Jonathan Wakely [Fri, 4 Jan 2019 11:06:49 +0000 (11:06 +0000)]
Avoid spurious test failures when -fno-inline in test flags
These tests rely on inlining, so if -fno-inline is added to the compiler
flags the tests fail. Use the predefined __NO_INLINE__ macro to detect
that situation, and don't bother testing the move assignment.
* testsuite/21_strings/basic_string/modifiers/assign/char/
move_assign_optim.cc: Avoid spurious failure when -fno-inline added
to test flags.
* testsuite/21_strings/basic_string/modifiers/assign/wchar_t/
move_assign_optim.cc: Likewise.
Jan Beulich [Fri, 4 Jan 2019 08:50:56 +0000 (08:50 +0000)]
x86: relax mask register constraints
While their use for masking is indeed restricted to %k1...%k7, use as
"normal" insn operands also permits %k0. Remove the unnecessary
limitations, requiring quite a few testsuite adjustments.
Oddly enough some AVX512{F,DQ} test cases already check for %k[0-7],
while others did permit {%k0} - where they get touched here anyway this
gets fixed at the same time.
Jonathan Wakely [Thu, 3 Jan 2019 20:38:11 +0000 (20:38 +0000)]
PR libstdc++/88607 replace or remove unnecessary UTF-8 characters
There are a number of UTF-8 characters in comments which add no value
and can be replaced with ASCII equivalents, or removed entirely for the
section sign (U+00A7).
Tom de Vries [Thu, 3 Jan 2019 15:08:25 +0000 (15:08 +0000)]
[nvptx] Factor out populate_offload_attrs
Factor out populate_offload_attrs from nvptx_reorg.
2019-01-03 Tom de Vries <tdevries@suse.de>
* config/nvptx/nvptx.c (struct offload_attrs): New.
(populate_offload_attrs): New function. Factor mask extraction out of
nvptx_reorg. Add extraction of dimensions.
(nvptx_reorg): Use populate_offload_attrs.
Tom de Vries [Thu, 3 Jan 2019 15:08:15 +0000 (15:08 +0000)]
[nvptx] Add early-out cases in nvptx_goacc_validate_dims
Add early-out cases for for oacc_min_dims_p and routine_p in
nvptx_goacc_validate_dims, allowing simplification of the rest of the
function.
2019-01-03 Tom de Vries <tdevries@suse.de>
* config/nvptx/nvptx.c (nvptx_goacc_validate_dims_1): Add early-out
cases for oacc_min_dims_p and routine_p. Add asserts for
oacc_default_dims_p and offload_region_p.
Martin Sebor [Wed, 2 Jan 2019 21:38:56 +0000 (21:38 +0000)]
gimple-fold.c (gimple_fold_builtin_strlen): Use set_strlen_range rather than set_range_info.
* gimple-fold.c (gimple_fold_builtin_strlen): Use set_strlen_range
rather than set_range_info.
* tree-ssa-strlen.c (set_strlen_range): Extracted from
maybe_set_strlen_range. Handle potentially boundary crossing
cases more conservatively.
(maybe_set_strlen_range): Parts refactored into set_strlen_range.
Call set_strlen_range.
* tree-ssa-strlen.h (set_strlen_range): Add prototype.
* gcc.dg/strlenopt-36.c: Update.
* gcc.dg/strlenopt-45.c: Update.
* gcc.c-torture/execute/strlen-5.c: New test.
* gcc.c-torture/execute/strlen-6.c: New test.
* gcc.c-torture/execute/strlen-7.c: New test.
Co-Authored-By: Jeff Law <law@redhat.com>
From-SVN: r267531
Martin Sebor [Wed, 2 Jan 2019 18:30:50 +0000 (18:30 +0000)]
re PR middle-end/88663 (internal compiler error: in check, at tree-vrp.c:188)
PR middle-end/88663
* gimple-fold.c (get_range_strlen): Update prototype to no longer
need the flexp argument.
