platform/upstream/gcc.git
6 years agoPR c++/69560 - wrong alignof(double) on x86.
Jason Merrill [Mon, 23 Apr 2018 20:49:38 +0000 (16:49 -0400)]
PR c++/69560 - wrong alignof(double) on x86.

CWG 1879 - Inadequate definition of alignment requirement.
* cp-tree.h (ALIGNOF_EXPR_STD_P): New.
* typeck.c (cxx_sizeof_or_alignof_type): Add std_alignof parm.
(cxx_sizeof_expr, cxx_sizeof_nowarn, cxx_alignas_expr)
(cxx_alignof_expr): Pass it.
* parser.c (cp_parser_unary_expression): Pass it.
* pt.c (tsubst_copy): Copy it.
(tsubst_copy_and_build): Pass it.
* decl.c (fold_sizeof_expr): Pass it.

From-SVN: r259578

6 years agoRegenerate .pot files.
Joseph Myers [Mon, 23 Apr 2018 20:27:02 +0000 (21:27 +0100)]
Regenerate .pot files.

gcc/po:
* gcc.pot: Regenerate.

libcpp/po:
* cpplib.pot: Regenerate.

From-SVN: r259575

6 years agore PR middle-end/85496 (internal compiler error: in emit_move_insn, at expr.c:3722)
Eric Botcazou [Mon, 23 Apr 2018 20:19:39 +0000 (20:19 +0000)]
re PR middle-end/85496 (internal compiler error: in emit_move_insn, at expr.c:3722)

PR middle-end/85496
* expr.c (store_field): In the bitfield case, if the value comes from
a function call and is returned in registers by means of a PARALLEL,
do not change the mode of the temporary unless BLKmode and VOIDmode.

From-SVN: r259574

6 years agoPR c++/85470 - wrong error with static data member.
Jakub Jelinek [Mon, 23 Apr 2018 19:11:22 +0000 (21:11 +0200)]
PR c++/85470 - wrong error with static data member.

* decl.c (check_initializer): Check DECL_INITIALIZED_IN_CLASS_P.
* typeck2.c (store_init_value): Likewise.

Co-Authored-By: Jason Merrill <jason@redhat.com>
From-SVN: r259571

6 years agoAdd myself to MAINTAINERS.
Paul Clarke [Mon, 23 Apr 2018 18:53:34 +0000 (18:53 +0000)]
Add myself to MAINTAINERS.

* MAINTAINERS (write after approval): Add myself.

From-SVN: r259570

6 years agore PR rtl-optimization/85423 (ICE in code_motion_process_successors, at sel-sched...
Andrey Belevantsev [Mon, 23 Apr 2018 15:19:06 +0000 (18:19 +0300)]
re PR rtl-optimization/85423 (ICE in code_motion_process_successors, at sel-sched.c:6403)

        PR rtl-optimization/85423

        * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
        dependencies to debug insns when the previous insn is non-debug.

        * gcc.dg/pr85423.c: New test.

From-SVN: r259563

6 years agox86/cet: Use unsigned integer to unwind shadow stack
H.J. Lu [Mon, 23 Apr 2018 12:29:00 +0000 (12:29 +0000)]
x86/cet: Use unsigned integer to unwind shadow stack

Use unsigned integer to unwind shadow stack by replacing jle with jbe and
jg with ja.

PR target/85489
* config/x86/sjlj.S (GTM_longjmp): Replace jle/jg with jbe/ja.

From-SVN: r259559

6 years ago[ARC] Fix FLS, SETI patterns.
Claudiu Zissulescu [Mon, 23 Apr 2018 09:55:14 +0000 (11:55 +0200)]
[ARC] Fix FLS, SETI patterns.

Claudiu Zissulescu  <claziss@synopsys.com>

* config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
enums into a single definition.
(fls): Fix predicates and printing.
(seti): Likewise.

From-SVN: r259558

6 years ago[ARC] Update sleep builtin.
Claudiu Zissulescu [Mon, 23 Apr 2018 09:54:59 +0000 (11:54 +0200)]
[ARC] Update sleep builtin.

gcc/
2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>

* config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
* config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
and short u6 immediate.
(check_if_valid_sleep_operand): Remove.
* config/arc/arc.md (Sleep): Accepts registers and u6 immediates.

changelog

From-SVN: r259557

6 years agoDaily bump.
GCC Administrator [Mon, 23 Apr 2018 00:16:27 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r259556

6 years ago[NDS32] Add new option -malways-save-lp.
Chung-Ju Wu [Sun, 22 Apr 2018 10:10:00 +0000 (10:10 +0000)]
[NDS32] Add new option -malways-save-lp.

gcc/
* config/nds32/nds32.c (nds32_compute_stack_frame): Consider
flag_always_save_lp condition.
* config/nds32/nds32.opt (malways-save-lp): New option.

From-SVN: r259553

6 years ago[NDS32] Implment USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT, USE_STORE_POST_DEC...
Shiva Chen [Sun, 22 Apr 2018 09:05:10 +0000 (09:05 +0000)]
[NDS32] Implment USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT, USE_STORE_POST_DECREMENT and USE_STORE_POST_INCREMENT.

gcc/
* config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
* config/nds32/nds32.c (nds32_use_load_post_increment): New.
* config/nds32/nds32.h
(USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
(USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.

From-SVN: r259552

6 years ago[NDS32] Remove unused function: nds32_ls_333_p().
Shiva Chen [Sun, 22 Apr 2018 08:51:26 +0000 (08:51 +0000)]
[NDS32] Remove unused function: nds32_ls_333_p().

gcc/
* config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
* config/nds32/nds32.c (nds32_ls_333_p): Remove.

From-SVN: r259551

6 years ago[NDS32] Refine CASE_VECTOR_SHORTEN_MODE to function.
Shiva Chen [Sun, 22 Apr 2018 08:42:18 +0000 (08:42 +0000)]
[NDS32] Refine CASE_VECTOR_SHORTEN_MODE to function.

gcc/
* config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
Declare.
* config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
* config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.

Co-Authored-By: Chung-Ju Wu <jasonwucj@gmail.com>
From-SVN: r259550

6 years ago[NDS32] Fix wrong input value of parameter from stack for variable-length function.
Chung-Ju Wu [Sun, 22 Apr 2018 08:25:47 +0000 (08:25 +0000)]
[NDS32] Fix wrong input value of parameter from stack for variable-length function.

gcc/
* config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.

From-SVN: r259549

6 years ago[NDS32] Implement DATA_ALIGNMENT, LOCAL_ALIGNMENT and TARGET_CONSTANT_ALIGNMENT.
Chung-Ju Wu [Sun, 22 Apr 2018 08:19:38 +0000 (08:19 +0000)]
[NDS32] Implement DATA_ALIGNMENT, LOCAL_ALIGNMENT and TARGET_CONSTANT_ALIGNMENT.

gcc/
* config/nds32/nds32-protos.h (nds32_data_alignment,
nds32_local_alignment): Declare.
* config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
nds32_local_alignment): New functions.
(TARGET_CONSTANT_ALIGNMENT): Define.
* config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.

From-SVN: r259548

6 years ago[NDS32] Move some target hooks to the bottom of nds32.c file.
Chung-Ju Wu [Sun, 22 Apr 2018 08:07:08 +0000 (08:07 +0000)]
[NDS32] Move some target hooks to the bottom of nds32.c file.

gcc/
* config/nds32/nds32.c
(TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
(TARGET_MODES_TIEABLE_P): Likewise.

From-SVN: r259547

6 years ago[NDS32] Consider -Ofast and -Og in nds32_asm_file_start function.
Chung-Ju Wu [Sun, 22 Apr 2018 07:57:06 +0000 (07:57 +0000)]
[NDS32] Consider -Ofast and -Og in nds32_asm_file_start function.

gcc/
* config/nds32/nds32.c (nds32_asm_file_start): Display optimization
level Ofast and Og.

From-SVN: r259546

6 years ago[NDS32] Add unaligned access support.
Monk Chiang [Sun, 22 Apr 2018 07:46:39 +0000 (07:46 +0000)]
[NDS32] Add unaligned access support.

gcc/
* config/nds32/constants.md (unspec_volatile_element): Add enum values
for unaligned access.
* config/nds32/nds32-intrinsic.c: Implementation of expanding
unaligned access.
* config/nds32/nds32-intrinsic.md: Likewise.
* config/nds32/nds32_intrinsic.h: Likewise.
* config/nds32/nds32.h (nds32_builtins): Likewise.
* config/nds32/nds32.opt (munaligned-access): New option.
* config/nds32/nds32.c (nds32_asm_file_start): Display
flag_unaligned_access status.

