tree-optimization/119960 - fix and guard get_later_stmt

The following makes get_later_stmt handle stmts from different
basic-blocks in the case they are orderd and otherwise asserts.

* tree-vectorizer.h (get_later_stmt): Robustify against
stmts in different BBs, assert when they are unordered.

tree-optimization/119960 - failed external SLP promotion

The following addresses a too conservative sanity check of SLP nodes
we want to promote external.  The issue lies in code generation
for such external which relies on get_later_stmt to figure an
insert location.  But get_later_stmt relies on the ability to
totally order stmts, specifically implementation-wise that they
are all from the same BB, which is what is verified at the moment.

The patch changes this to require stmts to be orderable by
dominance queries.  For simplicity and seemingly enough for the
testcase in PR119960, this handles the case of two distinct BBs.

PR tree-optimization/119960
* tree-vect-slp.cc (vect_slp_can_convert_to_external):
Handle cases where defs from multiple BBs are ordered
by their dominance relation.

* gcc.dg/vect/bb-slp-pr119960-1.c: New testcase.

ipa/120006 - wrong code with IPA PTA

When PTA gets support for special-handling more builtins in
find_func_aliases the corresponding code in find_func_clobbers
needs updating as well since for unhandled cases it assumes
the former will populate ESCAPED accordingly.  The following
fixes a few omissions, the testcase runs into the missing strdup
handling.  I believe the more advanced handling using modref
results and fnspecs opened a larger gap, the proper fix is to
merge both functions, gating the clobber/use part on a parameter
to avoid diverging.

PR ipa/120006
* tree-ssa-structalias.cc (find_func_clobbers): Handle
strdup, strndup, realloc, index, strchr, strrchr, memchr,
strstr, strpbrk builtins like find_func_aliases does.

* gcc.dg/torture/pr120006.c: New testcase.

tree-optimization/120003 - missed jump threading

The following allows the entry and exit block of a jump thread path
to be equal, which can easily happen when there isn't a forwarder
on the interesting edge for an FSM thread conditional.  We just
don't want to enlarge the path from such a block.

PR tree-optimization/120003
* tree-ssa-threadbackward.cc (back_threader::find_paths_to_names):
Allow block re-use but do not enlarge the path beyond such a
re-use.

* gcc.dg/tree-ssa/ssa-thread-23.c: New testcase.
* gcc.dg/tree-ssa/ssa-dom-thread-7.c: Adjust.

Fix GNAT build failure for x86/FreeBSD

gcc/ada/
PR ada/112958
* Makefile.rtl (LIBGNAT_TARGET_PAIRS) [x86 FreeBSD]: Add specific
version of s-dorepr.adb.
* libgnat/s-dorepr__freebsd.adb: New file.

AVR: target/119989 - Add missing clobbers to xload_<mode>_libgcc.

libgcc's __xload_1...4 is clobbering Z (and also R21 is some cases),
but avr.md had clobbers of respective GPRs only up to reload.
Outcome was that code reading from the same __memx address twice
could be wrong.  This patch adds respective clobbers.

      Forward-port from 2025-04-30 r14-11703
      PR target/119989
gcc/
* config/avr/avr.md (xload_<mode>_libgcc): Clobber R21, Z.

gcc/testsuite/
* gcc.target/avr/torture/pr119989.h: New file.
* gcc.target/avr/torture/pr119989-memx-1.c: New test.
* gcc.target/avr/torture/pr119989-memx-2.c: New test.
* gcc.target/avr/torture/pr119989-memx-3.c: New test.
* gcc.target/avr/torture/pr119989-memx-4.c: New test.
* gcc.target/avr/torture/pr119989-flashx-1.c: New test.
* gcc.target/avr/torture/pr119989-flashx-2.c: New test.
* gcc.target/avr/torture/pr119989-flashx-3.c: New test.
* gcc.target/avr/torture/pr119989-flashx-4.c: New test.

(cherry picked from commit 1ca1c1fc3b58ae5e1d3db4f5a2014132fe69f82a)

RISC-V: Allow different dynamic floating point mode to be merged [PR119832]

Although we already try to set the mode needed to FRM_DYN after a function call,
there are still some corner cases where both FRM_DYN and FRM_DYN_CALL may appear
on incoming edges.

Therefore, we use TARGET_MODE_CONFLUENCE to tell GCC that FRM_DYN, FRM_DYN_CALL,
and FRM_DYN_EXIT modes are compatible.

gcc/ChangeLog:

PR target/119832
* config/riscv/riscv.cc (riscv_dynamic_frm_mode_p): New.
(riscv_mode_confluence): New.
(TARGET_MODE_CONFLUENCE): Define to riscv_mode_confluence.

gcc/testsuite/ChangeLog:

PR target/119832
* g++.target/riscv/pr119832.C: New test.

RISC-V: Fix missing implied Zicsr from Zve32x

The Zve32x extension depends on the Zicsr extension.
Currently, enabling Zve32x alone does not automatically imply Zicsr in GCC.

gcc/ChangeLog:

* common/config/riscv/riscv-common.cc: Add Zve32x depends on Zicsr

gcc/testsuite/ChangeLog:

* gcc.target/riscv/predef-19.c: set the march to rv64im_zve32x
instead of rv64gc_zve32x to avoid Zicsr implied by g. Extra m is
added to avoid current 'V' extension requires 'M' extension

Signed-off-by: Jerry Zhang Jian <jerry.zhangjian@sifive.com>

AArch64: Fold LD1/ST1 with ptrue to LDR/STR for 128-bit VLS

If -msve-vector-bits=128, SVE loads and stores (LD1 and ST1) with a
ptrue predicate can be replaced by neon instructions (LDR and STR),
thus avoiding the predicate altogether. This also enables formation of
LDP/STP pairs.

For example, the test cases

svfloat64_t
ptrue_load (float64_t *x)
{
  svbool_t pg = svptrue_b64 ();
  return svld1_f64 (pg, x);
}
void
ptrue_store (float64_t *x, svfloat64_t data)
{
  svbool_t pg = svptrue_b64 ();
  return svst1_f64 (pg, x, data);
}

were previously compiled to
(with -O2 -march=armv8.2-a+sve -msve-vector-bits=128):

ptrue_load:
        ptrue   p3.b, vl16
        ld1d    z0.d, p3/z, [x0]
        ret
ptrue_store:
        ptrue   p3.b, vl16
        st1d    z0.d, p3, [x0]
        ret

Now the are compiled to:

ptrue_load:
        ldr     q0, [x0]
        ret
ptrue_store:
        str     q0, [x0]
        ret

The implementation includes the if-statement
if (known_eq (GET_MODE_SIZE (mode), 16)
    && aarch64_classify_vector_mode (mode) == VEC_SVE_DATA)
which checks for 128-bit VLS and excludes partial modes with a
mode size < 128 (e.g. VNx2QI).

The patch was bootstrapped and tested on aarch64-linux-gnu, no regression.
OK for mainline?

Signed-off-by: Jennifer Schmitz <jschmitz@nvidia.com>
gcc/
* config/aarch64/aarch64.cc (aarch64_emit_sve_pred_move):
Fold LD1/ST1 with ptrue to LDR/STR for 128-bit VLS.

gcc/testsuite/
* gcc.target/aarch64/sve/ldst_ptrue_128_to_neon.c: New test.
* gcc.target/aarch64/sve/cond_arith_6.c: Adjust expected outcome.
* gcc.target/aarch64/sve/pcs/return_4_128.c: Likewise.
* gcc.target/aarch64/sve/pcs/return_5_128.c: Likewise.
* gcc.target/aarch64/sve/pcs/struct_3_128.c: Likewise.

RISC-V: Add intrinsics testcases for SiFive Xsfvcp extensions.

This commit adds testcases for Xsfvcp.

Co-Authored by: Jiawei Chen <jiawei@iscas.ac.cn>
Co-Authored by: Shihua Liao <shihua@iscas.ac.cn>
Co-Authored by: Yixuan Chen <chenyixuan@iscas.ac.cn>

gcc/testsuite/ChangeLog:

* gcc.target/riscv/rvv/xsfvector/sf_vc_f.c: New test.
* gcc.target/riscv/rvv/xsfvector/sf_vc_i.c: New test.
* gcc.target/riscv/rvv/xsfvector/sf_vc_v.c: New test.
* gcc.target/riscv/rvv/xsfvector/sf_vc_x.c: New test.

RISC-V: Add intrinsics support for SiFive Xsfvcp extensions.

This version is same as v5, but rebase to trunk, send out to trigger CI.

This commit adds intrinsics support for Xsfvcp extension.
Diff with V4: Delete the sifive_vector.h file.

Co-Authored by: Jiawei Chen <jiawei@iscas.ac.cn>
Co-Authored by: Shihua Liao <shihua@iscas.ac.cn>
Co-Authored by: Yixuan Chen <chenyixuan@iscas.ac.cn>

gcc/ChangeLog:

* config/riscv/constraints.md (Ou01): New constraint.
(Ou02): Ditto.
* config/riscv/generic-vector-ooo.md (vec_sf_vcp): New reservation.
* config/riscv/genrvv-type-indexer.cc (main): New type.
* config/riscv/riscv-c.cc (riscv_pragma_intrinsic): Add xsfvcp strings.
* config/riscv/riscv-vector-builtins-shapes.cc (struct sf_vcix_se_def):
New function.
(struct sf_vcix_def): Ditto.
(SHAPE): Ditto.
* config/riscv/riscv-vector-builtins-shapes.h: Ditto.
* config/riscv/riscv-vector-builtins-types.def (DEF_RVV_X2_U_OPS): New type.
(DEF_RVV_X2_WU_OPS): Ditto.
(vuint8mf8_t): Ditto.
(vuint8mf4_t): Ditto.
(vuint8mf2_t): Ditto.
(vuint8m1_t): Ditto.
(vuint8m2_t): Ditto.
(vuint8m4_t): Ditto.
(vuint16mf4_t): Ditto.
(vuint16mf2_t): Ditto.
(vuint16m1_t): Ditto.
(vuint16m2_t): Ditto.
(vuint16m4_t): Ditto.
(vuint32mf2_t): Ditto.
(vuint32m1_t): Ditto.
(vuint32m2_t): Ditto.
(vuint32m4_t): Ditto.
* config/riscv/riscv-vector-builtins.cc (DEF_RVV_X2_U_OPS): New builtins
def.
(DEF_RVV_X2_WU_OPS): Ditto.
(rvv_arg_type_info::get_scalar_float_type): Ditto.
(function_instance::modifies_global_state_p): Ditto.
* config/riscv/riscv-vector-builtins.def (v_x): New base type.
(i): Ditto.
(v_i): Ditto.
(xv): Ditto.
(iv): Ditto.
(fv): Ditto.
(vvv): Ditto.
(xvv): Ditto.
(ivv): Ditto.
(fvv): Ditto.
(vvw): Ditto.
(xvw): Ditto.
(ivw): Ditto.
(fvw): Ditto.
(v_vv): Ditto.
(v_xv): Ditto.
(v_iv): Ditto.
(v_fv): Ditto.
(v_vvv): Ditto.
(v_xvv): Ditto.
(v_ivv): Ditto.
(v_fvv): Ditto.
(v_vvw): Ditto.
(v_xvw): Ditto.
(v_ivw): Ditto.
(v_fvw): Ditto.
(x2_vector): Ditto.
(scalar_float): Ditto.
* config/riscv/riscv-vector-builtins.h (enum required_ext): New extension.
(required_ext_to_isa_name): Ditto.
(required_extensions_specified): Ditto.
(struct rvv_arg_type_info): Ditto.
(struct function_group_info): Ditto.
* config/riscv/riscv.md: New attr.
* config/riscv/sifive-vector-builtins-bases.cc (class sf_vc): New function.
(BASE): New base_name.
* config/riscv/sifive-vector-builtins-bases.h: New function_base.
* config/riscv/sifive-vector-builtins-functions.def
(REQUIRED_EXTENSIONS): New intrinsics def.
(sf_vc): Ditto.
* config/riscv/sifive-vector.md (@sf_vc_x_se<mode>): New RTL mode.
(@sf_vc_v_x_se<mode>): Ditto.
(@sf_vc_v_x<mode>): Ditto.
(@sf_vc_i_se<mode>): Ditto.
(@sf_vc_v_i_se<mode>): Ditto.
(@sf_vc_v_i<mode>): Ditto.
(@sf_vc_vv_se<mode>): Ditto.
(@sf_vc_v_vv_se<mode>): Ditto.
(@sf_vc_v_vv<mode>): Ditto.
(@sf_vc_xv_se<mode>): Ditto.
(@sf_vc_v_xv_se<mode>): Ditto.
(@sf_vc_v_xv<mode>): Ditto.
(@sf_vc_iv_se<mode>): Ditto.
(@sf_vc_v_iv_se<mode>): Ditto.
(@sf_vc_v_iv<mode>): Ditto.
(@sf_vc_fv_se<mode>): Ditto.
(@sf_vc_v_fv_se<mode>): Ditto.
(@sf_vc_v_fv<mode>): Ditto.
(@sf_vc_vvv_se<mode>): Ditto.
(@sf_vc_v_vvv_se<mode>): Ditto.
(@sf_vc_v_vvv<mode>): Ditto.
(@sf_vc_xvv_se<mode>): Ditto.
(@sf_vc_v_xvv_se<mode>): Ditto.
(@sf_vc_v_xvv<mode>): Ditto.
(@sf_vc_ivv_se<mode>): Ditto.
(@sf_vc_v_ivv_se<mode>): Ditto.
(@sf_vc_v_ivv<mode>): Ditto.
(@sf_vc_fvv_se<mode>): Ditto.
(@sf_vc_v_fvv_se<mode>): Ditto.
(@sf_vc_v_fvv<mode>): Ditto.
(@sf_vc_vvw_se<mode>): Ditto.
(@sf_vc_v_vvw_se<mode>): Ditto.
(@sf_vc_v_vvw<mode>): Ditto.
(@sf_vc_xvw_se<mode>): Ditto.
(@sf_vc_v_xvw_se<mode>): Ditto.
(@sf_vc_v_xvw<mode>): Ditto.
(@sf_vc_ivw_se<mode>): Ditto.
(@sf_vc_v_ivw_se<mode>): Ditto.
(@sf_vc_v_ivw<mode>): Ditto.
(@sf_vc_fvw_se<mode>): Ditto.
(@sf_vc_v_fvw_se<mode>): Ditto.
(@sf_vc_v_fvw<mode>): Ditto.
* config/riscv/vector-iterators.md: New iterator.
* config/riscv/vector.md: New vtype.