(get_range_strlen_tree): Drop flexp argument. Drop flexp argument
from calls to get_range_strlen. Update comments. Just update
VAL for an unterminated const char array and let the reset of the
code handle it normally. No longer try to set *flexp. Adjust
return value.
(get_range_strlen): Update for the new get_range_strlen API.
(get_maxval_strlen): Similarly.
(gimple_fold_builtin_strlen): Handle update meaning of return value
from get_range_strlen.
* gimple-ssa-sprintf.c (get_string_length): Update for the new
get_range_strlen API.
Co-Authored-By: Jeff Law <law@redhat.com>
From-SVN: r267520
Jonathan Wakely [Wed, 2 Jan 2019 16:30:49 +0000 (16:30 +0000)]
Add more testcases for class template argument deduction of maps
This adds additional tests for std::map and std::multimap CTAD. The
tests ensure that deduction works for braced-init-list of value_type
objects, and for pairs of input iterators (with both std::pair<Key, T>
and value_type as the iterator's value_type). This ensures deduction
from value_type still works, as well as the non-value_type cases in LWG
3025.
Similar tests for unordered maps do not work, apparently because the
constructor taking an initializer_list<value_type> is not usable for
deduction, and the deduction guide taking initializer_list<pair<Key, T>>
deduces key_type to be const. I am not addressing that.
* testsuite/23_containers/map/cons/deduction.cc: Test deduction from
initializer_list<value_type> and from input iterator ranges.
* testsuite/23_containers/multimap/cons/deduction.cc: Likewise.
Jan Hubicka [Wed, 2 Jan 2019 15:23:27 +0000 (16:23 +0100)]
re PR lto/88130 (ICE in copy_function_or_variable, at lto-streamer-out.c:2315 since r260963)
PR lto/88130
* varpool.c (varpool_node::ctor_useable_for_folding_p): Also return
false at WPA time when body was removed.
* g++.dg/torture/pr88130.C: New testcase.
Martin Liska [Wed, 2 Jan 2019 14:03:53 +0000 (15:03 +0100)]
Calculate prediction remainder at proper place (PR tree-optimization/88650).
2019-01-02 Martin Liska <mliska@suse.cz>
PR tree-optimization/88650
* predict.c (set_even_probabilities): Calculate probability
remainer only when really used.
2019-01-02 Martin Liska <mliska@suse.cz>
PR tree-optimization/88650
* gfortran.dg/predict-3.f90: New test.
Richard Biener [Wed, 2 Jan 2019 13:52:03 +0000 (13:52 +0000)]
re PR tree-optimization/88651 (tree-data-ref.c:3764:26: runtime error: signed integer overflow: 9223372036854775802 - -6 cannot be represented in type 'long int')
2019-01-02 Richard Biener <rguenther@suse.de>
PR middle-end/88651
* tree-data-ref.c (analyze_subscript_affine_affine): Use
widest_ints when mangling max_stmt_execution results.
Richard Biener [Wed, 2 Jan 2019 08:49:07 +0000 (08:49 +0000)]
re PR lto/85574 (LTO bootstapped binaries differ)
2019-01-02 Richard Biener <rguenther@suse.de>
PR ipa/85574
* ipa-icf.h (sem_item_optimizer::sort_congruence_split): Declare.
* ipa-icf.c (sem_item_optimizer::sort_congruence_split): New
function.
(sem_item_optimizer::do_congruence_step_f): Sort the congruence
set after UIDs before splitting them.
Martin Sebor [Wed, 2 Jan 2019 06:17:54 +0000 (06:17 +0000)]
gimple-fold.c (get_range_strlen_tree): Record if the computed length is optimistic.
* gimple-fold.c (get_range_strlen_tree): Record if the computed
length is optimistic. If it is, then arrange to compute the
conservative length as well.
* gimple-fold.h (get_range_strlen): Update prototype.
* builtins.c (check_access): Update call to get_range_strlen to use
c_strlen_data pointer. Change various variable accesses to instead
pull data from the c_strlen_data structure.
(check_strncat_sizes, expand_builtin_strncat): Likewise.
