 May 2010 17:47:47 +0000 (17:47 +0000)]
tm.texi (defmac SMALL_REGISTER_CLASSES): Remove.
* doc/tm.texi (defmac SMALL_REGISTER_CLASSES): Remove.
(TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P): Add documentation,
based on the above, for new target hook.
* config/sh/sh.h: Replace SMALL_REGISTER_CLASSES with new target
hook definition.
* config/sh/sh.c (sh_small_register_classes_for_mode_p): Add
implementation of the hook that considers all register classes
small except for SH64.
(sh_override_options): Use the new hook.
* config/sh/sh-protos.h (sh_small_register_classes_for_mode_p):
Add prototype.
* config/arm/arm.h: Replace SMALL_REGISTER_CLASSES with new target
hook definition.
* config/arm/arm.c (arm_small_register_classes_for_mode_p): Add
implementation of the hook that considers all register classes
small for THUMB1.
* config/arm/arm-protos.h (arm_small_register_classes_for_mode_p):
Add prototype.
* config/mips/mips.h: Replace SMALL_REGISTER_CLASSES with new target
hook definition.
* config/mips/mips.c (mips_small_register_classes_for_mode_p): Add
implementation of the hook that considers all register classes
small for MIPS16.
* config/mips/mips-protos.h (mips_small_register_classes_for_mode_p):
Add prototype.
projects / thirdparty / gcc.git / log