Co-Authored-By: Chung-Ju Wu <jasonwucj@gmail.com>
From-SVN: r259545

6 years agoDaily bump.
GCC Administrator [Sun, 22 Apr 2018 00:16:24 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r259544

6 years agoRISC-V: Pass --no-relax to linker if -mno-relax is present.
Kito Cheng [Sat, 21 Apr 2018 01:03:30 +0000 (01:03 +0000)]
RISC-V: Pass --no-relax to linker if -mno-relax is present.

gcc/
2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
* config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
-mno-relax is present.
* config/riscv/linux.h (LINK_SPEC): Ditto.

From-SVN: r259540

6 years agoDaily bump.
GCC Administrator [Sat, 21 Apr 2018 00:16:14 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r259539

6 years agoPR c/85365 - -Wrestrict false positives with -fsanitize=undefined
Martin Sebor [Fri, 20 Apr 2018 23:43:51 +0000 (23:43 +0000)]
PR c/85365 -  -Wrestrict false positives with -fsanitize=undefined

gcc/ChangeLog:

PR c/85365
* gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
for null pointers.
(gimple_fold_builtin_stxcpy_chk): Same.
* gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.

gcc/testsuite/ChangeLog:

PR c/85365
* gcc.dg/Wrestrict-15.c: New test.

From-SVN: r259535

6 years agore PR target/85456 (PowerPC: Using -mabi=ieeelongdouble calls wrong function for...
Michael Meissner [Fri, 20 Apr 2018 21:27:08 +0000 (21:27 +0000)]
re PR target/85456 (PowerPC: Using -mabi=ieeelongdouble calls wrong function for __builtin_powi.)

[libgcc]
2018-04-20  Michael Meissner  <meissner@linux.ibm.com>

PR target/85456
* config/rs6000/_powikf2.c: New file.  Add support for the
__builtin_powil function when long double is IEEE 128-bit floating
point.
* config/rs6000/float128-ifunc.c (__powikf2_resolve): Add
__powikf2 support.
(__powikf2): Likewise.
* config/rs6000/quad-float128.h (__powikf2_sw): Likewise.
(__powikf2_hw): Likewise.
(__powikf2): Likewise.
* config/rs6000/t-float128 (fp128_ppc_funcs): Likewise.
* config/rs6000/t-float128-hw (fp128_hw_func): Likewise.
(_powikf2-hw.c): Likewise.

[gcc]
2018-04-20  Michael Meissner  <meissner@linux.ibm.com>

PR target/85456
* config/rs6000/rs6000.c (init_float128_ieee): Add support to call
__powikf2 when long double is IEEE 128-bit.

[gcc/testsuite]
2018-04-20  Michael Meissner  <meissner@linux.ibm.com>

PR target/85456
* gcc.target/powerpc/pr85456.c: New test.

From-SVN: r259533

6 years agoundef-bool-1.C: Require lp64.
Bill Schmidt [Fri, 20 Apr 2018 21:00:41 +0000 (21:00 +0000)]
undef-bool-1.C: Require lp64.

[gcc/testsuite]

2018-04-20  Bill Schmidt  <wschmidt@linux.ibm.com>

* g++.dg/ext/undef-bool-1.C: Require lp64.
* gcc.target/powerpc/undef-bool-2.c: Likewise.

From-SVN: r259532

6 years agogotest: only use [TD] on big-endian PPC64 non-AIX systems
Ian Lance Taylor [Fri, 20 Apr 2018 20:40:27 +0000 (20:40 +0000)]
gotest: only use [TD] on big-endian PPC64 non-AIX systems

    Reviewed-on: https://go-review.googlesource.com/108457

From-SVN: r259531

6 years agoRISC-V: Make sure stack is always aligned during adjusting stack.
Kito Cheng [Fri, 20 Apr 2018 19:03:19 +0000 (19:03 +0000)]
RISC-V: Make sure stack is always aligned during adjusting stack.

gcc/
2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
* config/riscv/riscv.c (riscv_first_stack_step): Round up min
step to make sure stack always aligned.

From-SVN: r259530

6 years agostack-check-5.c: Improve dg-skip-if selector for the stack protector.
Jeff Law [Fri, 20 Apr 2018 17:20:46 +0000 (11:20 -0600)]
stack-check-5.c: Improve dg-skip-if selector for the stack protector.

       * gcc.dg/stack-check-5.c: Improve dg-skip-if selector for the
        stack protector.
        * gcc.dg/stack-check-6.c: Likewise.
        * gcc.dg/stack-check-6a.c: Likewise.
        * gcc.target/i386/stack-check-17.c: Add dg-skip-if selector.
        * gcc.target/i386/stack-check-18.c: Likewise.
        * gcc.target/i386/stack-check-19.c: Likewise.

From-SVN: r259528

6 years agoFix missing hunk from previous commit
Kyrylo Tkachov [Fri, 20 Apr 2018 16:33:53 +0000 (16:33 +0000)]
Fix missing hunk from previous commit

From-SVN: r259527

6 years agoPR testsuite/85483: Move aarch64/sve/vcond_1.c test to g++.dg/other/
Kyrylo Tkachov [Fri, 20 Apr 2018 16:31:19 +0000 (16:31 +0000)]
PR testsuite/85483: Move aarch64/sve/vcond_1.c test to g++.dg/other/

I totally botched up this sve test file in 259437.
It needs C++, so move it to g++.dg/other and make it a .C file.
Also adds the target guards to prevent it from running on non-aarch64 targets.

Tested that it passes on aarch64-none-elf and doesn't get run on arm-none-eabi.

Committing to trunk as obvious.

PR testsuite/85483
* gcc.target/aarch64/sve/vcond_1.c: Move to...
* g++.dg/other/sve_vcond_1.C: ... Here.  Add target directives.
* gcc.target/aarch64/sve/vcond_1_run.c: Move to...
* g++.dg/other/sve_vcond_1_run.C: ... Here.  Change include file name.

From-SVN: r259526

6 years agore PR target/83402 (PPC64 implementation of ./rs6000/emmintrin.h gives out of range...
Carl Love [Fri, 20 Apr 2018 15:18:24 +0000 (15:18 +0000)]
re PR target/83402 (PPC64 implementation of ./rs6000/emmintrin.h gives out of range for _mm_slli_epi32)

gcc/ChangeLog:

2018-04-20  Carl Love  <cel@us.ibm.com>

        PR target/83402
        * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
        size check for arg0.

From-SVN: r259524

6 years ago[nvptx] Fix calls to vector and worker routines
Nathan Sidwell [Fri, 20 Apr 2018 13:46:07 +0000 (13:46 +0000)]
[nvptx] Fix calls to vector and worker routines

2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
    Tom de Vries  <tom@codesourcery.com>

PR target/85445
* config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
Emit insns for calls too.
(nvptx_find_par): Always look for worker-level predecessor insn.
(nvptx_propagate): Add is_call parm, return bool.  Copy frame for
calls.
(nvptx_vpropagate, nvptx_wpropagate): Adjust.
(nvptx_process_pars): Propagate frames for calls.

* testsuite/libgomp.oacc-c++/ref-1.C: New.

Co-Authored-By: Tom de Vries <tom@codesourcery.com>
From-SVN: r259523

6 years agoDefine __CET__ for -fcf-protection and remove -mibt
H.J. Lu [Fri, 20 Apr 2018 13:30:13 +0000 (13:30 +0000)]
Define __CET__ for -fcf-protection and remove -mibt

With revision 259496:

commit b1384095a7c1d06a44b70853372ebe037b2f7867
Author: hjl <hjl@138bc75d-0d04-0410-961f-82ee72b054a4>
Date:   Thu Apr 19 15:15:04 2018 +0000

    x86: Enable -fcf-protection with multi-byte NOPs

-mibt does nothing and can be removed.  Define __CET__ to indicate level
protection with -fcf-protection:

(__CET__ & 1) != 0: -fcf-protection=branch or -fcf-protection=full
(__CET__ & 2) != 0: -fcf-protection=return or -fcf-protection=full

gcc/

PR target/85469
* common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
Removed.
(OPTION_MASK_ISA_IBT_UNSET): Likewise.
(ix86_handle_option): Don't handle OPT_mibt.
* config/i386/cet.h: Check __CET__ instead of __IBT__ and
__SHSTK__.
* config/i386/driver-i386.c (host_detect_local_cpu): Remove
has_ibt and ibt.
* config/i386/i386-c.c (ix86_target_macros_internal): Don't
check OPTION_MASK_ISA_IBT nor flag_cf_protection.
(ix86_target_macros): Define __CET__ with flag_cf_protection
for -fcf-protection.
* config/i386/i386.c (isa2_opts): Remove -mibt.
* config/i386/i386.h (TARGET_IBT): Removed.
(TARGET_IBT_P): Likewise.
(ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
* config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
* config/i386/i386.opt (mcet): Update help message.
(mshstk): Likewise.
(mibt): Removed.
* doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
-mcet as an alias for -mshstk.

gcc/testsuite/

PR target/85469
* gcc.target/i386/pr85044.c (dg-options): Remove -mibt.
* gcc.target/i386/sse-26.c (dg-options): Remove -mno-ibt.