Fix compilation failure on FreeBSD

[changelog]
PR ada/112958
* init.c (__gnat_error_handler) [__FreeBSD__]: Fix typo.

libstdc++: [_GLIBCXX_INLINE_VERSION] Fix tests failures

Adapt testsuite v3_target_compile to strip version namespace from compiler
output so that dg-error and dg-warning directives do not need to consider it.

Avoid a aligned_storage check as behavior has been fixed only when using
gnu-versioned-namespace as it is an abi breaking change.

libstdc++-v3/ChangeLog:

* testsuite/lib/libstdc++.exp (v3_target_compile): Strip version namespace
from compiler output.
* testsuite/20_util/aligned_storage/value.cc [_GLIBCXX_INLINE_VERSION]:
Avoid align_msa check.
* testsuite/20_util/function/cons/70692.cc: Remove now useless __8 namespace
pattern.
* testsuite/23_containers/map/48101_neg.cc: Likewise.
* testsuite/23_containers/multimap/48101_neg.cc: Likewise.

Co-authored-by: Jonathan Wakely <jwakely@redhat.com>

libstdc++: Fix _Padding_sink in case when predicted width is between padwidth and maxwidth [PR109162]

The _Padding_sink was behaving incorrectly, when the predicted width (based on
code units count) was higher than _M_maxwidth, but lower than _M_padwidth.
In this case _M_update() returned without calling _M_force_update() and computing
field width for Unicode encoding, because _M_buffering() returned 'true'.
As a consequence we switched to _M_ignoring() mode, while storing a sequence
with more code units but smaller field width than _M_maxwidth.

We now call _M_force_update() if predicted width is greater or equal to either
_M_padwidth or _M_maxwidth.

This happened for existing test case on 32bit architecture.

PR libstdc++/109162

libstdc++-v3/ChangeLog:

* include/std/format (_Padding_sink::_M_update): Fixed condition for
calling _M_force_update.
* testsuite/std/format/debug.cc: Add test that reproduces this issue
on 64bit architecture.
* testsuite/std/format/ranges/sequence.cc: Another edge value test.

Daily bump.

Fix cs_interesting_for_ipcp_p wrt flag_profile_partial_training.

As noticed by Martin Jambor, I introduced a bug while simplifying
cs_interesting_for_ipcp_p and reversed condition for
flag_profile_partial_training.  Also I noticed that we probably want to
consider calls with unintialized counts for cloning so the pass does somehting
with -fno-guess-branch-probability even thugh this is probably not very useful
in practice.

gcc/ChangeLog:

* ipa-cp.cc (cs_interesting_for_ipcp_p): Fix handling of uninitialized
counts and 0 IPA cost wrt flag_profile_partial_training.

analyzer: fix for older version of GCC

Having both an enum and a variable with the same name triggers an error with
gcc 5.

  exploded-graph.h:351:29: error: ‘status’ is not a class, namespace, or enumeration

gcc/analyzer/ChangeLog:
* exploded-graph.h (set_status): Rename parameter.
* constraint-manager.cc (bound::ensure_closed): Likewise.
(range::add_bound): Likewise.

Signed-off-by: Marc Poulhiès <dkm@kataplop.net>
Reviewed-by: David Malcolm <dmalcolm@redhat.com>
Signed-off-by: Marc Poulhiès <dkm@kataplop.net>

d: Use __builtin_clear_padding for zeroing alignment holes after set

In an earlier change, a wrapper function was added to set
CONSTRUCTOR_ZERO_PADDING_BITS on all CONSTRUCTOR nodes. This removes all
the old generated calls to built-in memset and memcpy as zero padding is
now taken care of by the middle-end.

The remaining constructors that weren't getting zero padded was
ARRAY_TYPEs, so now `__builtin_clear_padding' is used to fill in all
alignment holes in constructed array literals where required.

PR d/103044

gcc/d/ChangeLog:

* d-tree.h (build_clear_padding_call): New prototype.
* d-codegen.cc (build_clear_padding_call): New function.
(build_memset_call): Remove generated call to __builtin_memcpy.
(build_address): Replace generated call to __builtin_memset with
__builtin_clear_padding.
(build_array_from_exprs): Likewise.
* expr.cc (ExprVisitor::visit (AssignExp *)): Remove generated call to
__builtin_memset.
(ExprVisitor::visit (ArrayLiteralExp *)): Likewise.  Insert call to
__builtin_clear_padding after copying array into GC memory.
(ExprVisitor::visit (StructLiteralExp *)): Remove generated call to
__builtin_memset.
* toir.cc (IRVisitor::visit (ReturnStatement *)): Likewise.

libstdc++: Use no_stdname for make_obj_using_allocator feature test macro

This is a non-standard feature test macro only used internally, so use
the new no_stdname property for it.

libstdc++-v3/ChangeLog:

* include/bits/version.def (make_obj_using_allocator): Use
no_stdname.
* include/bits/version.h: Regenerate.

libstdc++: Use constexpr-if to slightly simplify <regex>

This will hardly make a dent in the very slow compile times for <regex>
but it seems worth doing anyway.

libstdc++-v3/ChangeLog:

* include/bits/regex_compiler.h: Replace _GLIBCXX17_CONSTEXPR
with constexpr and disable diagnostics with pragmas.
(_AnyMatcher::operator()): Use constexpr-if instead of tag
dispatching. Postpone calls to _M_translate until after checking
result of earlier calls.
(_AnyMatcher::_M_apply): Remove both overloads.
(_BracketMatcher::operator(), _BracketMatcher::_M_ready):
Replace tag dispatching with 'if constexpr'.
(_BracketMatcher::_M_apply(_CharT, true_type)): Remove.
(_BracketMatcher::_M_apply(_CharT, false_type)): Remove second
parameter.
(_BracketMatcher::_M_make_cache): Remove both overloads.
* include/bits/regex_compiler.tcc (_BracketMatcher::_M_apply):
Remove second parameter.
* include/bits/regex_executor.tcc: Replace _GLIBCXX17_CONSTEXPR
with constexpr and disable diagnostics with pragmas.
(_Executor::_M_handle_backref): Replace __glibcxx_assert with
static_assert.
(_Executor::_M_handle_accept): Mark _S_opcode_backref case as
unreachable for non-DFS mode and do not instantiate
_M_handle_backref for that case.

Reviewed-by: Tomasz Kamiński <tkaminsk@redhat.com>

simplify-rtx: Combine bitwise operations in more cases

This patch transforms RTL expressions of the form (subreg (not X)) into
(not (subreg X)) if the subreg is an operand of another binary logical
operation. This transformation can expose opportunities to combine more
logical operations.

For example, it improves the codegen of the following AArch64 NEON
intrinsics:
vandq_s64(vreinterpretq_s64_s32(vmvnq_s32(a)),
  vreinterpretq_s64_s32(b));
from:
not     v0.16b, v0.16b
and     v0.16b, v0.16b, v1.16b
to:
bic     v0.16b, v1.16b, v0.16b

Regression tested on x86_64-linux-gnu, arm-linux-gnueabihf and
aarch64-linux-gnu.

gcc/ChangeLog:

* simplify-rtx.cc (non_paradoxical_subreg_not_p): New function
for pattern match of (subreg (not X)).
(simplify_with_subreg_not): New function for simplification.

i386: Disable string insn from non-default AS for Pmode != word_mode [PR111657]

0x67 prefix is applied before segment register. That is in

rep movsq %gs:(%esi), (%edi)

the address is %gs + %esi. In case Pmode != word_mode (x32 with a default
-maddress-mode=short) instructions should not allow segment override prefixes.

Also, remove explicit addr32 prefix from asm templates because address
mode can be determined from explicit instruction operands. Also note that
Pmode != word_mode only with TARGET_64BIT, so the check in ix86_print_operand
is not needed.

PR target/111657

gcc/ChangeLog:

* config/i386/i386-expand.cc (alg_usable_p): For Pmode != word_mode
reject rep_prefix_{1,4,8}_byte algorithms with src_as in the
non-default address space.
* config/i386/i386-protos.h (ix86_check_movs): New prototype.
* config/i386/i386.cc (ix86_check_movs): New function.
(ix86_print_operand) [case '^']: Remove excess check for TARGET_64BIT.
* config/i386/i386.md (strmov): For Pmode != word_mode expand with
gen_strmov_single only when operands[3] (source) is in the default
address space.
(*strmovdi_rex_1) Use ix86_check_movs.  Remove %^ from asm template.
(*strmovsi_1): Ditto.
(*strmovhi_1): DItto.
(*strmovqi_1): Ditto.
(*rep_movdi_rex64): Ditto.
(*rep_movsi): Ditto.
(*rep_movqi): Ditto.

gcc/testsuite/ChangeLog:

* gcc.target/i386/pr111657-1.c: Check that segment override is not
generated for "rep movsq" for x32 target.

libstdc++: Optimize removal from unique assoc containers [PR112934]

Previously, calling erase(key) on both std::map and std::set
would execute that same code that std::multi{map,set} would.
However, doing that is unnecessary because std::{map,set}
guarantee that all elements are unique.

It is reasonable to expect that erase(key) is equivalent
or better than:

  auto it = m.find(key);
  if (it != m.end())
    m.erase(it);

However, this was not the case. Fix that by adding a new
function _Rb_tree<>::_M_erase_unique() that is essentially
equivalent to the above snippet, and use this from both
std::map and std::set.

libstdc++-v3/ChangeLog:

PR libstdc++/112934
* include/bits/stl_map.h (map::erase): Use _M_erase_unique.
* include/bits/stl_set.h (set::erase): Likewise.
* include/bits/stl_tree.h (_Rb_tree::_M_erase_unique): Add.

RISC-V: Fix register move cost for SIBCALL_REGS/JALR_REGS

SIBCALL_REGS/JALR_REGS are also subset of GR_REGS and need to
be taken into acount in riscv_register_move_cost, otherwise it
will get a incorrect cost.

gcc/ChangeLog:

* config/riscv/riscv.cc (riscv_register_move_cost): Use
reg_class_subset_p to check the reg class.

libstdc++: Fix availability of std::erase_if(std::flat_foo) [PR119427]

These std::erase_if overloads were wrongly implemented as hidden
friends, visible only via ADL, so erase_if(x) would work but not
std::erase_if(x).

PR libstdc++/119427

libstdc++-v3/ChangeLog:

* include/std/flat_map (_Flat_map_impl::erase_if): Replace
this hidden friend with ...
(_Flat_map_impl::_M_erase_if): ... this member function.
(flat_map): Export _Flat_map_impl::_M_erase_if.
(erase_if(flat_map)): Define.
(flat_multimap): Export _Flat_map_impl::_M_erase_if.
(erase_if(flat_multimap)): Define.
* include/std/flat_set (_Flat_set_impl::erase_if): Replace
with ...
(_Flat_set_impl::_M_erase_if): ... this member function.
(flat_set): Export _Flat_set_impl::_M_erase_if.
(erase_if(flat_set)): Define.
(flat_multiset): Export _Flat_set_impl::_M_erase_if.
(erase_if(flat_multiset)): Define.
* testsuite/23_containers/flat_map/1.cc (test07): New test.
* testsuite/23_containers/flat_multimap/1.cc (test07): New test.
* testsuite/23_containers/flat_multiset/1.cc (test09): New test.
* testsuite/23_containers/flat_set/1.cc (test09): New test.