* calls.c (maybe_warn_nonstring_arg): Likewise.
* tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Likewise. Reset
minimum length if maximum lengh is unknown.
* gimple-ssa-sprintf.c (get_string_length): Likewise. Drop code
that used c_strlen, it's no longer needed. Restructure slightly.
(format_string): Set unlikely range appropriately.
* gimple-fold.c (get_range_strlen): Update comments. Fix minor
formatting issues.
(get_range_strlen): Accept c_strlen_data pointer for external
call sites as well. Pass through to call to internal get_range_strlen.
Adjust minlen, maxlen and maxbound as needed.
(get_maxval_strlen): Update comments.
(gimple_fold_builtin_strlen): Update call to get_range_strlen
to use c_strlen_data pointer. Change variable accesses to instead
use c_strlen_data data members.
* gcc.dg/strlenopt-40.c: Disable a couple tests.
* gcc.dg/strlenopt-48.c: Twiddle test.
* gcc.dg/strlenopt-59.c: New test.
* gcc.dg/tree-ssa/builtin-snprintf-5.c: New test.
* g++.dg/init/strlen.C: New test.
Co-Authored-By: Jeff Law <law@redhat.com>
From-SVN: r267503
Thomas Koenig [Tue, 1 Jan 2019 21:19:53 +0000 (21:19 +0000)]
re PR fortran/82743 (uncaught character truncation in derived type initialization)
2019-01-01 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/82743
* primary.c (gfc_convert_to_structure_constructor): If a character
in a constructor is too long, add a warning with
-Wcharacter-truncation.
2019-01-01 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/82743
* gfortran.dg/structure_constructor_16.f90: New test.
* gimple-fold.c (get_range_strlen): Update prototype.
(get_range_strlen_tree): Update prototype. Drop minlen/maxlen
local variables. Use pdata to return information to caller.
Update calls to get_range_strlen. Update pdata->maxbound.
(get_range_strlen -- static version): Similarly.
(get_range_strlen -- extern version): Update for internal
get_range_strlen API change. Convert to external data format.
(get_maxval_strlen): Similarly.
Co-Authored-By: Jeff Law <law@redhat.com>
From-SVN: r267498
Jan Hubicka [Tue, 1 Jan 2019 13:10:39 +0000 (14:10 +0100)]
devirt-13_0.C: Drop broken scan of ssa dump.
* g++.dg/lto/devirt-13_0.C: Drop broken scan of ssa dump.
* g++.dg/lto/devirt-14_0.C: Drop broken scan of ssa dump.
* g++.dg/lto/devirt-23_0.C: Add -fdump-ipa-cp.
Thomas Koenig [Mon, 31 Dec 2018 14:59:46 +0000 (14:59 +0000)]
re PR fortran/82995 (Segmentation fault passing optional argument to intrinsic sum function)
2018-12-31 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/82995
* trans-expr.c (gfc_conv_procedure_call): Pass NULL pointer
for missing optional dummy arguments for library routines.
* trans-intinsic.c (conv_mask_condition): New function.
(gfc_conv_intrinsic_arith): Detect and handle optional mask.
(gfc_conv_intrinsic_minmaxloc): Likewise.
(gfc_conv_intrinsic_findloc): Likewise.
(gfc_conv_intrinsic_minmaxval): Likewise.
(gfc_inline_intrinsic_function_p): Do not inline for rank > 1 if
an optional mask is present.
Martin Liska [Mon, 31 Dec 2018 13:46:08 +0000 (14:46 +0100)]
predict.c (set_even_probabilities): Include also unlikely_count in calculation.
).
2018-12-31 Martin Liska <mliska@suse.cz>
* predict.c (set_even_probabilities): Include also
unlikely_count in calculation.
(combine_predictions_for_bb): Consider also HOT and
COLD labels predictions.
* predict.def (PRED_HOT_LABEL): Move it just after
__builtin_expect_with_probability predictor.
(PRED_COLD_LABEL): Likewise.