From-SVN: r259522

6 years agoDo not overflow string buffer (PR objc/85476).
Martin Liska [Fri, 20 Apr 2018 12:55:36 +0000 (14:55 +0200)]
Do not overflow string buffer (PR objc/85476).

2018-04-20  Martin Liska  <mliska@suse.cz>

PR objc/85476
* objc-act.c (finish_class): Do not overflow string buffer.

From-SVN: r259521

6 years agore PR tree-optimization/85475 (Compile time hog w/ -O1 -fpeel-loops)
Richard Biener [Fri, 20 Apr 2018 10:31:23 +0000 (10:31 +0000)]
re PR tree-optimization/85475 (Compile time hog w/ -O1 -fpeel-loops)

2018-04-20  Richard Biener <rguenther@suse.de>

PR middle-end/85475
* match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
complexity by forcing a single use of the multiply operand.

* gcc.dg/torture/pr85475.c: New testcase.

From-SVN: r259519

6 years agoFix IPA-CP test for self-feeding recursive dependency
Martin Jambor [Fri, 20 Apr 2018 09:35:33 +0000 (11:35 +0200)]
Fix IPA-CP test for self-feeding recursive dependency

2018-04-20  Martin Jambor  <mjambor@suse.cz>

ipa/85449
* ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
recursion dependency to only apply to non-clones.

testsuite/
* gcc.dg/ipa/pr85449.c: New test.

From-SVN: r259518

6 years agoCheck that clones of edges exist during IPA-CP
Martin Jambor [Fri, 20 Apr 2018 09:19:39 +0000 (11:19 +0200)]
Check that clones of edges exist during IPA-CP

2018-04-20  Martin Jambor  <mjambor@suse.cz>

ipa/85447
* ipa-cp.c (create_specialized_node): Check that clones of
self-recursive edges exist during IPA-CP.

testsuite/
* g++.dg/ipa/pr85447.C: New file.
* gcc.dg/ipa/ipcp-self-recursion-1.c: Likewise.

From-SVN: r259517

6 years agore PR c++/85462 (internal compiler error: in inc_refcount_use, at cp/pt.c:8955)
Jakub Jelinek [Fri, 20 Apr 2018 07:56:52 +0000 (09:56 +0200)]
re PR c++/85462 (internal compiler error: in inc_refcount_use, at cp/pt.c:8955)

PR c++/85462
* cp-tree.h (tinst_level): Remove in_system_header_p member,
change refcount member from unsigned char to unsigned short,
add refcount_infinity static data member, adjust comments.
* pt.c (tinst_level::refcount_infinity): Define.
(inc_refcount_use): Remove assert, don't increment if refcount
is already refcount_infinity, adjust comment.
(dec_refcount_use): Remove assert, don't decrement if refcount
is refcount_infinity, adjust comment.
(push_tinst_level_loc): Formatting fix.

* g++.dg/cpp0x/pr85462.C: New test.

From-SVN: r259516

6 years agoDaily bump.
GCC Administrator [Fri, 20 Apr 2018 00:16:26 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r259515

6 years agoinvoke.texi: Add -floop-unroll-and-jam to options enabled by -O3.
Toon Moene [Thu, 19 Apr 2018 19:29:26 +0000 (19:29 +0000)]
invoke.texi: Add -floop-unroll-and-jam to options enabled by -O3.

2018-04-19  Toon Moene  <toon@moene.org>

* doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
by -O3.

From-SVN: r259509

6 years agoRequire CET target on gcc.target/i386/pr85404.c
H.J. Lu [Thu, 19 Apr 2018 19:20:37 +0000 (19:20 +0000)]
Require CET target on gcc.target/i386/pr85404.c

Need a working CET assembler to assemble the compiler output.

* gcc.target/i386/pr85404.c: Require CET target.

From-SVN: r259508

6 years agore PR tree-optimization/85467 (ICE: verify_gimple failed: non-trivial conversion...
Jakub Jelinek [Thu, 19 Apr 2018 19:16:18 +0000 (21:16 +0200)]
re PR tree-optimization/85467 (ICE: verify_gimple failed: non-trivial conversion at assignment with -O2 -fno-tree-ccp --param=sccvn-max-scc-size=10)

PR tree-optimization/85467
* fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
VECTOR_CST element to type.

* gcc.dg/pr85467.c: New test.

From-SVN: r259507

6 years agore PR c++/84611 (ICE in operator[], at vec.h:826 (local_class_index()))
Paolo Carlini [Thu, 19 Apr 2018 18:09:45 +0000 (18:09 +0000)]
re PR c++/84611 (ICE in operator[], at vec.h:826 (local_class_index()))

/cp
2018-04-19  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/84611
* pt.c (lookup_template_class_1): Check pushtag return value for
error_mark_node.

/testsuite
2018-04-19  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/84611
* g++.dg/parse/crash68.C: New.

From-SVN: r259505

6 years agolibgcc/CET: Skip signal frames when unwinding shadow stack
H.J. Lu [Thu, 19 Apr 2018 17:05:39 +0000 (17:05 +0000)]
libgcc/CET: Skip signal frames when unwinding shadow stack

When -fcf-protection -mcet is used, I got

FAIL: g++.dg/eh/sighandle.C

(gdb) bt
 #0  _Unwind_RaiseException (exc=exc@entry=0x416ed0)
    at /export/gnu/import/git/sources/gcc/libgcc/unwind.inc:140
 #1  0x00007ffff7d9936b in __cxxabiv1::__cxa_throw (obj=<optimized out>,
    tinfo=0x403dd0 <typeinfo for int@@CXXABI_1.3>, dest=0x0)
    at /export/gnu/import/git/sources/gcc/libstdc++-v3/libsupc++/eh_throw.cc:90
 #2  0x0000000000401255 in sighandler (signo=11, si=0x7fffffffd6f8,
    uc=0x7fffffffd5c0)
    at /export/gnu/import/git/sources/gcc/gcc/testsuite/g++.dg/eh/sighandle.C:9
 #3  <signal handler called> <<<< Signal frame which isn't on shadow stack
 #4  dosegv ()
    at /export/gnu/import/git/sources/gcc/gcc/testsuite/g++.dg/eh/sighandle.C:14
 #5  0x00000000004012e3 in main ()
    at /export/gnu/import/git/sources/gcc/gcc/testsuite/g++.dg/eh/sighandle.C:30
(gdb) p frames
$6 = 5
(gdb)

frame count should be 4, not 5.  This patch skips signal frames when
unwinding shadow stack.

gcc/testsuite/

PR libgcc/85334
* g++.dg/torture/pr85334.C: New test.

libgcc/

PR libgcc/85334
* unwind-generic.h (_Unwind_Frames_Increment): New.
* config/i386/shadow-stack-unwind.h (_Unwind_Frames_Increment):
Likewise.
* unwind.inc (_Unwind_RaiseException_Phase2): Increment frame
count with _Unwind_Frames_Increment.
(_Unwind_ForcedUnwind_Phase2): Likewise.

From-SVN: r259502

6 years agoPR c++/85464 - missing location for -Wignored-qualifiers diagnostic
Jonathan Wakely [Thu, 19 Apr 2018 17:00:37 +0000 (18:00 +0100)]
PR c++/85464 - missing location for -Wignored-qualifiers diagnostic

* g++.dg/diagnostic/pr85464.C: New.

From-SVN: r259501

6 years agoi386: Add save_stack_nonlocal and restore_stack_nonlocal
H.J. Lu [Thu, 19 Apr 2018 16:36:34 +0000 (16:36 +0000)]
i386: Add save_stack_nonlocal and restore_stack_nonlocal

Define STACK_SAVEAREA_MODE to hold both shadow stack and stack pointers.
Replace builtin_setjmp_setup and builtin_longjmp with save_stack_nonlocal
and restore_stack_nonlocal to support both builtin setjmp/longjmp as well
as non-local goto in nested functions.

gcc/

PR target/85397
* config/i386/i386.h (STACK_SAVEAREA_MODE): New.
* config/i386/i386.md (builtin_setjmp_setup): Removed.
(builtin_longjmp): Likewise.
(save_stack_nonlocal): New pattern.
(restore_stack_nonlocal): Likewise.

gcc/testsuite/

PR target/85397
* gcc.dg/torture/pr85397-1.c: New test.
* gcc.target/i386/cet-sjlj-6a.c: Adjusted.
* gcc.target/i386/cet-sjlj-6b.c: Likewise.