Reviewed-by: Jonathan Wakely <jwakely@redhat.com>

tree-optimization/119997 - &ptr->field no longer subject to PRE

The following makes PRE handle &ptr->field the same as VN by
treating it as a POINTER_PLUS_EXPR when possible and thus as
'nary'.  To facilitate this the patch splits out vn_pp_nary_for_addr
and adds const overloads for vec::last.  The patch also avoids
handling an effective zero offset as POINTER_PLUS_EXPR.

PR tree-optimization/119997
* vec.h (vec<T, A, vl_embed>::last): Provide const overload.
(vec<T, va_heap, vl_ptr>::last): Likewise.
* tree-ssa-sccvn.h (vn_pp_nary_for_addr): Declare.
* tree-ssa-sccvn.cc (vn_pp_nary_for_addr): Split out from ...
(vn_reference_lookup): ... here.
(vn_reference_insert): ... and duplicate here.  Do not handle
zero offset as POINTER_PLUS_EXPR.
* tree-ssa-pre.cc (compute_avail): Implement
ADDR_EXPR-as-POINTER_PLUS_EXPR special casing.

* gcc.dg/tree-ssa/ssa-pre-35.c: New testcase.

libstdc++: Use constexpr-if for C++11 and C++14

Replace remaining uses of _GLIBCXX17_CONSTEXPR for constexpr-if, so that
we always use constexpr-if in C++11 and C++14. Diagnostic pragmas are
used to suppress diagnostics.

libstdc++-v3/ChangeLog:

* include/bits/char_traits.h (char_traits::assign): Use
constexpr-if unconditionally for C++11 and C++14.
* include/bits/locale_conv.h (__do_str_codecvt): Likewise.
* include/bits/ostream.h (basic_ostream::_S_cast_flt): Likewise.
* include/bits/random.tcc (shuffle_order_engine::operator())
(seed_seq::seed_seq(Iter, Iter)): Likewise.
* include/bits/shared_ptr_base.h (_Sp_counted_base::_M_release):
Likewise.
* include/bits/stl_tree.h (_Rb_tree::_M_move_data): Likewise.
* include/bits/uniform_int_dist.h
(uniform_int_distribution::operator()): Likewise.
* include/bits/valarray_array.h (__valarray_default_construct)
(__valarray_fill_construct, __valarray_copy_construct)
(__valarray_copy_construct, __valarray_destroy_elements):
Likewise.
* include/experimental/numeric (lcm): Likewise.
* include/std/bit (__rotl, __rotr, __countl_zero, __countr_zero)
(__popcount, __bit_ceil) Likewise.:
* include/std/bitset (operator>>): Likewise.
* include/std/charconv (__to_chars_8, __to_chars_i)
(__from_chars_alnum_to_val, from_chars): Likewise.
* include/tr2/dynamic_bitset (__dynamic_bitset_base): Likewise.
* testsuite/26_numerics/random/pr60037-neg.cc: Adjust dg-error
line number.

Reviewed-by: Tomasz Kamiński <tkaminsk@redhat.com>

libstdc++: Use constexpr-if in std::function for C++11 and C++14

This allows removing the _Target_handler class template, because it's no
longer needed to prevent instantiating invalid specializations of
_Function_handler.

libstdc++-v3/ChangeLog:

* include/bits/std_function.h (_Target_handler): Remove.
(function::target): Use constexpr-if for C++11 and
C++14, with diagnostic pragmas to suppress warnings.

Reviewed-by: Tomasz Kamiński <tkaminsk@redhat.com>

libstdc++: Use constexpr-if to simplify std::vector relocation

Simplify std::vector's use of std::__relocate_a by using 'if constexpr'
even in C++11 and C++14, with diagnostic pragmas to disable warnings.
This allows us to call std::__relocate_a directly, instead of via
_S_relocate and tag distpatching.

Preserve _S_relocate so that explicit instantiations still get it, but
make it a no-op when _S_use_relocate() is false, so that we don't
instantiate __relocate_a if it isn't needed.

libstdc++-v3/ChangeLog:

* include/bits/stl_vector.h (_S_do_relocate): Remove.
(_S_relocate): Remove tag dispatching path.
* include/bits/vector.tcc (reserve, _M_realloc_insert)
(_M_realloc_append, _M_default_append): Add diagnostic pragmas
and use 'if constexpr' in C++11 and C++14. Call
std::__relocate_a directly instead of _S_relocate.

Reviewed-by: Tomasz Kamiński <tkaminsk@redhat.com>

libstdc++: Fix allocator propagation for rvalue+rvalue string concatenation

I made a last-minute change to Nina's r10-200-gf4e678ef74b272
implementation of P1165R1 (consistent allocator propagation for
operator+ on strings), so that the rvalue+rvalue case assumes that COW
strings do not support stateful allocators. I don't think that was true
when the change went in, and isn't true now. COW strings don't support
allocator propagation on assignment and swap, but they do support
non-equal stateful allocators, which are correctly propagated on move
construction.

This removes the preprocessor conditional in the rvalue+rvalue overload
so that COW strings are handled equivalently. Also use constexpr-if
unconditionally, disabling diagnostics with pragmas.

libstdc++-v3/ChangeLog:

* include/bits/basic_string.h (operator+(string&&, string&&)):
Do not assume that COW strings have equal allocators. Use
constexpr-if unconditionally.
* testsuite/21_strings/basic_string/allocator/char/operator_plus.cc:
Remove cxx11_abi effective-target check.
* testsuite/21_strings/basic_string/allocator/wchar_t/operator_plus.cc:
Likewise.

Reviewed-by: Tomasz Kamiński <tkaminsk@redhat.com>

i386: Allow string instructions from non-default address space [PR111657]

MOVS instructions allow segment override of their source operand, e.g.:

rep movsq %gs:(%rsi), (%rdi)

where %rsi is the address of the source location (with %gs segment override)
and %rdi is the address of the destination location.

The testcase improves from (-O2 -mno-sse -mtune=generic):

xorl %eax, %eax
.L2:
movl %eax, %edx
addl $8, %eax
movq %gs:m(%rdx), %rcx
movq %rcx, (%rdi,%rdx)
cmpl $240, %eax
jb .L2
ret

to:
movl $m, %esi
movl $30, %ecx
rep movsq %gs:(%rsi), (%rdi)
ret

PR target/111657

gcc/ChangeLog:

* config/i386/i386-expand.cc (alg_usable_p): Remove have_as bool
argument and add dst_as and src_as address space arguments.  Reject
libcall algorithm with dst_as and src_as in the non-default address
spaces.  Reject rep_prefix_{1,4,8}_byte algorithms with dst_as in
the non-default address space.
(decide_alg): Remove have_as bool argument and add dst_as and src_as
address space arguments.  Update calls to alg_usable_p.
(ix86_expand_set_or_cpymem): Update call to decide_alg.
* config/i386/i386.md (strmov): Do not fail if operand[3] (source)
is in the non-default address space.  Expand with gen_strmov_singleop
only when operand[1] (destination) is in the default address space.
(*strmovdi_rex_1): Determine memory operands from insn pattern.
Allow only when destination is in the default address space.
Rewrite asm template to use explicit operands.
(*strmovsi_1): Ditto.
(*strmovhi_1): DItto.
(*strmovqi_1): Ditto.
(*rep_movdi_rex64): Ditto.
(*rep_movsi): Ditto.
(*rep_movqi): Ditto.
(*strsetdi_rex_1): Determine memory operands from insn pattern.
Allow only when destination is in the default address space.
(*strsetsi_1): Ditto.
(*strsethi_1): Ditto.
(*strsetqi_1): Ditto.
(*rep_stosdi_rex64): Ditto.
(*rep_stossi): Ditto.
(*rep_stosqi): Ditto.

gcc/testsuite/ChangeLog:

* gcc.target/i386/pr111657-1.c: New test.

target.def: Remove TARGET_PROMOTE_FUNCTION_RETURN reference

Since TARGET_PROMOTE_FUNCTION_RETURN is no longer used, remove its
reference from target.def.

PR target/119985
* target.def: Remove TARGET_PROMOTE_FUNCTION_RETURN reference.
* doc/tm.texi: Regenerated.

Signed-off-by: H.J. Lu <hjl.tools@gmail.com>

Annotate empty bb with all debug_stmt with location of phi in the single_succ.

For an empty BB with all debug_stmt, it will be ignored by
afdo_set_bb_count, but it can be set with count of single successors
PHIs which edge from the BB.

gcc/ChangeLog:

PR gcov-profile/118581
* auto-profile.cc (autofdo_source_profile::get_count_info):
Overload the function with parameter gimple location instead
of stmt.
(afdo_set_bb_count): For !has_annotated BB, Check single
successors PHIs corresponding to the block and use those
count.

debug/78685 - reword -Og documentation

The following rewords the documentation for -Og which over-promises
the ability to debug the generated code.  While -Og enables
var-tracking and thus improves debugging in some areas the experience
is usually worse than -O0 for standard C code.

PR debug/78685
* doc/invoke.texi (-Og): Reword.

i386: Skip sub-RTXes of memory operand in ix86_update_stack_alignment

Skip sub-RTXes of the memory operand if stack access register is
not mentioned in the operand.

gcc/ChangeLog:

* config/i386/i386.cc (ix86_update_stack_alignment): Skip sub-RTXes
of the memory operand if stack access register is not mentioned in
the operand.

x86: Add a pass to remove redundant all 0s/1s vector load

For all different modes of all 0s/1s vectors, we can use the single widest
all 0s/1s vector register for all 0s/1s vector uses in the whole function.
Add a pass to generate a single widest all 0s/1s vector set instruction at
entry of the nearest common dominator for basic blocks with all 0s/1s
vector uses.  On Linux/x86-64, in cc1plus, this patch reduces the number
of vector xor instructions from 4803 to 4714 and pcmpeq instructions from
144 to 142.

NB: PR target/92080 and PR target/117839 aren't same.  PR target/117839
is for vectors of all 0s and all 1s with different sizes and different
components.  PR target/92080 is for broadcast of the same component to
different vector sizes.  This patch covers only all 0s and all 1s cases
of PR target/92080.

gcc/

PR target/92080
PR target/117839
* config/i386/i386-features.cc (ix86_place_single_vector_set):
New function.
(remove_partial_avx_dependency): Use it.
(ix86_get_vector_load_mode): New function.
(replace_vector_const): Likewise.
(remove_redundant_vector_load): Likewise.
(pass_data_remove_redundant_vector_load): Likewise.
(pass_remove_redundant_vector_load): Likewise.
(make_pass_remove_redundant_vector_load): Likewise.
* config/i386/i386-passes.def: Add
pass_remove_redundant_vector_load after
pass_remove_partial_avx_dependency.
* config/i386/i386-protos.h
(make_pass_remove_redundant_vector_load): New.
* config/i386/i386.cc (ix86_modes_tieable_p): Return true for
narrower non-scalar-integer modes in SSE registers.

gcc/testsuite/

PR target/92080
PR target/117839
* gcc.target/i386/pr117839-1a.c: New test.
* gcc.target/i386/pr117839-1b.c: Likewise.
* gcc.target/i386/pr117839-2.c: Likewise.
* gcc.target/i386/pr92080-1.c: Likewise.
* gcc.target/i386/pr92080-2.c: Likewise.
* gcc.target/i386/pr92080-3.c: Likewise.

Signed-off-by: H.J. Lu <hjl.tools@gmail.com>

i386: Add ix86_expand_unsigned_small_int_cst_argument

When passing 0xff as an unsigned char function argument with the C frontend
promotion, expand_normal used to get

<integer_cst 0x7fffe6aa23a8 type <integer_type 0x7fffe98225e8 int> constant 255>

and returned the rtx value using the sign-extended representation:

(const_int 255 [0xff])

But after

commit a670ebde3995481225ec62b29686ec07a21e5c10
Author: H.J. Lu <hjl.tools@gmail.com>
Date:   Thu Nov 21 07:54:35 2024 +0800

    Drop targetm.promote_prototypes from C, C++ and Ada frontends

expand_normal now gets

<integer_cst 0x7fffe9824018 type <integer_type 0x7fffe9822348 unsigned char > constant 255>

and returns

     (const_int -1 [0xffffffffffffffff])

which doesn't work with the predicates nor the instruction templates which
expect the unsigned expanded value.  Extract the unsigned char and short
integer constants to return

(const_int 255 [0xff])

so that the expanded value is always unsigned, without the C frontend
promotion.