2018-12-31 Martin Liska <mliska@suse.cz>
* g++.dg/predict-2.C: New test.
* g++.dg/predict-3.C: New test.
* g++.dg/predict-4.C: New test.
* gcc.dg/tree-ssa/attr-hotcold-2.c: Adjust test-case.
Martin Sebor [Mon, 31 Dec 2018 01:01:10 +0000 (01:01 +0000)]
gimple-fold.c (strlen_range_kind): New enum.
* gimple-fold.c (strlen_range_kind): New enum.
(get_range_strlen): Update signature to use strlen_range_kind
instead of type+fuzzy.
(get_range_strlen_tree): Update signature to use
strlen_range_kind instead of type+fuzzy. Pass rkind down to
get_range_strlen. Check the rkind enum value instead of
type+fuzzy.
(get_range_strlen): Similarly.
(get_maxval_strlen): Update signature to make NONSTR argument have
a default value and make static. Add asserts to ensure sanity of
arguments. Update calls to get_range_strlen.
(gimple_fold_builtin_strcpy): Update calls to get_maxval_strlen.
(gimple_fold_builtin_strcat, gimple_fold_builtin_fputs): Similarly.
(gimple_fold_builtin_memory_chk): Similarly.
(gimple_fold_builtin_stxcpy_chk): Similarly.
(gimple_fold_builtin_snprintf_chk): Similarly.
(gimple_fold_builtin_sprintf, gimple_fold_builtin_snprintf): Similarly.
* gimple-fold.h (get_maxval_strlen): Delete prototype.
Co-Authored-By: Jeff Law <law@redhat.com>
From-SVN: r267482
msp430.c (msp430_attr): Warn when the critical and interrupt function attributes are used together.
2018-12-29 Jozef Lawrynowicz <jozef.l@mittosystems.com>
gcc/ChangeLog:
* config/msp430/msp430.c (msp430_attr): Warn when the critical and
interrupt function attributes are used together.
* gcc/doc/extend.texi: Update documentation on the critical attribute.
gcc/testsuite/ChangeLog:
* gcc.target/msp430/critical-interrupt.c: New test.
Steven G. Kargl [Sat, 29 Dec 2018 17:43:02 +0000 (17:43 +0000)]
cpp.c (gfc_cpp_init): Add pre-defined macros for INTEGER(1) INTEGER(2)...
2018-12-25 Steven G. Kargl <kargl@gcc.gnu.org>
* cpp.c (gfc_cpp_init): Add pre-defined macros for INTEGER(1)
INTEGER(2), INTEGER(8) and INTEGER(16) if supported. Add pre-defined
macros for REAL(10) and REAL(16) if available.
* gfortran.texi: Document new macros.
M ChangeLog
M gcc.dg/torture/pr25967-1.c
M gcc.dg/torture/pr25967-2.c
M gcc.dg/torture/pr68037-1.c
M gcc.dg/torture/pr68037-2.c
M gcc.dg/torture/pr68037-3.c
M gcc.dg/torture/pr68264.c
Jakub Jelinek [Sat, 29 Dec 2018 10:53:36 +0000 (11:53 +0100)]
re PR ipa/88586 (ICE: Segmentation fault (in free_lang_data_in_decl))
PR ipa/88586
* omp-low.c (lower_omp_task_reductions): Set DECL_CONTEXT on field
and ifield. Update TYPE_ALIGN from alignment of field, ifield or
bfield.
Ian Lance Taylor [Sat, 29 Dec 2018 00:07:06 +0000 (00:07 +0000)]
runtime: prevent deadlock when profiling signal arrives in stack scan
Precise stack scan needs to unwind the stack. When it is
unwinding the stack, if a profiling signal arrives, which also
does a traceback, it may deadlock in dl_iterate_phdr. Prevent
this deadlock by setting up runtime_in_callers before traceback.
Thomas Schwinge [Fri, 28 Dec 2018 11:33:54 +0000 (12:33 +0100)]
Cleanup libgomp's coalesce chunk data structures
libgomp/
* target.c (struct gomp_coalesce_chunk): New structure.