From-SVN: r259500

6 years agox86/cet: Properly output labels in property note section
H.J. Lu [Thu, 19 Apr 2018 15:24:53 +0000 (15:24 +0000)]
x86/cet: Properly output labels in property note section

Replace ASM_OUTPUT_LABEL with fprintf so that internal labels in property
note section are unchanged by -fleading-underscore.

gcc/

PR target/85404
* config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
Replace ASM_OUTPUT_LABEL with fprintf.

gcc/testsuite/

PR target/85404
* gcc.target/i386/pr85404.c: New test.

From-SVN: r259498

6 years agolibgcc/CET: Add _CET_ENDBR to __stack_split_initialize
H.J. Lu [Thu, 19 Apr 2018 15:22:27 +0000 (15:22 +0000)]
libgcc/CET: Add _CET_ENDBR to __stack_split_initialize

Program received signal SIGSEGV, Segmentation fault.
__stack_split_initialize ()
    at /export/gnu/import/git/sources/gcc/libgcc/config/i386/morestack.S:751
751 leaq -16000(%rsp),%rax # We should have at least 16K.
Missing separate debuginfos, use: dnf debuginfo-install libgcc-8.0.1-0.21.0.fc28.x86_64
(gdb) disass
Dump of assembler code for function __stack_split_initialize:
=> 0x0000000000402858 <+0>: lea    -0x3e80(%rsp),%rax
   0x0000000000402860 <+8>: mov    %rax,%fs:0x70
   0x0000000000402869 <+17>: sub    $0x8,%rsp
   0x000000000040286d <+21>: mov    %rsp,%rdi
   0x0000000000402870 <+24>: mov    $0x3e80,%esi
   0x0000000000402875 <+29>: callq  0x401810 <__generic_morestack_set_initial_sp>
   0x000000000040287a <+34>: add    $0x8,%rsp
   0x000000000040287e <+38>: retq
End of assembler dump.
(gdb)

This patch adds the missing ENDBR to __stack_split_initialize.

PR libgcc/85379
* config/i386/morestack.S (__stack_split_initialize): Add
_CET_ENDBR.

From-SVN: r259497

6 years agox86: Enable -fcf-protection with multi-byte NOPs
H.J. Lu [Thu, 19 Apr 2018 15:15:04 +0000 (15:15 +0000)]
x86: Enable -fcf-protection with multi-byte NOPs

-fcf-protection -mcet can't be used with IFUNC features, like symbol
multiversioning or target clone, since IBT/SHSTK are applied to the whole
program and they may be disabled in some functions.  But -fcf-protection
is implemented with multi-byte NOPs on all 64-bit processors as well as
32-bit processors starting with Pentium Pro.  If -fcf-protection requires
-mcet, IFUNC features can't be used on Linux when -fcf-protection is
enabled by default.

This patch changes -fcf-protection to implement indirect branch and
return address tracking with multi-byte NOPs.  -mibt and -mshstk are
changed to only enable CET built-in functions.  CET tests are updated
to allow -fcf-protection without -mibt, -mshstk and -mcet on x86.
-fcf-protection=none are also added to tests which fail with
-fcf-protection so that -fcf-protection can be added to RUNTESTFLAGS
to verify -fcf-protection implementation.

gcc/

PR target/85417
* config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
* config/i386/i386-c.c (ix86_target_macros_internal): Also
define __IBT__ and __SHSTK__ for -fcf-protection.
* config/i386/i386.c (pass_insert_endbranch::gate): Don't check
TARGET_IBT.
(ix86_trampoline_init): Likewise.
(x86_output_mi_thunk): Likewise.
(ix86_notrack_prefixed_insn_p): Likewise.
(ix86_option_override_internal): Don't disallow -fcf-protection.
* config/i386/i386.md (rdssp<mode>): Also enable for
-fcf-protection.
(incssp<mode>): Likewise.
(nop_endbr): Likewise.
* config/i386/i386.opt (mcet): Change help message to built-in
functions only.
(mibt): Likewise.
(mshstk): Likewise.
* doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
enable CET built-in functions.

gcc/testsuite/

PR target/85417
* c-c++-common/attr-nocf-check-1.c: Compile with
-fcf-protection=none.
* c-c++-common/attr-nocf-check-3.c: Likewise.
* gcc.dg/march-generic.c: Likewise.
* gcc.target/i386/align-limit.c: Likewise.
* gcc.target/i386/cet-notrack-icf-1.c: Likewise.
* gcc.target/i386/cet-notrack-icf-3.c: Likewise.
* gcc.target/i386/cet-property-2.c: Likewise.
* gcc.target/i386/ret-thunk-26.c: Likewise.
* c-c++-common/fcf-protection-1.c: Remove dg-error for x86
targets.
* c-c++-common/fcf-protection-2.c: Likewise.
* c-c++-common/fcf-protection-3.c: Likewise.
* c-c++-common/fcf-protection-5.c: Likewise.
* c-c++-common/fcf-protection-6.c: Likewise.
* c-c++-common/fcf-protection-7.c: Likewise.
* gcc.target/i386/cet-label-3.c: New test.
* gcc.target/i386/cet-property-3.c: Likewise.
* gcc.target/i386/cet-sjlj-7.c: Likewise.
* gcc.target/i386/pr85417-1.c: Likewise.
* gcc.target/i386/indirect-thunk-attr-7.c: Also expect
__x86_indirect_thunk_nt_(r|e)ax
* gcc.target/i386/indirect-thunk-extern-7.c: Likewise.
* gcc.target/i386/pr85403.c: Remove dg-error,

From-SVN: r259496

6 years agoi386-common.c (OPTION_MASK_ISA_MOVDIRI_SET, [...]): New defines.
Sebastian Peryt [Thu, 19 Apr 2018 15:08:02 +0000 (17:08 +0200)]
i386-common.c (OPTION_MASK_ISA_MOVDIRI_SET, [...]): New defines.

2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>

gcc/

* common/config/i386/i386-common.c
(OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
OPTION_MASK_ISA_MOVDIRI_UNSET,
OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
(ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
* config.gcc (movdirintrin.h): New header.
* config/i386/cpuid.h (bit_MOVDIRI,
bit_MOVDIR64B): New bits.
* config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
and -mmvodir64b.
* config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
(VOID, PVOID, PCVOID)): New function types.
* config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
__builtin_ia32_directstoreu_u64,
__builtin_ia32_movdir64b): New builtins.
* config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
* config/i386/i386.c (ix86_target_string): Added -mmovdir64b
and -mmovdiri.
(ix86_valid_target_attribute_inner_p): Ditto.
(ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
and VOID_FTYPE_PUNSIGNED_UNSIGNED.
(ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
* config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
* config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
(movdiri<mode>, movdir64b_<mode>): New.
* config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
* config/i386/immintrin.h: Include movdirintrin.h.
* config/i386/movdirintrin.h: New file.
* doc/invoke.texi: Added -mmovdiri and -mmovdir64b.

gcc/testsuite/

* gcc.target/i386/movdir-1.c: New test.

From-SVN: r259495

6 years agore PR rtl-optimization/85455 (ICE in verify_loop_structure, at cfgloop.c:1708 (error...
Richard Biener [Thu, 19 Apr 2018 13:53:06 +0000 (13:53 +0000)]
re PR rtl-optimization/85455 (ICE in verify_loop_structure, at cfgloop.c:1708 (error: basic block 3 should be marked irreducible))

2018-04-19  Richard Biener  <rguenther@suse.de>

PR middle-end/85455
* cfg.c (clear_bb_flags): When loop state says we have
marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.

* gcc.dg/pr85455.c: New testcase.

From-SVN: r259494

6 years agore PR tree-optimization/84737 (20% degradation in CPU2000 172.mgrid starting with...
Richard Biener [Thu, 19 Apr 2018 12:41:42 +0000 (12:41 +0000)]
re PR tree-optimization/84737 (20% degradation in CPU2000 172.mgrid starting with r256888)

2018-04-19  Richard Biener  <rguenther@suse.de>

PR tree-optimization/84737
* tree-vect-data-refs.c (vect_copy_ref_info): New function
copying restrict info.
(vect_setup_realignment): Use it.
* tree-vectorizer.h (vect_copy_ref_info): Declare.
* tree-vect-stmts.c (vectorizable_store): Copy ref info from
the first DR to all generated stores.
(vectorizable_load): Likewise for loads.

From-SVN: r259493

6 years ago[testsuite] Fix tests for pr84805 and pr85405.
Christophe Lyon [Thu, 19 Apr 2018 08:59:36 +0000 (08:59 +0000)]
[testsuite] Fix tests for pr84805 and pr85405.