PR target/117547
* config/i386/i386-expand.cc (ix86_expand_unsigned_small_int_cst_argument):
New function.
(ix86_expand_args_builtin): Call
ix86_expand_unsigned_small_int_cst_argument to expand the argument
before calling fixup_modeless_constant.
(ix86_expand_round_builtin): Likewise.
(ix86_expand_special_args_builtin): Likewise.
(ix86_expand_builtin): Likewise.

Signed-off-by: H.J. Lu <hjl.tools@gmail.com>

Remove other processors from X86_TUNE_DEST_FALSE_DEP_FOR_GLC except GLC

Since the tune if only for GLC(sapphirerapids and alderlake-P).

gcc/ChangeLog:

* config/i386/x86-tune.def (X86_TUNE_DEST_FALSE_DEP_FOR_GLC):
Remove other processor except for GLC since this one is only
for GLC.

Daily bump.

libstdc++: centralize and improve testing of shared_ptr/weak_ptr conversions

Since the conversions are under the same constraints, centralize the
test in one file instead of two, testing both smart pointer classes, to
ease future maintenance. This is used right away: more tests are added.
Amends r15-8048-gdf0e6509bf7442.

libstdc++-v3/ChangeLog:

* testsuite/20_util/shared_ptr/requirements/1.cc: Test both
shared_ptr and weak_ptr.
Add more tests.
* testsuite/20_util/weak_ptr/requirements/1.cc: Removed as
superseded by the other test.

Signed-off-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>

analyzer: handle NRVO and DECL_BY_REFERENCE [PR111536]

The analyzer was issuing false warnings about uninitialized variables
in C++ in places where NRVO was marking DECL_RESULT with
DECL_BY_REFERENCE.

Fixed thusly.

gcc/analyzer/ChangeLog:
PR analyzer/111536
* engine.cc (maybe_update_for_edge): Update for new call_stmt
param to region_model::push_frame.
* program-state.cc (program_state::push_frame): Likewise.
* region-model.cc (region_model::update_for_gcall): Likewise.
(region_model::push_frame): Add "call_stmt" param.
Handle DECL_RESULT with DECL_BY_REFERENCE set on it by stashing
the region of the lhs of the call_stmt in the caller frame,
and writing a reference to it within the "result" in the callee
frame.
(region_model::pop_frame): Don't write back to the LHS for
DECL_BY_REFERENCE results.
(selftest::test_stack_frames): Update for new call_stmt param to
region_model::push_frame.
(selftest::test_get_representative_path_var): Likewise.
(selftest::test_state_merging): Likewise.
(selftest::test_alloca): Likewise.
* region-model.h (region_model::push_frame): Add "call_stmt"
param.
* region.cc: Include "tree-ssa.h".
(region::can_have_initial_svalue_p): Use ssa_defined_default_def_p
for ssa names, rather than special-casing it for just parameters.
This should now also cover DECL_RESULT with DECL_BY_REFERENCE and
hard registers.
* sm-signal.cc (update_model_for_signal_handler): Update for new
call_stmt param to region_model::push_frame.
* state-purge.cc (state_purge_per_decl::process_worklists):
Likewise.

gcc/testsuite/ChangeLog:
PR analyzer/111536
* c-c++-common/analyzer/hard-reg-1.c: New test.
* g++.dg/analyzer/nrvo-1.C: New test.
* g++.dg/analyzer/nrvo-2.C: New test.
* g++.dg/analyzer/nrvo-pr111536-1.C: New test.
* g++.dg/analyzer/nrvo-pr111536-1b.C: New test.
* g++.dg/analyzer/nrvo-pr111536-2.C: New test.
* g++.dg/analyzer/nrvo-pr111536-2b.C: New test.

Signed-off-by: David Malcolm <dmalcolm@redhat.com>

analyzer: fix null deref false negative on std::unique_ptr [PR109366]

gcc/analyzer/ChangeLog:
PR analyzer/109366
* region-model-manager.cc
(region_model_manager::maybe_fold_sub_svalue): Sub-values of zero
constants are zero.

gcc/testsuite/ChangeLog:
PR analyzer/109366
* g++.dg/analyzer/unique_ptr-1.C: New test.
* g++.dg/analyzer/unique_ptr-2.C: New test.

Signed-off-by: David Malcolm <dmalcolm@redhat.com>

analyzer: initial implementation of exception handling [PR97111]

This patch adds initial support for exception-handling to -fanalyzer,
handling eh_dispatch for regions of type ERT_TRY and
ERT_ALLOWED_EXCEPTIONS.   I haven't managed yet seen eh_dispatch for
regions of type ERT_CLEANUP and ERT_MUST_NOT_THROW in the analyzer; with
this patch it will ICE if it sees those.

Additionally, this patch only checks for exact matches of exception
types, rather than supporting subclasses and references.  I'm deferring
fixing this for now whilst figuring out how best to interact with the C++
type system; I'm tracking it as PR analyzer/119697.

The patch adds event classes for throwing and catching exceptions, and
seems to generate readable warnings for the kinds of leak that might
occur due to trying to manage resources manually and forgetting about
exceptions; for example:

exception-leak-1.C: In function ‘int test()’:
exception-leak-1.C:7:9: warning: leak of ‘ptr’ [CWE-401] [-Wanalyzer-malloc-leak]
    7 |   throw 42;
      |         ^~
  ‘int test()’: events 1-3
    5 |   void *ptr = __builtin_malloc (1024);
      |               ~~~~~~~~~~~~~~~~~^~~~~~
      |                                |
      |                                (1) allocated here
    6 |
    7 |   throw 42;
      |         ~~
      |         |
      |         (2) throwing exception of type ‘int’ here...
      |         (3) ⚠️  ‘ptr’ leaks here; was allocated at (1)

Although dynamic exception specifications are only available in C++14
and earlier, the need to support them meant it seemed relatively easy to
add a warning to check them, hence the patch adds a new warning
for code paths that throw an exception that doesn't match a dynamic
exception specification:  -Wanalyzer-throw-of-unexpected-type.

gcc/analyzer/ChangeLog:
PR analyzer/97111
* analyzer.cc (is_cxa_throw_p): New.
(is_cxa_rethrow_p): New.
* analyzer.opt (Wanalyzer-throw-of-unexpected-type): New.
* analyzer.opt.urls: Regenerate.
* call-info.cc (custom_edge_info::create_enode): New.
* call-info.h (call_info::print): Drop "final".
(call_info::add_events_to_path): Likewise.
* checker-event.cc (event_kind_to_string): Add cases for
event_kind::catch_, event_kind::throw_, and event_kind::unwind.
(explicit_throw_event::print_desc): New.
(throw_from_call_to_external_fn_event::print_desc): New.
(unwind_event::print_desc): New.
* checker-event.h (enum class event_kind): Add catch_, throw_,
and unwind.
(class catch_cfg_edge_event): New.
(class throw_event): New.
(class explicit_throw_event): New.
(class throw_from_call_to_external_fn_event): New.
(class unwind_event): New.
* common.h (class eh_dispatch_cfg_superedge): New forward decl.
(class eh_dispatch_try_cfg_superedge): New forward decl.
(class eh_dispatch_allowed_cfg_superedge): New forward decl.
(custom_edge_info::create_enode): New vfunc decl.
(is_cxa_throw_p): New decl.
(is_cxa_rethrow_p): New decl.
* diagnostic-manager.cc
(diagnostic_manager::add_events_for_superedge): Special-case edges
for eh_dispach_try.
(diagnostic_manager::prune_path): Call consolidate_unwind_events.
(diagnostic_manager::prune_for_sm_diagnostic): Don't filter the new
event_kinds.
(diagnostic_manager::consolidate_unwind_events): New.
* diagnostic-manager.h
(diagnostic_manager::consolidate_unwind_events): New decl.
* engine.cc (exploded_node::on_stmt_pre): Handle "__cxa_throw",
"__cxa_rethrow", and resx statements.
(class throw_custom_edge): New.
(class unwind_custom_edge): New.
(get_eh_outedge): New.
(exploded_graph::unwind_from_exception): New.
(exploded_node::on_throw): New.
(exploded_node::on_resx): New.
(exploded_graph::get_or_create_node): Add "add_to_worklist" param
and use it.
(exploded_graph::process_node): Use edge_info's create_enode vfunc
to create enodes, rather than calling get_or_create_node directly.
Ignore CFG edges in the sgraph flagged with EH whilst we're
exploring the egraph.
(exploded_graph_annotator::print_enode): Handle case
exploded_node::status::special.
* exploded-graph.h (exploded_node::status): Add value "special".
(exploded_node::on_throw): New decl.
(exploded_node::on_resx): New decl.
(exploded_graph::get_or_create_node): Add optional
"add_to_worklist" param.
(exploded_graph::unwind_from_exception): New decl.
* kf-lang-cp.cc (class kf_cxa_allocate_exception): New.
(class kf_cxa_begin_catch): New.
(class kf_cxa_end_catch): New.
(class throw_of_unexpected_type): New.
(class kf_cxa_call_unexpected): New.
(register_known_functions_lang_cp): Register known functions
"__cxa_allocate_exception", "__cxa_begin_catch",
"__cxa_end_catch", and "__cxa_call_unexpected".
* kf.cc (class kf_eh_pointer): New.
(register_known_functions): Register it for BUILT_IN_EH_POINTER.
* region-model.cc: Include "analyzer/function-set.h".
(exception_node::operator==): New.
(exception_node::dump_to_pp): New.
(exception_node::dump): New.
(exception_node::to_json): New.
(exception_node::make_dump_widget): New.
(exception_node::maybe_get_type): New.
(exception_node::add_to_reachable_regions): New.
(region_model::region_model): Initialize
m_thrown_exceptions_stack and m_caught_exceptions_stack.
(region_model::operator=): Likewise.
(region_model::operator==): Compare them.
(region_model::dump_to_pp): Dump exception stacks.
(region_model::to_json): Add exception stacks.
(region_model::make_dump_widget): Likewise.
(class exception_thrown_from_unrecognized_call): New.
(get_fns_assumed_not_to_throw): New.
(can_throw_p): New.
(region_model::check_for_throw_inside_call): New.
(region_model::on_call_pre): Call check_for_throw_inside_call
on unknown fns or those we don't have a body for.
(region_model::maybe_update_for_edge): Handle eh_dispatch_stmt
statements.  Drop old code that called
apply_constraints_for_exception on EDGE_EH edges.
(class rejected_eh_dispatch): New.
(exception_matches_type_p): New.
(matches_any_exception_type_p): New.
(region_model::apply_constraints_for_eh_dispatch): New.
(region_model::apply_constraints_for_eh_dispatch_try): New.
(region_model::apply_constraints_for_eh_dispatch_allowed): New.
(region_model::apply_constraints_for_exception): Delete.
(region_model::can_merge_with_p): Don't merge models with
non-equal exception stacks.
(region_model::get_referenced_base_regions): Add regions from
exception stacks.
* region-model.h (struct exception_node): New.
(region_model::push_thrown_exception): New.
(region_model::get_current_thrown_exception): New.
(region_model::pop_thrown_exception): New.
(region_model::push_caught_exception): New.
(region_model::get_current_caught_exception): New.
(region_model::pop_caught_exception): New.
(region_model::apply_constraints_for_eh_dispatch_try): New decl.
(region_model::apply_constraints_for_eh_dispatch_allowed) New decl.
(region_model::apply_constraints_for_exception): Delete.
(region_model::apply_constraints_for_eh_dispatch): New decl.
(region_model::check_for_throw_inside_call): New decl.
(region_model::m_thrown_exceptions_stack): New field.
(region_model::m_caught_exceptions_stack): New field.
* supergraph.cc: Include "except.h" and "analyzer/region-model.h".
(supergraph::add_cfg_edge): Special-case eh_dispatch edges.
(superedge::get_description): Use default_tree_printer.
(get_catch): New.
(eh_dispatch_cfg_superedge::make): New.
(eh_dispatch_cfg_superedge::eh_dispatch_cfg_superedge): New.
(eh_dispatch_cfg_superedge::get_eh_status): New.
(eh_dispatch_try_cfg_superedge::dump_label_to_pp): New.
(eh_dispatch_try_cfg_superedge::apply_constraints): New.
(eh_dispatch_allowed_cfg_superedge::eh_dispatch_allowed_cfg_superedge):
New.
(eh_dispatch_allowed_cfg_superedge::dump_label_to_pp): New.
(eh_dispatch_allowed_cfg_superedge::apply_constraints): New.
* supergraph.h: Include "except.h".
(superedge::dyn_cast_eh_dispatch_cfg_superedge): New vfunc.
(superedge::dyn_cast_eh_dispatch_try_cfg_superedge): New vfunc.
(superedge::dyn_cast_eh_dispatch_allowed_cfg_superedge): New
vfunc.
(class eh_dispatch_cfg_superedge): New.
(is_a_helper <const eh_dispatch_cfg_superedge *>::test): New.
(class eh_dispatch_try_cfg_superedge): New.
(is_a_helper <const eh_dispatch_try_cfg_superedge *>::test): New.
(class eh_dispatch_allowed_cfg_superedge): New.
(is_a_helper <const eh_dispatch_allowed_cfg_superedge *>::test):
New.
* svalue.cc (svalue::maybe_get_type_from_typeinfo): New.
* svalue.h (svalue::maybe_get_type_from_typeinfo): New decl.