(struct gomp_coalesce_buf): Update the chunks member to use that
type. Adjust all users.
Co-Authored-By: Cesar Philippidis <cesar@codesourcery.com>
From-SVN: r267446
Tom de Vries [Fri, 28 Dec 2018 03:43:56 +0000 (03:43 +0000)]
[libbacktrace] Reduce memory usage in build_address_map
In build_address_map we allocate a unit, and then look for addresses in the
unit, which we store in the addrs vector, with the elements pointing to the
unit. However, if we cannot find addresses in the unit, the allocated unit is
not used.
Fix this by detecting if the allocated unit has been used, and reusing it
otherwise.
Tom de Vries [Fri, 28 Dec 2018 03:43:41 +0000 (03:43 +0000)]
[libbacktrace] Simplify memory management in build_address_map
In the main loop in build_address_map, we first read the abbrevs into a local
variable abbrevs, and then allocate the corresponding unit, after which we assign
the abbrevs to the unit. This results in dedicated free-upon-failure
handling for the variable, and extra code to make sure that free-upon-failure
doesn't trigger once the unit has taken ownership of the abbrevs.
Simplify this by reversing the order of abbrev reading and unit allocation,
and eliminating the abbrevs local variable.
Bootstrapped and reg-tested on x86_64.
2018-12-28 Tom de Vries <tdevries@suse.de>
* dwarf.c (build_address_map): Simplify by removing local variable
abbrevs.
Ian Lance Taylor [Fri, 28 Dec 2018 03:43:26 +0000 (03:43 +0000)]
[libbacktrace] Fix memory leak in loop in build_address_map
When failing in build_address_map, we free the unit that's currently being
handled in the loop, but the ones that already have been allocated are leaked.
Fix this by keeping track of allocated units in a vector, and releasing them
upon failure.
Also, now that we have a vector of allocated units, move the freeing upon
failure of the abbrevs associated with each unit to build_address_map, and
remove the now redundant call to free_unit_addrs_vector.
Bootstrapped and reg-tested on x86_64.
2018-12-28 Ian Lance Taylor <iant@golang.org>
Tom de Vries <tdevries@suse.de>
PR libbacktrace/88063
* dwarf.c (free_unit_addrs_vector): Remove.
(build_address_map): Keep track of allocated units in vector. Free
allocated units and corresponding abbrevs upon failure. Remove now
redundant call to free_unit_addrs_vector. Free addrs vector upon
failure. Free allocated unit vector.
Co-Authored-By: Tom de Vries <tdevries@suse.de>
From-SVN: r267443
Tom de Vries [Fri, 28 Dec 2018 03:43:15 +0000 (03:43 +0000)]
[libbacktrace] Fix memory leak in build_address_map
While upon failure in build_address_map we call free_unit_addrs_vector, this
does not actually free the addrs vector, but merely the abbrevs of the units
pointed at by the elements of the addrs vector.
Fix this by adding code to build_address_map to make sure that the addrs vector
is freed upon failure.
Bootstrapped and reg-tested on x86_64.
2018-12-28 Tom de Vries <tdevries@suse.de>
* dwarf.c (build_address_map): Free addrs vector upon failure.
* ipa-devirt.c (polymorphic_call_target_d): Add n_odr_types.
(polymorphic_call_target_hasher::hash): Hash it.
(polymorphic_call_target_hasher::equal): Compare it.
(possible_polymorphic_call_targets): Set it.
* tree.c (free_lang_data): Rebuild type inheritance graph even on
non-LTO path.
Ian Lance Taylor [Thu, 27 Dec 2018 16:31:50 +0000 (16:31 +0000)]
runtime: let ARM32 EABI personality function continue unwind when called from traceback
On ARM32 EABI, unlike other platforms, the personality function is
called during _Unwind_Backtrace (libgcc/unwind-arm-common.inc:581).
In this case, simply unwind the frame without returning any
handlers. Otherwise traceback will loop if there is a frame with
a defer on stack.