2018-04-19  Christophe Lyon  <christophe.lyon@linaro.org>

* g++.dg/lto/pr85405_0.C: Require shared and fpic effective
targets.
* g++.dg/lto/pr85405b_0.C: Likewise.
* g++.dg/lto/pr84805_0.C: Likewise.

From-SVN: r259492

6 years agoPR85463 '[nvptx] "exit" in offloaded region doesn't terminate process'
Thomas Schwinge [Thu, 19 Apr 2018 08:53:38 +0000 (10:53 +0200)]
PR85463 '[nvptx] "exit" in offloaded region doesn't terminate process'

libgomp/
PR libfortran/85166
* testsuite/libgomp.oacc-fortran/abort-1.f90: Switch back to "call
abort".
* testsuite/libgomp.oacc-fortran/abort-2.f90: Likewise.

libgfortran/
PR libfortran/85166
PR libgomp/85463
* runtime/minimal.c (stop_numeric): Reimplement.
(stop_string, error_stop_string, error_stop_numeric): New
functions.
libgomp/
PR libgomp/85463
* testsuite/libgomp.oacc-fortran/error_stop-1.f: New file.
* testsuite/libgomp.oacc-fortran/error_stop-2.f: Likewise.
* testsuite/libgomp.oacc-fortran/error_stop-3.f: Likewise.
* testsuite/libgomp.oacc-fortran/stop-1.f: Likewise.
* testsuite/libgomp.oacc-fortran/stop-2.f: Likewise.
* testsuite/libgomp.oacc-fortran/stop-3.f: Likewise.

From-SVN: r259491

6 years agoDo not bail out for multiple PREVAILING_DEF_IRONLY for common symbols.
Martin Liska [Thu, 19 Apr 2018 08:42:52 +0000 (10:42 +0200)]
Do not bail out for multiple PREVAILING_DEF_IRONLY for common symbols.

2018-04-19  Martin Liska  <mliska@suse.cz>

* lto-symtab.c (lto_symtab_resolve_symbols): Do not bail out
for multiple PREVAILING_DEF_IRONLY for common symbols.

From-SVN: r259490

6 years agore PR tree-optimization/85446 (wrong-code on riscv64)
Jakub Jelinek [Thu, 19 Apr 2018 07:46:54 +0000 (09:46 +0200)]
re PR tree-optimization/85446 (wrong-code on riscv64)

PR tree-optimization/85446
* match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
the integral and pointer types to have the same precision.

From-SVN: r259488

6 years agocet.m4 (GCC_CET_FLAGS): Default to --disable-cet, replace --enable-cet=default with...
Jakub Jelinek [Thu, 19 Apr 2018 07:45:51 +0000 (09:45 +0200)]
cet.m4 (GCC_CET_FLAGS): Default to --disable-cet, replace --enable-cet=default with --enable-cet=auto.

* config/cet.m4 (GCC_CET_FLAGS): Default to --disable-cet, replace
--enable-cet=default with --enable-cet=auto.

* doc/install.texi: Document --disable-cet being the default and
--enable-cet=auto.

* configure: Regenerated.

From-SVN: r259487

6 years agore PR c++/80290 (g++ uses unreasonable amount of memory compiling nested string maps)
Alexandre Oliva [Thu, 19 Apr 2018 06:39:06 +0000 (06:39 +0000)]
re PR c++/80290 (g++ uses unreasonable amount of memory compiling nested string maps)

PR c++/80290
* cp-tree.h (tinst_level::free): Fix whitespace.

From-SVN: r259486

6 years agoDaily bump.
GCC Administrator [Thu, 19 Apr 2018 00:16:23 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r259483

6 years agoFix coding style and typos.
Martin Liska [Wed, 18 Apr 2018 20:09:33 +0000 (22:09 +0200)]
Fix coding style and typos.

2018-04-18  Martin Liska  <mliska@suse.cz>

* ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
style.
2018-04-18  Martin Liska  <mliska@suse.cz>

* lto.c (lto_read_decls): Fix GNU coding style and typos.

From-SVN: r259480

6 years agoMake Wodr warnings stable.
Martin Liska [Wed, 18 Apr 2018 20:08:44 +0000 (22:08 +0200)]
Make Wodr warnings stable.

2018-04-18  Martin Liska  <mliska@suse.cz>

PR ipa/83983
PR ipa/85391
* lto.c (cmp_type_location): New function.
(lto_read_decls): First collect all types, then
sort them according by location before register_odr_type
is called.
2018-04-18  Martin Liska  <mliska@suse.cz>

PR ipa/83983
PR ipa/85391
* g++.dg/lto/pr83121_1.C (struct Environment): Adjust expected
output.

From-SVN: r259479

6 years agoRevert r258133.
Martin Liska [Wed, 18 Apr 2018 20:06:07 +0000 (22:06 +0200)]
Revert r258133.

2018-04-18  Martin Liska  <mliska@suse.cz>

Revert
2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>

PR ipa/83983
* ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
arguments if they are comparable.

From-SVN: r259478

6 years agoRevert r25841.
Martin Liska [Wed, 18 Apr 2018 20:05:10 +0000 (22:05 +0200)]
Revert r25841.

2018-04-18  Martin Liska  <mliska@suse.cz>

Revert
2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>

PR lto/84805
* ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
incomplete types.

From-SVN: r259477

6 years agoi386: Insert ENDBR after __morestack call
H.J. Lu [Wed, 18 Apr 2018 19:14:28 +0000 (19:14 +0000)]
i386: Insert ENDBR after __morestack call

Since __morestack will jump back to its callee via indirect call, we
need to insert ENDBR after calling __morestack.

gcc/

PR target/85388
* config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
ENDBR after calling __morestack.

gcc/testsuite/

PR target/85388
* gcc.dg/pr85388-1.c: New test.
* gcc.dg/pr85388-2.c: Likewise.
* gcc.dg/pr85388-3.c: Likewise.
* gcc.dg/pr85388-4.c: Likewise.
* gcc.dg/pr85388-5.c: Likewise.
* gcc.dg/pr85388-6.c: Likewise.

From-SVN: r259475

6 years agoPR libstdc++/85442 fix duplicate debuginfo for cxx11-ios_failure.cc
Jonathan Wakely [Wed, 18 Apr 2018 12:14:58 +0000 (13:14 +0100)]
PR libstdc++/85442 fix duplicate debuginfo for cxx11-ios_failure.cc

2018-04-18  Jonathan Wakely  <jwakely@redhat.com>
    Jakub Jelinek  <jakub@redhat.com>

PR libstdc++/85442
* src/c++11/Makefile.am: Don't generate debuginfo again for
cxx11-ios_failure-lt.s and cxx11-ios_failure.s files.
* src/c++11/Makefile.in: Regenerate.

Co-Authored-By: Jakub Jelinek <jakub@redhat.com>
From-SVN: r259467

6 years agore PR lto/85391 (ICE in add_type_duplicate, at ipa-devirt.c:1887)
Jan Hubicka [Wed, 18 Apr 2018 11:29:26 +0000 (13:29 +0200)]
re PR lto/85391 (ICE in add_type_duplicate, at ipa-devirt.c:1887)

PR lto/85391
* lto.c (lto_read_decls): Do not test TYPE_CANONICAL before registering odr
types.
* g++.dg/lto/pr83121_0.C: Update template.
* g++.dg/lto/pr83121_1.C: Update template.
* g++.dg/lto/pr84805_0.C: Update template.
* g++.dg/lto/pr84805_1.C: Update template.
* g++.dg/lto/pr84805_2.C: Update template.

Co-Authored-By: Martin Liska <mliska@suse.cz>
From-SVN: r259464

6 years agoPR libstdc++/84442 if _Exit isn't declared then use _exit instead
Jonathan Wakely [Wed, 18 Apr 2018 11:15:38 +0000 (12:15 +0100)]
PR libstdc++/84442 if _Exit isn't declared then use _exit instead

PR libstdc++/84442
* testsuite/30_threads/thread/cons/terminate.cc
[!_GLIBCXX_USE_C99_STDLIB] : Use _exit or std::exit instead of _Exit.

From-SVN: r259463

6 years agore PR jit/85384 (libgccjit does not work if --with-gcc-major-version is used)
David Malcolm [Wed, 18 Apr 2018 09:46:58 +0000 (09:46 +0000)]
re PR jit/85384 (libgccjit does not work if --with-gcc-major-version is used)

PR jit/85384
* acx.m4 (GCC_BASE_VER): Remove \$\$ from sed expression.

* configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
by using gcc_base_ver to generate a gcc_driver_version, and use
it when generating GCC_DRIVER_NAME.
* configure: Regenerate.