gcc/ChangeLog:
PR analyzer/97111
* doc/invoke.texi: Add -Wanalyzer-throw-of-unexpected-type.
* gimple.h (gimple_call_nothrow_p): Make arg const.

gcc/testsuite/ChangeLog:
PR analyzer/97111
* c-c++-common/analyzer/analyzer-verbosity-2a.c: Add
-fno-exceptions.
* c-c++-common/analyzer/analyzer-verbosity-3a.c: Likewise.
* c-c++-common/analyzer/attr-const-2.c: Add
__attribute__((nothrow)).
* c-c++-common/analyzer/attr-malloc-4.c: Likewise.
* c-c++-common/analyzer/attr-malloc-5.c: Likewise.
* c-c++-common/analyzer/attr-malloc-6.c: Add -fno-exceptions.
* c-c++-common/analyzer/attr-malloc-CVE-2019-19078-usb-leak.c:
Likewise.
* c-c++-common/analyzer/attr-malloc-exception.c: New test.
* c-c++-common/analyzer/call-summaries-pr107158-2.c: Add
-fno-exceptions.
* c-c++-common/analyzer/call-summaries-pr107158.c: Likewise.
* c-c++-common/analyzer/capacity-2.c: Likewise.
* c-c++-common/analyzer/coreutils-sum-pr108666.c: Likewise.
* c-c++-common/analyzer/data-model-22.c: Likewise.
* c-c++-common/analyzer/data-model-5d.c: Likewise.
* c-c++-common/analyzer/deref-before-check-pr108455-git-pack-revindex.c:
Likewise.
* c-c++-common/analyzer/deref-before-check-pr108475-haproxy-tcpcheck.c:
Likewise.
* c-c++-common/analyzer/edges-2.c: Likewise.
* c-c++-common/analyzer/fd-2.c: Likewise.
* c-c++-common/analyzer/fd-3.c: Likewise.
* c-c++-common/analyzer/fd-meaning.c: Likewise.
* c-c++-common/analyzer/file-1.c: Likewise.
* c-c++-common/analyzer/file-3.c: Likewise.
* c-c++-common/analyzer/file-meaning-1.c: Likewise.
* c-c++-common/analyzer/infinite-recursion.c: Likewise.
* c-c++-common/analyzer/leak-3.c: Likewise.
* c-c++-common/analyzer/malloc-dedupe-1.c: Likewise.
* c-c++-common/analyzer/malloc-in-loop.c: Likewise.
* c-c++-common/analyzer/malloc-many-paths-3.c: Likewise.
* c-c++-common/analyzer/malloc-paths-5.c: Likewise.
* c-c++-common/analyzer/malloc-paths-7.c: Likewise.
* c-c++-common/analyzer/malloc-paths-8.c: Likewise.
* c-c++-common/analyzer/malloc-vs-local-1a.c: Likewise.
* c-c++-common/analyzer/malloc-vs-local-2.c: Likewise.
* c-c++-common/analyzer/malloc-vs-local-3.c: Likewise.
* c-c++-common/analyzer/paths-7.c: Likewise.
* c-c++-common/analyzer/pr110830.c: Likewise.
* c-c++-common/analyzer/pr93032-mztools-simplified.c: Likewise.
* c-c++-common/analyzer/pr93355-localealias-feasibility-3.c:
Likewise.
* c-c++-common/analyzer/pr93355-localealias-simplified.c:
Likewise.
* c-c++-common/analyzer/pr96650-1-trans.c: Likewise.
* c-c++-common/analyzer/pr97072.c: Add __attribute__((nothrow)).
* c-c++-common/analyzer/pr98575-1.c: Likewise.
* c-c++-common/analyzer/pr99716-1.c: Add -fno-exceptions.
* c-c++-common/analyzer/pr99716-2.c: Likewise.
* c-c++-common/analyzer/pr99716-3.c: Likewise.
* c-c++-common/analyzer/pragma-2.c: Likewise.
* c-c++-common/analyzer/rhbz1878600.c: Likewise.
* c-c++-common/analyzer/strndup-1.c: Likewise.
* c-c++-common/analyzer/write-to-string-literal-4-disabled.c:
Likewise.
* c-c++-common/analyzer/write-to-string-literal-4.c: Likewise.
* c-c++-common/analyzer/write-to-string-literal-5.c: Likewise.
* c-c++-common/analyzer/zlib-5.c: Likewise.
* g++.dg/analyzer/exception-could-throw-1.C: New test.
* g++.dg/analyzer/exception-could-throw-2.C: New test.
* g++.dg/analyzer/exception-dynamic-spec.C: New test.
* g++.dg/analyzer/exception-leak-1.C: New test.
* g++.dg/analyzer/exception-leak-2.C: New test.
* g++.dg/analyzer/exception-leak-3.C: New test.
* g++.dg/analyzer/exception-leak-4.C: New test.
* g++.dg/analyzer/exception-leak-5.C: New test.
* g++.dg/analyzer/exception-leak-6.C: New test.
* g++.dg/analyzer/exception-nothrow.C: New test.
* g++.dg/analyzer/exception-path-1.C: New test.
* g++.dg/analyzer/exception-path-catch-all-1.C: New test.
* g++.dg/analyzer/exception-path-catch-all-2.C: New test.
* g++.dg/analyzer/exception-path-unwind-multiple-2.C: New test.
* g++.dg/analyzer/exception-path-unwind-multiple.C: New test.
* g++.dg/analyzer/exception-path-unwind-single.C: New test.
* g++.dg/analyzer/exception-path-with-cleanups.C: New test.
* g++.dg/analyzer/exception-rethrow-1.C: New test.
* g++.dg/analyzer/exception-rethrow-2.C: New test.
* g++.dg/analyzer/exception-stack-1.C: New test.
* g++.dg/analyzer/exception-stack-2.C: New test.
* g++.dg/analyzer/exception-subclass-1.C: New test.
* g++.dg/analyzer/exception-subclass-2.C: New test.
* g++.dg/analyzer/exception-value-1.C: New test.
* g++.dg/analyzer/exception-value-2.C: New test.
* g++.dg/analyzer/fno-exception.C: New test.
* g++.dg/analyzer/pr94028.C: Drop xfail.
* g++.dg/analyzer/std-unexpected.C: New test.
* g++.dg/coroutines/pr105287.C: Drop dg-excess-errors.

Signed-off-by: David Malcolm <dmalcolm@redhat.com>

analyzer,c++: add placeholder implementation of ana::translation_unit for C++

Implement ana::translation_unit for the C++ frontend with a
no-op placeholder implementation, for now.

No functional change intended; a follow-up may implement
things further.

gcc/cp/ChangeLog:
* parser.cc: Include "analyzer/analyzer-language.h".
(ana::cp_translation_unit): New class.
(cp_parser_translation_unit): Add call to
ana::on_finish_translation_unit.

Signed-off-by: David Malcolm <dmalcolm@redhat.com>

Eliminate make-unique.h and ::make_unique

C++11 does not provide a std::make_unique so in
r13-3627-g00d7c8ff16e683 I added a make-unique.h
declaring a ::make_unique.

As of r15-4719-ga9ec1bc06bd3cc we can use C++14, so make-unique.h is no
longer needed: we can use simply use std::make_unique instead.

This patch removes make-unique.h and updates every place using it
to use std::make_unique.

No functional change intended.

gcc/analyzer/ChangeLog:
* access-diagram.cc: Replace uses of ::make_unique with
std::make_unique.
* analyzer.cc: Likewise.
* bounds-checking.cc: Likewise.
* call-details.cc: Likewise.
* call-info.cc: Likewise.
* call-string.cc: Likewise.
* checker-path.cc: Likewise.
* common.h: Drop include of "make-unique.h".
* constraint-manager.cc: Replace uses of ::make_unique with
std::make_unique.
* diagnostic-manager.cc: Likewise.
* engine.cc: Likewise.
* infinite-loop.cc: Likewise.
* infinite-recursion.cc: Likewise.
* kf-analyzer.cc: Likewise.
* kf-lang-cp.cc: Likewise.
* kf.cc: Likewise.
* pending-diagnostic.cc: Likewise.
* program-point.cc: Likewise; drop #include.
* program-state.cc: Likewise.
* ranges.cc: Likewise.
* region-model.cc: Likewise.
* region.cc: Likewise; drop #include.
* sm-fd.cc: Likewise.
* sm-file.cc: Likewise.
* sm-malloc.cc: Likewise.
* sm-pattern-test.cc: Likewise.
* sm-sensitive.cc: Likewise.
* sm-signal.cc: Likewise.
* sm-taint.cc: Likewise.
* sm.cc: Likewise.
* store.cc: Likewise.
* supergraph.cc: Likewise.
* svalue.cc: Likewise; drop #include.
* varargs.cc: Likewise.

gcc/c-family/ChangeLog:
* c-pretty-print.cc: Drop include of "make-unique.h".
Replace uses of ::make_unique with std::make_unique.

gcc/c/ChangeLog:
* c-decl.cc: Drop include of "make-unique.h".
Replace uses of ::make_unique with std::make_unique.
* c-objc-common.cc: Likewise.
* c-parser.cc: Likewise.

gcc/cp/ChangeLog:
* cxx-pretty-print.cc: Drop include of "make-unique.h".
Replace uses of ::make_unique with std::make_unique.
* error.cc: Likewise.
* name-lookup.cc: Likewise.
* parser.cc: Likewise.

gcc/ChangeLog:
* diagnostic-format-json.cc: Drop include of "make-unique.h".
Replace uses of ::make_unique with std::make_unique.
* diagnostic-format-sarif.cc: Likewise.
* diagnostic-format-text.cc: Likewise.
* diagnostic.cc: Likewise.
* dumpfile.cc: Likewise.
* gcc-attribute-urlifier.cc: Likewise.
* gcc-urlifier.cc: Likewise.
* json-parsing.cc: Likewise.
* json.cc: Likewise.
* lazy-diagnostic-path.cc: Likewise.
* libgdiagnostics.cc: Likewise.
* libsarifreplay.cc: Likewise.
* lto-wrapper.cc: Likewise.
* make-unique.h: Delete.
* opts-diagnostic.cc: Drop include of "make-unique.h".
Replace uses of ::make_unique with std::make_unique.
* pretty-print.cc: Likewise.
* text-art/style.cc: Likewise.
* text-art/styled-string.cc: Likewise.
* text-art/table.cc: Likewise.
* text-art/tree-widget.cc: Likewise.
* text-art/widget.cc: Likewise.
* timevar.cc: Likewise.
* toplev.cc: Likewise.
* tree-diagnostic-client-data-hooks.cc: Likewise.

gcc/jit/ChangeLog:
* dummy-frontend.cc: Drop include of "make-unique.h".
Replace uses of ::make_unique with std::make_unique.

gcc/testsuite/ChangeLog:
* gcc.dg/plugin/analyzer_cpython_plugin.cc: Drop include of
"make-unique.h".  Replace uses of ::make_unique with
std::make_unique.
* gcc.dg/plugin/analyzer_gil_plugin.cc: Likewise.
* gcc.dg/plugin/analyzer_kernel_plugin.cc: Likewise.
* gcc.dg/plugin/analyzer_known_fns_plugin.cc: Likewise.
* gcc.dg/plugin/diagnostic_group_plugin.cc: Likewise.
* gcc.dg/plugin/diagnostic_plugin_xhtml_format.cc: Likewise.

Signed-off-by: David Malcolm <dmalcolm@redhat.com>

analyzer: use unique_ptr for state_machine instances

gcc/analyzer/ChangeLog:
* engine.cc (class plugin_analyzer_init_impl): Convert
"m_checkers" to use std::vector of std::unique_ptr.  Convert
"m_known_fn_mgr" to a reference.
(impl_run_checkers): Convert "checkers" to use std::vector of
std::unique_ptr and move it into the extrinsic_state.
* program-state.cc (extrinsic_state::dump_to_pp): Update for
changes to m_checkers.
(extrinsic_state::to_json): Likewise.
(extrinsic_state::get_sm_idx_by_name): Likewise.
(selftest::test_sm_state_map): Update to use std::unique_ptr
for state machines.
(selftest::test_program_state_1): Likewise.
(selftest::test_program_state_2): Likewise.
(selftest::test_program_state_merging): Likewise.
(selftest::test_program_state_merging_2): Likewise.
* program-state.h (class extrinsic_state): Convert "m_checkers" to
use std::vector of std::unique_ptr and to be owned by this object,
rather than a reference.  Add ctor for use in selftests.
* sm-fd.cc (make_fd_state_machine): Update to use std::unique_ptr.
* sm-file.cc (make_fileptr_state_machine): Likewise.
* sm-malloc.cc (make_malloc_state_machine): Likewise.
* sm-pattern-test.cc (make_pattern_test_state_machine): Likewise.
* sm-sensitive.cc (make_sensitive_state_machine): Likewise.
* sm-signal.cc (make_signal_state_machine): Likewise.
* sm-taint.cc (make_taint_state_machine): Likewise.
* sm.cc: Define INCLUDE_LIST.
(make_checkers): Return the vector directly, rather than pass it
in by reference.  Update to use std::unique_ptr throughout.  Use
an intermediate list, and use that to filter with
flag_analyzer_checker, fixing memory leak for this case.
* sm.h: (make_checkers): Return the vector directly, rather than
pass it in by reference, and use std::vector of std::unique_ptr.
(make_malloc_state_machine): Convert return type to use std::unique_ptr.
(make_fileptr_state_machine): Likewise.
(make_taint_state_machine): Likewise.
(make_sensitive_state_machine): Likewise.
(make_signal_state_machine): Likewise.
(make_pattern_test_state_machine): Likewise.
(make_va_list_state_machine): Likewise.
(make_fd_state_machine): Likewise.
* varargs.cc (make_va_list_state_machine): Update to use
std::unique_ptr.

Signed-off-by: David Malcolm <dmalcolm@redhat.com>

analyzer: convert various call_summary * to call_summary &

No functional change intended.

gcc/analyzer/ChangeLog:
* call-summary.cc (call_summary_replay::call_summary_replay):
Convert "summary" from call_summary * to call_summary &.
(call_summary_replay::dump_to_pp): Likewise for m_summary.
* call-summary.h (call_summary_replay::call_summary_replay):
Likewise for "summary".
(call_summary_replay::m_summary): Likewise.
* engine.cc (call_summary_edge_info::call_summary_edge_info):
Likewise.
(call_summary_edge_info::update_state): Likewise.
(call_summary_edge_info::update_model): Likewise.
(call_summary_edge_info::print_desc): Likewise for m_summary.
(call_summary_edge_info::m_summary): Likewise.
(exploded_node::replay_call_summaries): Update for change to
replay_call_summary.
(exploded_node::replay_call_summary): Convert "summary" from
call_summary * to call_summary &.
* exploded-graph.h (exploded_node::replay_call_summary): Likewise.

Signed-off-by: David Malcolm <dmalcolm@redhat.com>

analyzer: convert gcall * to gcall & in many places

No functional change intended.

gcc/analyzer/ChangeLog:
* analyzer.cc: Convert gcall * to gcall & where we know the
pointer must be non-null.
* call-details.cc: Likewise.
* call-details.h: Likewise.
* call-info.cc: Likewise.
* call-info.h: Likewise.
* call-summary.h: Likewise.
* checker-event.cc: Likewise.
* checker-event.h: Likewise.
* common.h: Likewise.
* diagnostic-manager.cc: Likewise.
* engine.cc: Likewise.
* exploded-graph.h: Likewise.
* kf-analyzer.cc: Likewise.
* kf-lang-cp.cc: Likewise.
* kf.cc: Likewise.
* known-function-manager.cc: Likewise.
* program-state.cc: Likewise.
* program-state.h: Likewise.
* region-model.cc: Likewise.
* region-model.h: Likewise.
* sm-fd.cc: Likewise.
* sm-file.cc: Likewise.
* sm-malloc.cc: Likewise.
* sm-sensitive.cc: Likewise.
* sm-signal.cc: Likewise.
* sm-taint.cc: Likewise.
* sm.h: Likewise.
* store.cc: Likewise.
* store.h: Likewise.
* supergraph.cc: Likewise.
* supergraph.h: Likewise.
* svalue.h: Likewise.
* varargs.cc: Likewise.

gcc/testsuite/ChangeLog:
* gcc.dg/plugin/analyzer_gil_plugin.cc: Convert gcall * to gcall &
where we know the pointer must be non-null.

Signed-off-by: David Malcolm <dmalcolm@redhat.com>

analyzer: convert various enums to "enum class"

Modernization; no functional change intended.

gcc/analyzer/ChangeLog:
* access-diagram.cc: Convert enum access_direction to
"enum class".
* bounds-checking.cc: Likewise.
* checker-event.cc: Convert enum event_kind to "enum class".
* checker-event.h: Likewise.
* checker-path.cc: Likewise.
* common.h: Convert enum access_direction to "enum class".
* constraint-manager.cc: Convert enum bound_kind to "enum class".
* constraint-manager.h: Likewise.
* diagnostic-manager.cc: Convert enum event_kind to "enum class".
* engine.cc: Convert enum status to "enum class".
* exploded-graph.h: Likewise.
* infinite-loop.cc: Likewise.
* kf-lang-cp.cc: Convert enum poison_kind to "enum class".
* kf.cc: Likewise.
* region-model-manager.cc: Likewise.
* region-model.cc: Likewise; also for enum access_direction.
* svalue.cc: Likewise.
* svalue.h: Likewise.

gcc/testsuite/ChangeLog:
* gcc.dg/plugin/analyzer_cpython_plugin.cc: Convert
enum poison_kind to "enum class".

Signed-off-by: David Malcolm <dmalcolm@redhat.com>

analyzer: use analyzer/common.h as a common header

Our headers are a major pain to work with: many require certain other
headers to be included in a particular (undocumented) order in order
to be includable.

Simplify includes in the analyzer by renaming analyzer/analyzer.h to
analyzer/common.h and have it include all the common headers needed
throughout the analyzer, thus encapsulating the rules about e.g. being
able to include "gimple.h" in one place in the analyzer subdirectory.

Doing so also makes it easier to e.g. define INCLUDE_SET in one place,
rather than in many source files.

gcc/analyzer/ChangeLog:
* analyzer.h: Rename to...
* common.h: ...this.  Add define of INCLUDE_VECTOR, includes of
"config.h", "system.h", "coretypes.h", "make-unique.h", "tree.h",
"function.h", "basic-block.h", "gimple.h", "options.h",
"bitmap.h", "diagnostic-core.h", and "diagnostic-path.h".
* access-diagram.h: Don't include "analyzer/analyzer.h".
* access-diagram.cc: Reorganize includes to #include
"analyzer/common.h" first, then group by subsystem, dropping
redundant headers.
* analysis-plan.cc: Likewise.
* analyzer-language.cc: Likewise.
* analyzer-pass.cc: Likewise.
* analyzer-selftests.cc: Likewise.
* analyzer.cc: Likewise.
* bounds-checking.cc: Likewise.
* call-details.cc: Likewise.
* call-info.cc: Likewise.
* call-string.cc: Likewise.
* call-summary.cc: Likewise.
* checker-event.cc: Likewise.
* checker-path.cc: Likewise.
* complexity.cc: Likewise.
* constraint-manager.cc: Likewise.
* diagnostic-manager.cc: Likewise.
* engine.cc: Likewise.
* feasible-graph.cc: Likewise.
* infinite-loop.cc: Likewise.
* infinite-recursion.cc: Likewise.
* kf-analyzer.cc: Likewise.
* kf-lang-cp.cc: Likewise.
* kf.cc: Likewise.
* known-function-manager.cc: Likewise.
* pending-diagnostic.cc: Likewise.
* program-point.cc: Likewise.
* program-state.cc: Likewise.
* ranges.cc: Likewise.
* record-layout.cc: Likewise.
* region-model-asm.cc: Likewise.
* region-model-manager.cc: Likewise.
* region-model-reachability.cc: Likewise.
* region-model.cc: Likewise.
* region.cc: Likewise.
* sm-fd.cc: Likewise.
* sm-file.cc: Likewise.
* sm-malloc.cc: Likewise.
* sm-pattern-test.cc: Likewise.
* sm-sensitive.cc: Likewise.
* sm-signal.cc: Likewise.
* sm-taint.cc: Likewise.
* sm.cc: Likewise.
* state-purge.cc: Likewise.
* store.cc: Likewise.
* supergraph.cc: Likewise.
* svalue.cc: Likewise.
* symbol.cc: Likewise.
* trimmed-graph.cc: Likewise.
* varargs.cc: Likewise.

gcc/testsuite/ChangeLog:
* gcc.dg/plugin/analyzer_cpython_plugin.cc: Update for renaming
of analyzer/analyzer.h to analyzer/common.h.
* gcc.dg/plugin/analyzer_gil_plugin.cc: Likewise.
* gcc.dg/plugin/analyzer_kernel_plugin.cc: Likewise.
* gcc.dg/plugin/analyzer_known_fns_plugin.cc: Likewise.

Signed-off-by: David Malcolm <dmalcolm@redhat.com>

analyzer: fail if we see unexpected gimple stmt codes

gcc/analyzer/ChangeLog:
* region-model.cc (region_model::on_stmt_pre): Use internal_error
if we see an unexpected gimple stmt code.

Signed-off-by: David Malcolm <dmalcolm@redhat.com>

analyzer: add a call_details::dump (using text_art)

Add a new function to help debugging -fanalyzer.

No functional change intended.

gcc/analyzer/ChangeLog:
* call-details.cc (call_details::dump): New overload.
(call_details::make_dump_widget): New.
* call-details.h (call_details::dump): Declare new overload.
(call_details::make_dump_widget): New decl.

Signed-off-by: David Malcolm <dmalcolm@redhat.com>

pass_manager: add m_ prefix to pass fields accessed via passes.def

Make it clearer when we're accessing member data of pass_manager by
adding an "m"_ prefix to the fields handled via passes.def macros.

No functional change intended.

gcc/ChangeLog:
* pass_manager.h (class pass_manager): Add "m_" prefix to all pass
fields.
* passes.cc (pass_manager::execute_early_local_passes): Update
for added "m_" prefix.
(pass_manager::execute_pass_mode_switching): Likewise.
(pass_manager::finish_optimization_passes): Likewise.
(pass_manager::pass_manager): Likewise.
(pass_manager::dump_profile_report): Likewise.

Signed-off-by: David Malcolm <dmalcolm@redhat.com>

c,c++: use unique_ptr in name_hint to reduce naked 'new'

gcc/c-family/ChangeLog:
* name-hint.h (name_hint::name_hint): Use std::unique_ptr for
param.

gcc/c/ChangeLog:
* c-decl.cc: Include "make-unique.h".
(lookup_name_fuzzy): Use ::make_unique rather than "new" when
making suggest_missing_header and suggest_missing_option.
* c-parser.cc: Include "make-unique.h"
(c_parser_error_richloc): Use ::make_unique rather than "new" when
making suggest_missing_header.

gcc/cp/ChangeLog:
* name-lookup.cc: Include "make-unique.h".
(namespace_hints::convert_candidates_to_name_hint): Use
::make_unique rather than "new" when making
show_candidate_location and suggest_alternatives.
(namespace_hints::maybe_decorate_with_limit): Likewise when making
namespace_limit_reached.
(suggest_alternatives_for_1): Likewise when making
suggest_missing_option.
(maybe_suggest_missing_std_header): Likewise when making
missing_std_header.
(macro_use_before_def::maybe_make): Use std::unique_ptr.
(macro_use_before_def::macro_use_before_def): Make public.
(lookup_name_fuzzy): Use ::make_unique rather than "new" when
making suggest_missing_header.
* parser.cc: Include "make-unique.h".
(cp_parser_error_1): Use ::make_unique rather than "new" when
making suggest_missing_header.

Signed-off-by: David Malcolm <dmalcolm@redhat.com>

diagnostics: use diagnostic_option_id for m_opt_permissive

gcc/ChangeLog:
* diagnostic.h (diagnostic_context::m_opt_permissive): Convert
from int to diagnostic_option_id.  Update comment.

Signed-off-by: David Malcolm <dmalcolm@redhat.com>

diagnostics: make diagnostic_context::m_abort_on_error private

No functional change intended.

gcc/ChangeLog:
* diagnostic.h (diagnostic_context::set_abort_on_error): New.
(diagnostic_context::m_abort_on_error): Make private.
(diagnostic_abort_on_error): Delete.
* opts.cc (setup_core_dumping): Update for above changes.