* configure: Regenerate.

From-SVN: r259462

6 years agore PR target/81084 (powerpcspe port full of confusing configury / command-line option...
Jakub Jelinek [Wed, 18 Apr 2018 08:47:26 +0000 (10:47 +0200)]
re PR target/81084 (powerpcspe port full of confusing configury / command-line options not related to SPE)

PR target/81084
* config.gcc: Obsolete powerpc*-*-*spe*.

From-SVN: r259461

6 years agore PR c++/84630 (ICE: TYPE_NAME() used on error_mark_node in tsubst_lambda_expr,...
Paolo Carlini [Wed, 18 Apr 2018 08:03:45 +0000 (08:03 +0000)]
re PR c++/84630 (ICE: TYPE_NAME() used on error_mark_node in tsubst_lambda_expr, at cp/pt.c:17141)

/cp
2018-04-18  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/84630
* pt.c (tsubst_lambda_expr): Check begin_lambda_type return value
for error_mark_node.

/testsuite
2018-04-18  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/84630
* g++.dg/cpp0x/pr84630.C: New.

From-SVN: r259460

6 years agore PR sanitizer/85389 (posix_memalign() crash with address sanitizer when passing...
Bill Seurer [Wed, 18 Apr 2018 07:02:40 +0000 (07:02 +0000)]
re PR sanitizer/85389 (posix_memalign() crash with address sanitizer when passing invalid arguments)

PR sanitizer/85389
* asan/asan_allocator.h (kAllocatorSpace): For __powerpc64__ change
from 0xa0000000000ULL to ~(uptr)0.

From-SVN: r259459

6 years agore PR c++/84463 (Supposedly-incompliant "error: '* key0' is not a constant expression")
Jakub Jelinek [Wed, 18 Apr 2018 06:57:45 +0000 (08:57 +0200)]
re PR c++/84463 (Supposedly-incompliant "error: '* key0' is not a constant expression")

PR c++/84463
* typeck.c (cp_build_addr_expr_1): Move handling of offsetof-like
tricks from here to ...
* cp-gimplify.c (cp_fold) <case ADDR_EXPR>: ... here.  Only use it
if INDIRECT_REF's operand is INTEGER_CST cast to pointer type.

* g++.dg/cpp0x/constexpr-nullptr-1.C: Add -O1 to dg-options.
* g++.dg/cpp0x/constexpr-nullptr-2.C: Expect different diagnostics
in two cases.  Uncomment two other tests and add expected dg-error for
them.
* g++.dg/init/struct2.C: Cast to int rather than long to avoid
-Wnarrowing diagnostics on some targets for c++11.
* g++.dg/parse/array-size2.C: Remove xfail.
* g++.dg/cpp0x/constexpr-84463.C: New test.

From-SVN: r259458

6 years ago[PR c++/80290] recycle tinst garbage sooner
Alexandre Oliva [Wed, 18 Apr 2018 05:17:26 +0000 (05:17 +0000)]
[PR c++/80290] recycle tinst garbage sooner

tinst_level objects are created during template instantiation, and
they're most often quite short-lived, but since there's no intervening
garbage collection, they accumulate throughout the pass while most by
far could be recycled after a short while.  The original testcase in
PR80290, for example, creates almost 55 million tinst_level objects,
all but 10 thousand of them without intervening GC, but we don't need
more than 284 of them live at a time.

Furthermore, in many cases, TREE_LIST objects are created to stand for
the decl in tinst_level.  In most cases, those can be recycled as soon
as the tinst_level object is recycled; in some relatively common
cases, they are modified and reused in up to 3 tinst_level objects.
In the same testcase, TREE_LISTs are used in all but 3 thousand of the
tinst_level objects, and we don't need more than 89 tinst_level
objects with TREE_LISTs live at a time.  Furthermore, all but 2
thousand of those are created without intervening GC.

This patch arranges for tinst_level objects to be refcounted (I've
seen as many as 20 live references to a single tinst_level object in
my testing), and for pending_template, tinst_level and TREE_LIST
objects that can be recycled to be added to freelists; that's much
faster than ggc_free()ing them and reallocating them shortly
thereafter.  In fact, the introduction of such freelists is what makes
this entire patch lighter-weight, when it comes to memory use, and
faster.  With refcounting alone, the total memory footprint was still
about the same, and we spent more time.

In order to further reduce memory use, I've arranged for us to create
TREE_LISTs lazily, only at points that really need them (when printing
error messages).  This grows tinst_level by an additional pointer, but
since a TREE_LIST holds a lot more than an extra pointer, and
tinst_levels with TREE_LISTs used to be allocated tens of thousands of
times more often than plain decl ones, we still save memory overall.

I was still not quite happy about growing memory use in cases that
used template classes but not template overload resolution, so I
changed the type of the errors field to unsigned short, from int.
With that change, in_system_header_p and refcount move into the same
word or half-word that used to hold errors, releasing a full word,
bringing tinst_level back to its original size, now without any
padding.

The errors field is only used to test whether we've had any errors
since the expansion of some template, to skip the expansion of further
templates.  If we get 2^16 errors or more, it is probably reasonable
to refrain from expanding further templates, even if we would expand
them before this change.

With these changes, compile time for the original testcase at -O0,
with default checking enabled, is cut down by some 3.7%, total GCable
memory allocation is cut down by almost 20%, and total memory use (as
reported by GNU time as maxresident) is cut down by slightly over 15%.

for  gcc/cp/ChangeLog

PR c++/80290
* cp-tree.h (struct tinst_level): Split decl into tldcl and
targs.  Add private split_list_p, tree_list_p, and not_list_p
inline const predicates; to_list private member function
declaration; free public member function declaration; list_p,
get_node and maybe_get_node accessors, and refcount data
member.  Narrow errors to unsigned short.
* error.c (print_instantiation_full_context): Use new
accessors.
(print_instantiation_partial_context_line): Likewise.  Drop
const from tinst_level-typed parameter.
* mangle.c (mangle_decl_string): Likewise.
* pt.c (freelist): New template class.
(tree_list_freelist_head): New var.
(tree_list_freelist): New fn, along with specializations.
(tinst_level_freelist_head): New var.
(pending_template_freelist_head): Likewise.
(tinst_level_freelist, pending_template_freelist): New fns.
(tinst_level::to_list, tinst_level::free): Define.
(inc_refcount_use, dec_refcount_use): New fns for tinst_level.
(set_refcount_ptr): New template fn.
(add_pending_template): Adjust for refcounting, freelists and
new accessors.
(neglectable_inst_p): Take a NULL d as a non-DECL.
(limit_bad_template_recursion): Use new accessors.
(push_tinst_level): New overload to create the list.
(push_tinst_level_loc): Make it static, split decl into two
args, adjust tests and initialization to cope with split
lists, use freelist, adjust for refcounting.
(push_tinst_level_loc): New wrapper with the old interface.
(pop_tinst_level): Adjust for refcounting.
(record_last_problematic_instantiation): Likewise.
(reopen_tinst_level): Likewise.  Use new accessors.
(instantiate_alias_template): Adjust for split list.
(fn_type_unification): Likewise.
(get_partial_spec_bindings): Likewise.
(instantiate_pending_templates): Use new accessors.  Adjust
for refcount.  Release pending_template to freelist.
(instantiating_current_function_p): Use new accessors.

From-SVN: r259457

6 years agoundef-bool-2.c: Add -mvsx.
Bill Schmidt [Wed, 18 Apr 2018 01:56:59 +0000 (01:56 +0000)]
undef-bool-2.c: Add -mvsx.

[gcc/testsuite]

2018-04-17  Bill Schmidt  <wschmidt@linux.ibm.com>

* gcc.target/powerpc/undef-bool-2.c: Add -mvsx.
* gcc.target/g++.dg/ext/undef-bool-1.C: Likewise.

From-SVN: r259456

6 years agoDaily bump.
GCC Administrator [Wed, 18 Apr 2018 00:16:22 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r259455

6 years agoos/signal: disable loading of history during test
Ian Lance Taylor [Tue, 17 Apr 2018 23:55:17 +0000 (23:55 +0000)]
os/signal: disable loading of history during test

    Bring in https://golang.org/cl/98616 from gc tip.

    Original CL description:

        This change modifies Go to disable loading of users' shell history for
        TestTerminalSignal tests. TestTerminalSignal, as part of its workload,
        will execute a new interactive bash shell. Bash will attempt to load the
        user's history from the file pointed to by the HISTFILE environment
        variable. For users with large histories that may take up to several
        seconds, pushing the whole test past the 5 second timeout and causing
        it to fail.

    Reviewed-on: https://go-review.googlesource.com/107624

From-SVN: r259452

6 years agore PR debug/84637 (gcc/dbxout.c:684:14: runtime error: negation of -92233720368547758...
Jakub Jelinek [Tue, 17 Apr 2018 22:18:47 +0000 (00:18 +0200)]
re PR debug/84637 (gcc/dbxout.c:684:14: runtime error: negation of -9223372036854775808 cannot be represented in type 'long int'; cast to an unsigned type to negate this value to itself)

PR debug/84637
* dbxout.c (dbxout_int): Perform negation in unsigned int type.
(stabstr_D): Change type of unum from unsigned int to
unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
type.

From-SVN: r259451

6 years agoRISC-V: Fix 32-bit stack pointer alignment problem.
Jim Wilson [Tue, 17 Apr 2018 21:41:07 +0000 (21:41 +0000)]
RISC-V: Fix 32-bit stack pointer alignment problem.

gcc/
PR 84856
* config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
Set arg_pointer_offset after using pretend_args_size.

From-SVN: r259449

6 years agore PR rtl-optimization/85431 (UBSAN: ../../gcc/dse.c:303:15: runtime error: shift...
Jakub Jelinek [Tue, 17 Apr 2018 21:38:45 +0000 (23:38 +0200)]
re PR rtl-optimization/85431 (UBSAN: ../../gcc/dse.c:303:15: runtime error: shift exponent 64 is too large for 64-bit type 'long unsigned int')

PR rtl-optimization/85431
* dse.c (record_store): Ignore zero width stores.

From-SVN: r259448

6 years agore PR testsuite/85326 (`make check` fails with `--disable-bootstrap` and `--enable...
Jakub Jelinek [Tue, 17 Apr 2018 20:43:49 +0000 (22:43 +0200)]
re PR testsuite/85326 (`make check` fails with `--disable-bootstrap` and `--enable-languages=c`)

PR testsuite/85326
* g++.dg/other/pr81422.C: Require effective target tls and c++11.
* g++.dg/other/pr60675.C: Likewise.  Remove -std=c++11 from dg-options.
* g++.dg/other/sve_tls_2.C: Require effective target tls.

From-SVN: r259447

6 years agore PR sanitizer/85230 (asan: false positives in kernel on allocas)
Jakub Jelinek [Tue, 17 Apr 2018 20:22:50 +0000 (22:22 +0200)]
re PR sanitizer/85230 (asan: false positives in kernel on allocas)

PR sanitizer/85230
* asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
__asan_allocas_unpoison call and last_alloca_addr = new_sp before
__builtin_stack_restore rather than after it.
* builtins.c (expand_asan_emit_allocas_unpoison): Pass
arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
argument instead of virtual_dynamic_stack_rtx.

From-SVN: r259446

6 years agogccgo: suppress "ar rcD" and "-zdefs" on AIX
Ian Lance Taylor [Tue, 17 Apr 2018 20:10:49 +0000 (20:10 +0000)]
gccgo: suppress "ar rcD" and "-zdefs" on AIX

    Reviewed-on: https://go-review.googlesource.com/100955

From-SVN: r259445

6 years agoUpdate gcc .po files.
Joseph Myers [Tue, 17 Apr 2018 19:25:30 +0000 (20:25 +0100)]
Update gcc .po files.

* be.po, da.po, de.po, el.po, es.po, fi.po, fr.po, hr.po, id.po,
ja.po, nl.po, ru.po, sr.po, sv.po, tr.po, uk.po, vi.po, zh_CN.po,
zh_TW.po: Update.

From-SVN: r259443

6 years agors6000-protos.h (rs6000_builtin_is_supported_p): New prototype.
Kelvin Nilsen [Tue, 17 Apr 2018 18:58:08 +0000 (18:58 +0000)]
rs6000-protos.h (rs6000_builtin_is_supported_p): New prototype.

gcc/ChangeLog:

2018-04-13  Kelvin Nilsen  <kelvin@gcc.gnu.org>

* config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
New prototype.
* config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
Add note to error message to explain internal mapping of overloaded
built-in function name to non-overloaded built-in function name.
* config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
function.

gcc/testsuite/ChangeLog:

2018-04-13  Kelvin Nilsen  <kelvin@gcc.gnu.org>

* gcc.target/powerpc/bfp/scalar-extract-sig-5.c: Simplify to
prevent cascading of errors and change expected error message.
* gcc.target/powerpc/bfp/scalar-test-neg-4.c: Restrict this test
to 64-bit targets.
* gcc.target/powerpc/bfp/scalar-test-data-class-8.c: Likewise.
* gcc.target/powerpc/bfp/scalar-test-data-class-9.c: Likewise.
* gcc.target/powerpc/bfp/scalar-test-data-class-10.c: Likewise.
* gcc.target/powerpc/bfp/scalar-insert-exp-11.c: Change expected
error message.
* gcc.target/powerpc/bfp/scalar-extract-exp-5.c: Likewise.

From-SVN: r259442

6 years agore PR target/85424 (The __builtin_packlongdouble function might have issues with...
Michael Meissner [Tue, 17 Apr 2018 18:22:08 +0000 (18:22 +0000)]
re PR target/85424 (The __builtin_packlongdouble function might have issues with the output overlapping the inputs)

2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>

PR target/85424
* config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
where the inputs overlap with the output.

From-SVN: r259441

6 years agobacktrace.c: Revert last two changes.
Ian Lance Taylor [Tue, 17 Apr 2018 17:58:05 +0000 (17:58 +0000)]
backtrace.c: Revert last two changes.

* backtrace.c: Revert last two changes.  Don't call mmap
directly.

From-SVN: r259440

6 years agobacktrace.c: Include backtrace-supported.h before checking BACKTRACE_USES_MALLOC.
Ian Lance Taylor [Tue, 17 Apr 2018 17:29:27 +0000 (17:29 +0000)]
backtrace.c: Include backtrace-supported.h before checking BACKTRACE_USES_MALLOC.

* backtrace.c: Include backtrace-supported.h before checking
BACKTRACE_USES_MALLOC.

From-SVN: r259439

6 years agosse.md (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and explicit ...
Jakub Jelinek [Tue, 17 Apr 2018 17:18:20 +0000 (19:18 +0200)]
sse.md (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and explicit "memory" attribute.

* config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
(=v, v) alternative and explicit "memory" attribute.
(vec_extract_lo_<mode><mask_name>): Likewise.  Also add
"type", "prefix", "prefix_extra", "length_immediate" and "mode"
attributes.
(vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
"sselog1" type instead of "sselog".
(vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
"sselog".  Remove explicit "memory" attribute.
(vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
"type", "prefix", "prefix_extra", "length_immediate" and "mode"
attributes.
(vec_extract_hi_v32hi): Merge all alternatives into one, use
"sselog1" type instead of "sselog".  Remove explicit "memory"
attribute.
(vec_extract_hi_v16hi): Merge each pair of alternatives into one,
use "sselog1" type instead of "sselog".  Remove explicit "memory"
attribute.
(vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
"type", "prefix", "prefix_extra", "length_immediate" and "mode"
attributes.
(vec_extract_hi_v64qi): Merge all alternatives into one, use
"sselog1" type instead of "sselog".  Remove explicit "memory"
attribute.
(vec_extract_hi_v32qi): Merge each pair of alternatives into one,
use "sselog1" type instead of "sselog".  Remove explicit "memory"
attribute.

From-SVN: r259438

6 years agore PR testsuite/85326 (`make check` fails with `--disable-bootstrap` and `--enable...
Kyrylo Tkachov [Tue, 17 Apr 2018 17:06:36 +0000 (17:06 +0000)]
re PR testsuite/85326 (`make check` fails with `--disable-bootstrap` and `--enable-languages=c`)

PR testsuite/85326

Commit missing hunk from r259435.

From-SVN: r259437

6 years agore PR target/85430 (ICE: SIGSEGV in memory_operand at recog.c:1358/9 with -O2 -fno...
Jakub Jelinek [Tue, 17 Apr 2018 17:01:31 +0000 (19:01 +0200)]
re PR target/85430 (ICE: SIGSEGV in memory_operand at recog.c:1358/9 with -O2 -fno-tree-ccp -fno-tree-fre)

PR target/85430
* config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.

* gcc.dg/pr85430.c: New test.