Signed-off-by: David Malcolm <dmalcolm@redhat.com>

sarif output: introduce sarif_generation_options

Signed-off-by: David Malcolm <dmalcolm@redhat.com>
gcc/ChangeLog:
* diagnostic-format-sarif.cc (sarif_builder::get_opts): New
accessor.
(sarif_builder::get_version): Update for...
(sarif_builder::m_version): Replace this field...
(sarif_builder::m_m_sarif_gen_opts): ...with this.
(sarif_builder::sarif_builder): Replace version with
sarif_gen_opts throughout.
(sarif_builder::make_top_level_object): Use get_version.
(sarif_output_format::sarif_output_format): Replace version with
sarif_gen_opts throughout.
(sarif_stream_output_format::sarif_stream_output_format):
Likewise.
(sarif_file_output_format::sarif_file_output_format): Likewise.
(diagnostic_output_format_init_sarif_stderr): Drop version param
and use default for sarif_generation_options instead.
(diagnostic_output_format_init_sarif_file): Likewise.
(diagnostic_output_format_init_sarif_stream): Likewise.
(make_sarif_sink): Replace version with sarif_gen_opts throughout.
(sarif_generation_options::sarif_generation_options): New.
(selftest::test_sarif_diagnostic_context::test_sarif_diagnostic_context):
Replace version with sarif_gen_opts throughout.
(selftest::test_make_location_object): Likewise.
(selftest::test_simple_log): Likewise.
(selftest::test_simple_log_2): Likewise.
(selftest::test_message_with_embedded_link): Likewise.
(selftest::test_message_with_braces): Likewise.
(selftest::test_buffering): Likewise.
(selftest::run_tests_per_version): Replace with...
(selftest::for_each_sarif_gen_option): ...this...
(selftest::run_line_table_case_tests_per_version): ...and this.
(selftest::diagnostic_format_sarif_cc_tests): Update to use
for_each_sarif_gen_option and
run_line_table_case_tests_per_version.
* diagnostic-format-sarif.h (enum class sarif_version): Move lower
down.
(diagnostic_output_format_init_sarif_stderr): Drop "version"
param.
(diagnostic_output_format_init_sarif_file): Likewise.
(diagnostic_output_format_init_sarif_stream): Likewise.
(struct sarif_generation_options): New.
(make_sarif_sink): Add "formatted" param.  Replace version param
with sarif_gen_opts.
* diagnostic.cc (diagnostic_output_format_init): Drop hardcoded
sarif_version::v2_1_0 arguments from calls, which instead use
the default ctor for sarif_generation_options internally.
* libgdiagnostics.cc (sarif_sink::sarif_sink): Replace version
param with sarif_gen_opts, and update for changes to
make_sarif_sink.
(diagnostic_manager_add_sarif_sink): Use sarif_gen_opts rather
than version.
* opts-diagnostic.cc (sarif_scheme_handler::make_sink): Likewise.
Pass "true" for "formatted" param.

Signed-off-by: David Malcolm <dmalcolm@redhat.com>

Add a Relation iterator to the relation oracle.

This patch adds a relation iterator to query the oracle to list either
all the relations on exit to a block, or just ones involving a specified
SSA_NAME.   The oracle then uses this iterator internally as well.

* value-relation.cc (value_relation::swap): New.
(value_relation::negate): Remove.
(dom_oracle::next_relation): New.
(block_relation_iterator::block_relation_iterator): New.
(block_relation_iterator::get_next_relation): New.
(dom_oracle::dump): Use iterator.
* value-relation.h (relation_oracle::next_relation): New.
(dom_oracle::next_relation): New prototype.
(class block_relation_iterator): New.
(FOR_EACH_RELATION_BB): New.
(FOR_EACH_RELATION_NAME): New.

Add lhs_op1 relation to pointer_plus

When prange was split from irange, the functionality of lhs_op1_relation
did not get ported.  This patch adds that functionality back, and is
also good example of how to add new dispatch patterns to range-ops
as lhs_op1_relation had no prange/prange/irange combination.

* range-op-ptr.cc (range_operator::lhs_op1_relation): Add
prange/prange/irange (PPI) default.
(pointer_plus_operator::lhs_op1_relation): New.
* range-op.cc (range_op_handler::lhs_op1_relation): Add RO_PPI case.
* range-op.h (range_op_handler::lhs_op1_relation): Add prototype.

Do not overwrite relation in range_of_range_op.

when registering reltions between the lhs and op1/op2, the relation
between op1 and op2 is being overwritten by the result.  This could
result in either an incorrect relation being registered between lhs and op2,
or a correct relation not being recognized.

* gimple-range-fold.cc (fold_using_range::range_of_range_op): Use a
new local variable for intermediate relation results.

Use the current cache when creating inferred ranges.

Infer range processing was adjusted to allow a query to be specified,
but during VRP folding, ranger was not providing a query.  This results
in contextual ranges being missed.   Pass the cache in as the query
which provide a read-only query of the current state.

* gimple-range-cache.cc (ranger_cache::apply_inferred_ranges): Pass
'this' as the range-query to the inferred range constructor.

Infer non-zero for integral division RHS.

Adding op2_range for operator_div allows ranger to notice the divisor
is non-zero after execution.

PR tree-optimization/95801
gcc/
* range-op.cc (operator_div::op2_range): New.

gcc/testsuite/
* gcc.dg/tree-ssa/pr95801.c: New.

Always reflect lower bits from mask in subranges.

During intersection, we expand the subranges to exclude the lower values
from a bitmask with trailing zeros.  This leads to inconsistant evaluations
and in this case of this PR, that lead to an infinite cycle.

Always expand the lower subranges in set_range_from_bitmask instead.

PR tree-optimization/119712
gcc/
* value-range.cc (range_bitmask::adjust_range): Delete.
(irange::set_range_from_bitmask): Integrate adjust_range.
(irange::update_bitmask): Do nothing if bitmask doesnt change.
(irange:intersect_bitmask): Do not call adjust_range. Exit if there
is no second bitmask.
* value-range.h (adjust_range): Remove prototype.

gcc/testsuite/
* gcc.dg/pr119712.c: New.
* gcc.dg/pr83072-2.c: Adjust.
* gcc.dg/tree-ssa/phi-opt-value-5.c: Adjust.
* gcc.dg/tree-ssa/vrp122.c: Adjust

tailcall: Support ERF_RETURNS_ARG for tailcall [PR67797]

r15-6943-g9c4397cafc5ded added support to undo IPA-VRP return value optimization for tail calls,
using the same code ERF_RETURNS_ARG can be supported for functions which return one of their arguments.
This allows for tail calling of memset/memcpy in some cases which were not handled before.

Note this is very similar to https://gcc.gnu.org/legacy-ml/gcc-patches/2016-11/msg02485.html except
it has a few more checks.  Also on the question of expand vs tail call here is that this path is also
used by the IPA-VRP return value path and yes we get a tail call.
Note in the review in https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83142#c2 mentions about
re-instantiate a LHS on the call & propagate to dominating uses. Even though that can be done
for the ERF_RETURNS_ARG case, it is not done for the IPA-VRP return value case already so I don't think
there is anything to be done there.

Changes since v1:
* v2: Add an useless_type_conversion_p check as suggested by Jakub
      and add a testcase for that.
* v3: Fix the order of arguments to useless_type_conversion_p.

Bootstrapped and tested on x86_64-linux-gnu.

PR tree-optimization/67797

gcc/ChangeLog:

* tree-tailcall.cc (find_tail_calls): Add support for ERF_RETURNS_ARG.

gcc/testsuite/ChangeLog:

* gcc.dg/tree-ssa/tailcall-14.c: New test.
* gcc.dg/tree-ssa/tailcall-15.c: New test.

Signed-off-by: Andrew Pinski <quic_apinski@quicinc.com>

MAINTAINERS: Add my gcc.gnu.org username

ChangeLog:

* MAINTAINERS (kcy): Add gcc.gnu.org username.

gimple-verifier: Add check that comparison in GIMPLE_COND does not throw

While working on PR 119903, I noticed that there is code in replace_stmt_with_simplification
which makes sure that the comparison of a GIMPLE_COND does not throw
(non-call exceptions and trapping math) but the gimple verifier does not
verify this. So let's add it.

Bootstrapped and tested on x86_64-linux-gnu.

gcc/ChangeLog:

* tree-cfg.cc (verify_gimple_cond): Error out if the comparison
throws.

Signed-off-by: Andrew Pinski <quic_apinski@quicinc.com>

gimplefe: Round trip of rotates [PR119432]

This adds support for rotate left/right to the GIMPLE front-end
via __ROTATE_LEFT/__ROTATE_RIGHT oeprators.

PR c/119432
gcc/c/ChangeLog:

* gimple-parser.cc (gimple_binary_identifier_code): Add
__ROTATE_LEFT and __ROTATE_RIGHT.

gcc/ChangeLog:

* tree-pretty-print.cc (op_symbol_code): For LROTATE_EXPR,
output __ROTATE_LEFT for gimple.
For RROTATE_EXPR output __ROTATE_RIGHT for gimple.

gcc/testsuite/ChangeLog:

* gcc.dg/gimplefe-57.c: New test.

Signed-off-by: Andrew Pinski <quic_apinski@quicinc.com>

gimplefe: Simplify handling of identifier based binary operations

While looking into adding __ROTATE_LEFT and __ROTATE_RIGHT, I noticed
this code is just a bunch of if statments repeated. Instead we could just
use a simple lookup array to do the same thinga and it would be easier to
add to the array instead of duplicating the if sequence again.

Bootstrapped and tested on x86_64-linux-gnu.

gcc/c/ChangeLog:

* gimple-parser.cc (gimple_binary_identifier_code): New variable.
(c_parser_gimple_binary_expression): Use gimple_binary_identifier_code
instead of doing if statements on the strings.

Signed-off-by: Andrew Pinski <quic_apinski@quicinc.com>

Add testcase for bogus Warray-bounds warning dealing with __builtin_unreachable [PR100038]

After EVRP was switched to the ranger (r12-2305-g398572c1544d8b), we are better handling the case
where __builtin_unreachable comes after a loop. Instead of removing __builtin_unreachable and having
the loop become an infinite one; it is kept around longer and allows GCC to unroll the loop 2 times instead
of 3 times. When GCC unrolled the loop 3 times, GCC would produce a bogus Warray-bounds warning for the 3rd
iteration.
This adds the testcase to make sure we don't regress on this case. It is originally extracted from LLVM source
code too.

PR tree-optimization/100038

gcc/testsuite/ChangeLog:

* g++.dg/tree-ssa/pr100038.C: New test.

Signed-off-by: Andrew Pinski <quic_apinski@quicinc.com>

gccrs: Fix narrowing conversion warnings

Fixes PR#119641

gcc/rust/ChangeLog:

* checks/errors/borrowck/rust-bir-place.h
(IndexVec::size_type): Add.
(IndexVec::MAX_INDEX): Add.
(IndexVec::size): Change the return type to the type of the
internal value used by the index type.
(PlaceDB::lookup_or_add_variable): Use the return value from the
PlaceDB::add_place call.
* checks/errors/borrowck/rust-bir.h
(struct BasicBlockId): Move this definition before the
definition of the struct Function.

Signed-off-by: Owen Avery <powerboat9.gamer@gmail.com>

gccrs: Visit visibility in UseDeclaration

Default visitor should visit all it's children.

gcc/rust/ChangeLog:

* ast/rust-ast-visitor.cc (DefaultASTVisitor::visit): Visit visibility.

Signed-off-by: Pierre-Emmanuel Patry <pierre-emmanuel.patry@embecosm.com>

gccrs: Some assorted tweaks and bug fixes

gcc/rust/ChangeLog:

* ast/rust-ast-visitor.cc
(DefaultASTVisitor::visit): Visit the loop labels of
WhileLetLoopExpr instances before visiting their scrutinee
expressions.
* resolve/rust-early-name-resolver-2.0.cc
(Early::resolve_glob_import): Pass the glob import's path
directly to NameResolutionContext::resolve_path.
* resolve/rust-toplevel-name-resolver-2.0.cc
(TopLevel::visit): Remove unnecessary call to
Identifier::as_string.
(flatten_glob): Improve handling of cases where a glob use tree
has no path.

Signed-off-by: Owen Avery <powerboat9.gamer@gmail.com>

gccrs: Completely duplicate path node

Both nodes had the same id, this led to a resolution conflict.

gcc/rust/ChangeLog:

* expand/rust-derive-clone.cc (DeriveClone::clone_enum_struct): Clone
path to avoid using the same nodeid.

gcc/testsuite/ChangeLog:

* rust/compile/nr2/exclude: Remove now passing test from exclusion
list.

Signed-off-by: Pierre-Emmanuel Patry <pierre-emmanuel.patry@embecosm.com>

gccrs: Use specialized param visit function for params

This commit introduce a new public function to visit function parameters
in the default visitor. It allows visitors derived from DefaultVisitor
to override only a small part of the default visitor.

gcc/rust/ChangeLog:

* ast/rust-ast-visitor.cc (DefaultASTVisitor::visit_function_params):
Add specialized function to visit function parameters.
(DefaultASTVisitor::visit): Remove parameter visit and call specialized
function instead.
* ast/rust-ast-visitor.h: Add function prototye.
* resolve/rust-late-name-resolver-2.0.cc (Late::visit): Remove
function.
(Late::visit_function_params): Override specialized visit function.
* resolve/rust-late-name-resolver-2.0.h: Add overriden function
prototype.