From-SVN: r259436

6 years ago[AArch64/arm] PR testsuite/85326 Avoid C++ tests when C++ compiler not present
Kyrylo Tkachov [Tue, 17 Apr 2018 16:34:56 +0000 (16:34 +0000)]
[AArch64/arm] PR testsuite/85326 Avoid C++ tests when C++ compiler not present

PR testsuite/85326
* gcc.target/arm/pr54300.C: Move to...
* g++.dg/other/pr54300.C: ... Here.  Add target directives.
* gcc.target/arm/pr55073.C: Move to...
* g++.dg/other/pr55073.C: ... Here.  Add target directives.
* gcc.target/arm/pr56184.C: Move to...
* g++.dg/other/pr56184.C: ... Here.  Add target directives.
* gcc.target/arm/pr59985.C: Move to...
* g++.dg/other/pr59985.C: ... Here.  Add target directives.
* gcc.target/aarch64/pr60675.C: Move to...
* g++.dg/other/pr60675.C: ... Here.  Add target directives.
* gcc.target/aarch64/pr81422.C: Move to...
* g++.dg/other/pr81422.C: ... Here.  Add target directives.
* gcc.target/aarch64/sve/const_pred_1.C: Move to...
* g++.dg/other/sve_const_pred_1.C: ... Here.  Add target directives.
* gcc.target/aarch64/sve/const_pred_2.C: Move to...
* g++.dg/other/sve_const_pred_2.C: ... Here.  Add target directives.
* gcc.target/aarch64/sve/const_pred_3.C: Move to...
* g++.dg/other/sve_const_pred_3.C: ... Here.  Add target directives.
* gcc.target/aarch64/sve/const_pred_4.C: Move to...
* g++.dg/other/sve_const_pred_4.C: ... Here.  Add target directives.
* gcc.target/aarch64/sve/tls_2.C: Move to...
* g++.dg/other/sve_tls_2.C: ... Here.  Add target directives.
* gcc.target/aarch64/pr81414.C: Rename to...
* gcc.target/aarch64/pr81414.c: ... This.
* gcc.target/aarch64/simd/pr67896.C: Rename to...
* gcc.target/aarch64/simd/pr67896.c: ... This.  Update error expected
messages.
* gcc.target/aarch64/sve/vcond_1.C: Rename to...
* gcc.target/aarch64/sve/vcond_1.c: ... This.  Avoid use of stdint.h.
* gcc.target/aarch64/sve/vcond_1_run.C: Rename to...
* gcc.target/aarch64/sve/vcond_1_run.c: ... This.  Update include
file name.

From-SVN: r259435

6 years agobacktrace.c (backtrace_full): When testing whether we can allocate memory...
Ian Lance Taylor [Tue, 17 Apr 2018 13:59:38 +0000 (13:59 +0000)]
backtrace.c (backtrace_full): When testing whether we can allocate memory...

* backtrace.c (backtrace_full): When testing whether we can
allocate memory, call mmap directly, and munmap the memory.

Fixes https://github.com/ianlancetaylor/libbacktrace/issues/13 .

From-SVN: r259434

6 years agore PR middle-end/85414 (ICE: in ix86_expand_prologue, at config/i386/i386.c:13810...
Jakub Jelinek [Tue, 17 Apr 2018 09:16:48 +0000 (11:16 +0200)]
re PR middle-end/85414 (ICE: in ix86_expand_prologue, at config/i386/i386.c:13810 with -Og -fgcse)

PR middle-end/85414
* rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
on a SUBREG.

* gcc.dg/pr85414.c: New test.

From-SVN: r259433

6 years agoCall expand_all_artificial_thunks in ipa-cp if necessary
Martin Jambor [Tue, 17 Apr 2018 08:48:41 +0000 (10:48 +0200)]
Call expand_all_artificial_thunks in ipa-cp if necessary

2018-04-17  Martin Jambor  <mjambor@suse.cz>

PR ipa/85421
* ipa-cp.c (create_specialized_node): Call
expand_all_artificial_thunks if necessary.

testsuite/
* g++.dg/ipa/pr85421.C: New test.

From-SVN: r259432

6 years agoFix coding style and add a new test-case (PR lto/85405).
Martin Liska [Tue, 17 Apr 2018 08:28:21 +0000 (10:28 +0200)]
Fix coding style and add a new test-case (PR lto/85405).

2018-04-17  Martin Liska  <mliska@suse.cz>

PR lto/85405
* ipa-devirt.c (odr_types_equivalent_p): Remove trailing
in message, remote space in between '_G' and '('.
2018-04-17  Martin Liska  <mliska@suse.cz>

PR lto/85405
* g++.dg/lto/pr85405b_0.C: New test.
* g++.dg/lto/pr85405b_1.C: New test.

From-SVN: r259431

6 years agore PR target/85281 (Assembler messages: Error: operand size mismatch for `vpbroadcast...
Jakub Jelinek [Tue, 17 Apr 2018 07:08:06 +0000 (09:08 +0200)]
re PR target/85281 (Assembler messages: Error: operand size mismatch for `vpbroadcastb' with -mavx512bw -masm=intel)

PR target/85281
* config/i386/sse.md (reduces<mode><mask_scalar_name>,
avx512f_vmcmp<mode>3<round_saeonly_name>,
avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
avx512f_rndscale<mode><round_saeonly_name>,
avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
Use %<iptr>2 instead of %2 for -masm=intel.
(avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
avx512f_vcvttss2usi<round_saeonly_name>,
avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
-masm=intel.
(avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
avx512f_vcvttsd2usi<round_saeonly_name>,
avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
Use %q1 instead of %1 for -masm=intel.
(avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
of %3 for -masm=intel.
(sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
-masm=intel.
(*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
-masm=intel.
(*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
-masm=intel.
(avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
%k0 and %1 for -masm=intel rather than two patterns, one with %0 and
%g1.
(*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
-masm=intel.
(avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
%q0 and %1 for -masm=intel rather than two patterns, one with %0 and
%g1 and one with %0 and %1.
(avx512er_vmrcp28<mode><round_saeonly_name>,
avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
%1 for -masm=intel.
(avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
of %0 and %{%4%} for -masm=intel.
(avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
order of %0 and %{%5%}%{z%} for -masm=intel.

From-SVN: r259430

6 years agoSupport bitfields in Wodr machinery (PR lto/85405).
Martin Liska [Tue, 17 Apr 2018 05:41:40 +0000 (05:41 +0000)]
Support bitfields in Wodr machinery (PR lto/85405).

2018-04-17  Jan Hubicka  <jh@suse.cz>

PR lto/85405
* ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2018-04-17  Martin Liska  <mliska@suse.cz>

PR lto/85405
* g++.dg/lto/pr85405_0.C: New test.
* g++.dg/lto/pr85405_1.C: New test.

From-SVN: r259429

6 years agoMake redirection only for target_clones: V3 (PR ipa/85329).
Martin Liska [Tue, 17 Apr 2018 05:40:39 +0000 (07:40 +0200)]
Make redirection only for target_clones: V3 (PR ipa/85329).

2018-04-17  Martin Liska  <mliska@suse.cz>

PR ipa/85329
* multiple_target.c (create_dispatcher_calls): Set apostrophes
for target_clone error message.  Make default implementation
        clone to be a local declaration.
(separate_attrs): Add new argument and check for an empty
string.
(expand_target_clones): Handle it.
(ipa_target_clone): Make redirection just for target_clones
functions.
2018-04-17  Martin Liska  <mliska@suse.cz>

PR ipa/85329
* g++.dg/ext/pr85329-2.C: New test.
* g++.dg/ext/pr85329.C: New test.
* gcc.target/i386/mvc12.c: New test.

From-SVN: r259428

6 years agoDaily bump.
GCC Administrator [Tue, 17 Apr 2018 00:16:14 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r259426

6 years ago[PR c++/85039] no type definitions in builtin offsetof
Alexandre Oliva [Mon, 16 Apr 2018 21:35:34 +0000 (21:35 +0000)]
[PR c++/85039] no type definitions in builtin offsetof

Types defined within a __builtin_offsetof argument don't always get
properly recorded as members of their context types, so if they're
anonymous, we may fail to assign them an anon type index for mangling
and ICE.

We shouldn't allow types to be introduced in __builtin_offsetof, I
think, and Jason says the std committee agrees, so I've arranged for
us to reject them.

Even then, we still parse the definitions and attempt to assign
mangled names to its member functions, so the ICE remains.  Since
we've already reported an error, we might as well complete the name
assignment with an arbitrary index, thus avoiding the ICE.

We used to have a test that expected to be able to define types in
__builtin_offsetof; this patch removes that specific test.

for  gcc/cp/ChangeLog

PR c++/85039
* parser.c (cp_parser_builtin_offset): Reject type definitions.
* mangle.c (nested_anon_class_index): Avoid crash returning -1
if we've seen errors.

for  gcc/testsuite/ChangeLog

PR c++/85039
* g++.dg/pr85039-1.C: New.
* g++.dg/pr85039-2.C: New.
* g++.dg/parse/semicolon3.C: Remove test_offsetof.

From-SVN: r259423