Signed-off-by: Pierre-Emmanuel Patry <pierre-emmanuel.patry@embecosm.com>

gccrs: Remove passing test from exclusion list

gcc/testsuite/ChangeLog:

* rust/compile/nr2/exclude: Remove passing test from exclusion list.

Signed-off-by: Pierre-Emmanuel Patry <pierre-emmanuel.patry@embecosm.com>

gccrs: Use stacked context for nested bindings.

Binding context may be stacked when a new binding group is introduced
within a const expression.

gcc/rust/ChangeLog:

* resolve/rust-name-resolution-context.h: Use BindingLayer instead.
* resolve/rust-name-resolution-context.cc (BindingLayer::BindingLayer):
Add new constructor for binding layer.
(BindingLayer::bind_test): Add a function to test a binding constraint.
(BindingLayer::push): Push a new binding group.
(BindingLayer::and_binded): Add function to test and-binding
constraint.
(BindingLayer::or_binded): Add function to test or-binding constraints.
(BindingLayer::insert_ident): Insert a new identifier in the current
binding group.
(BindingLayer::merge): Merge current binding group with it's parent.
(BindingLayer::get_source): Get the source of the current binding
group.
* resolve/rust-late-name-resolver-2.0.cc: Use stacked context for
binding group.
* util/rust-stacked-contexts.h: Add mutable peek function.

Signed-off-by: Pierre-Emmanuel Patry <pierre-emmanuel.patry@embecosm.com>

gccrs: Add pattern bindings

gcc/rust/ChangeLog:

* resolve/rust-late-name-resolver-2.0.cc (Late::visit): Add binding
creation in visitor.
* resolve/rust-late-name-resolver-2.0.h: Add function prototypes.
* resolve/rust-name-resolution-context.h: Add binding context.

Signed-off-by: Pierre-Emmanuel Patry <pierre-emmanuel.patry@embecosm.com>

gccrs: Add binding context class

We need to differentiate bindings types, so the same binding cannot be
reused multiple time in a product binding.

gcc/rust/ChangeLog:

* resolve/rust-name-resolution-context.h (struct Binding): Add Binding
struct to differentiate Or and Product bindings in patterns.
(enum class): Add Binding kind.
(class BindingContext): Add binding context with Binding stack.

Signed-off-by: Pierre-Emmanuel Patry <pierre-emmanuel.patry@embecosm.com>

gccrs: Add hash function for Identifiers

gcc/rust/ChangeLog:

* ast/rust-ast.h: Add hash function.

Signed-off-by: Pierre-Emmanuel Patry <pierre-emmanuel.patry@embecosm.com>

gccrs: Add equality operator for identifiers

gcc/rust/ChangeLog:

* ast/rust-ast.h: Add equality operator.

Signed-off-by: Pierre-Emmanuel Patry <pierre-emmanuel.patry@embecosm.com>

gccrs: Add fn_once and Sized lang items to the test

gcc/testsuite/ChangeLog:

* rust/compile/multiple_bindings1.rs: Add missing lang items.

Signed-off-by: Pierre-Emmanuel Patry <pierre-emmanuel.patry@embecosm.com>

gccrs: Change expected error output to match nr2

Name resolution 2.0 message describes the context around the unresolved
items and should therefore be kept.

gcc/testsuite/ChangeLog:

* rust/compile/nr2/exclude: Remove test from exclusion list.
* rust/compile/use_1.rs: Change expected output and remove test from
nr1.

Signed-off-by: Pierre-Emmanuel Patry <pierre-emmanuel.patry@embecosm.com>

gccrs: Update generics9 expected error message

gcc/testsuite/ChangeLog:

* rust/compile/generics9.rs: Change expected error message.
* rust/compile/nr2/exclude: Remove test from exclusion list.

Signed-off-by: Pierre-Emmanuel Patry <pierre-emmanuel.patry@embecosm.com>

gccrs: Prevent forward declaration in type parameters

gcc/rust/ChangeLog:

* resolve/rust-default-resolver.cc (DefaultResolver::visit): Add visit
function for TypeParam.
* resolve/rust-default-resolver.h: Add function prototype.
* resolve/rust-forever-stack.h: Add function to check for forward
declaration ban.
* resolve/rust-late-name-resolver-2.0.cc (Late::visit): Check forward
declarations.

Signed-off-by: Pierre-Emmanuel Patry <pierre-emmanuel.patry@embecosm.com>

gccrs: Remove error state for GenericArg

gcc/rust/ChangeLog:

* ast/rust-ast-collector.cc (TokenCollector::visit): Remove error kind
and change function call.
* ast/rust-ast-visitor.cc (DefaultASTVisitor::visit): Change call name.
* ast/rust-path.cc (ConstGenericParam::as_string): Likewise.
* ast/rust-path.h: Remove error kind.
* hir/rust-ast-lower-type.cc (ASTLowerGenericParam::visit): Change call
name.
* parse/rust-parse-impl.h (Parser::parse_generic_param): Use optional
on parsing failure.
(Parser::parse_generic_arg): Likewise.
(Parser::parse_path_generic_args): Likewise.
* parse/rust-parse.h: Likewise.
* resolve/rust-ast-resolve-type.h: Change call name.

Signed-off-by: Pierre-Emmanuel Patry <pierre-emmanuel.patry@embecosm.com>

gccrs: Change error message to match expected test output

gcc/rust/ChangeLog:

* resolve/rust-late-name-resolver-2.0.cc (Late::visit): Change error
message.

gcc/testsuite/ChangeLog:

* rust/compile/nr2/exclude: Remove passing test from exclusion list.

Signed-off-by: Pierre-Emmanuel Patry <pierre-emmanuel.patry@embecosm.com>

gccrs: nr2.0: Improve error handling

gcc/rust/ChangeLog:

* resolve/rust-early-name-resolver-2.0.cc
(Early::build_import_mapping): Avoid outputting an "unresolved
import" error if other errors are outputted during resolution.
* resolve/rust-early-name-resolver-2.0.h
(Early::resolve_path_in_all_ns): Collect path resolution errors
while avoiding duplicate errors for resolutions in each
namespace.
* resolve/rust-forever-stack.h
(ForeverStack::resolve_path): Add parameter for collecting
errors.
(ForeverStack::find_starting_point): Likewise.
(ForeverStack::resolve_segments): Likewise.
* resolve/rust-forever-stack.hxx
(check_leading_kw_at_start): Likewise.
(ForeverStack::find_starting_point): Likewise.
(ForeverStack::resolve_segments): Likewise.
(ForeverStack::resolve_path): Likewise.
* resolve/rust-name-resolution-context.h
(NameResolutionContext::resolve_path): Add optional parameter
for collecting errors.

gcc/testsuite/ChangeLog:

* rust/compile/nr2/exclude: Remove entry.

Signed-off-by: Owen Avery <powerboat9.gamer@gmail.com>

gccrs: nr2.0: Fix derive-debug1.rs

gcc/testsuite/ChangeLog:

* rust/compile/derive-debug1.rs: Adjust a path.
* rust/compile/nr2/exclude: Remove derive-debug1.rs.

Signed-off-by: Owen Avery <powerboat9.gamer@gmail.com>

gccrs: prealloc the initilizer vector

There are two cases when initilizing an array, this is the
const context which means we need to build the array ctor,
which means using lots of memory, its super inefficient
because we are using a big wrapper over the GCC internals here
but preallocating the vectors here causes a:

  terminate called after throwing an instance of 'std::bad_alloc'

So this is a handy error condition to rely on for this senario.

Fixes Rust-GCC#3713
Fixes Rust-GCC#3727

gcc/rust/ChangeLog:

* backend/rust-compile-expr.cc (CompileExpr::array_copied_expr): prealloc the vector

Signed-off-by: Philip Herron <herron.philip@googlemail.com>

gccrs: Fix ICE with empty generic arguments

We have an assertion when accessing generic args if there are any which
is really useful so this adds the missing guards for the case where
they are specified but empty.

Fixes Rust-GCC#3649

gcc/rust/ChangeLog:

* ast/rust-ast-visitor.cc (DefaultASTVisitor::visit): add guard
* expand/rust-expand-visitor.cc (ExpandVisitor::visit): add guard

gcc/testsuite/ChangeLog:

* rust/compile/nr2/exclude: nr2 is missing error for this
* rust/compile/issue-3649.rs: New test.

Signed-off-by: Philip Herron <herron.philip@googlemail.com>

gccrs: Adjust segment start position errors

gcc/rust/ChangeLog:

* resolve/rust-ast-resolve-path.cc
(ResolvePath::resolve_path): Adjust error messages.
* resolve/rust-ast-resolve-type.cc
(ResolveRelativeTypePath::go): Likewise.
* resolve/rust-forever-stack.hxx
(check_leading_kw_at_start): Likewise.

gcc/testsuite/ChangeLog:

* rust/compile/issue-3568.rs: Adjust expected errors.
* rust/compile/name_resolution9.rs: Likewise.
* rust/compile/self-path2.rs: Likewise.

Signed-off-by: Owen Avery <powerboat9.gamer@gmail.com>

gccrs: Fix ICE when handling case of unknown field in HIR::FieldAccess

We were wrongly adding the assertion that this must not be an enum but
this is a pointless assertion we only care that there are variant in the
ADT and if the field exists in the first variant.

Fixes Rust-GCC#3581

gcc/rust/ChangeLog:

* typecheck/rust-hir-type-check-expr.cc (TypeCheckExpr::visit): fix bad assertion

gcc/testsuite/ChangeLog:

* rust/compile/nonexistent-field.rs: fix bad error message
* rust/compile/issue-3581-1.rs: New test.
* rust/compile/issue-3581-2.rs: New test.

Signed-off-by: Philip Herron <herron.philip@googlemail.com>

gccrs: Add test case to show ice is fixed

Fixes Rust-GCC#3652

gcc/testsuite/ChangeLog:

* rust/compile/nr2/exclude: nr2 does not error on the T it should require Self::T
* rust/compile/issue-3652.rs: New test.

Signed-off-by: Philip Herron <herron.philip@googlemail.com>

gccrs: nr2.0: Remove unnecessary copy of Node

gcc/rust/ChangeLog:

* resolve/rust-forever-stack.hxx
(ForeverStack::resolve_path): Pass instance of Node to lambda by
reference instead of by value.

Signed-off-by: Owen Avery <powerboat9.gamer@gmail.com>

gccrs: Fix ICE in struct expressions

The error handling here was done long ago when we didnt know how to do
any error handling very well. This removed bad fatal_errors and adds in
some nice rich_location error diagnostics instead.

Fixes Rust-GCC#3628

gcc/rust/ChangeLog:

* typecheck/rust-hir-type-check-struct-field.h: keep reference to parent expression
* typecheck/rust-hir-type-check-struct.cc (TypeCheckStructExpr::TypeCheckStructExpr):
update ctor
(TypeCheckStructExpr::resolve): remove bad rust_fatal_errors
(TypeCheckStructExpr::visit): cleanup errors

gcc/testsuite/ChangeLog:

* rust/compile/macros/mbe/macro-issue2983_2984.rs: cleanup error diagnotics
* rust/compile/struct_init1.rs: likewise
* rust/compile/issue-3628.rs: New test.

Signed-off-by: Philip Herron <herron.philip@googlemail.com>

gccrs: nr2.0: Handle StructPatternFieldIdent

gcc/rust/ChangeLog:

* resolve/rust-late-name-resolver-2.0.cc
(Late::visit): Handle StructPatternFieldIdent.
* resolve/rust-late-name-resolver-2.0.h
(Late::visit): Likewise.

gcc/testsuite/ChangeLog:

* rust/compile/nr2/exclude: Remove entry.

Signed-off-by: Owen Avery <powerboat9.gamer@gmail.com>

gccrs: Add test case to show ICE is fixed

Fixes Rust-GCC#3662

gcc/testsuite/ChangeLog:

* rust/compile/issue-3662.rs: New test.

Signed-off-by: Philip Herron <herron.philip@googlemail.com>

gccrs: Add test case to show issue is fixed

This was already fixed in: bb01719f0e1 but we require fn_once lang item
to be defined as we are working on libcore support still.

Fixes Rust-GCC#3711

gcc/testsuite/ChangeLog:

* rust/compile/issue-3711.rs: New test.

Signed-off-by: Philip Herron <herron.philip@googlemail.com>

gccrs: Add gimple test for black box intrinsic

gcc/testsuite/ChangeLog:

* rust/compile/black_box.rs: New test.

Signed-off-by: Pierre-Emmanuel Patry <pierre-emmanuel.patry@embecosm.com>

gccrs: Add execute test for black_box intrinsic

gcc/testsuite/ChangeLog:

* rust/execute/black_box.rs: New test.

Signed-off-by: Pierre-Emmanuel Patry <pierre-emmanuel.patry@embecosm.com>