PR93685 - ICE in gfc_constructor_append_expr, at fortran/constructor.c:135

Fix handling of F2018 enhancements to DATA statements that allows
initialization of pointer components to derived types, and adjust error
handling for the CHARACTER case.

gcc/fortran/ChangeLog:

* data.c (gfc_assign_data_value): Restrict use of
create_character_initializer to constant initializers.
* trans-expr.c (gfc_conv_initializer): Ensure that character
initializer is constant, otherwise fall through to get the same
error handling as for non-character cases.

gcc/testsuite/ChangeLog:

* gfortran.dg/pr93685_1.f90: New test.
* gfortran.dg/pr93685_2.f90: New test.

libstdc++: Convert three doxygen.nl links to https

libstdc++-v3:
2020-12-25  Gerald Pfeifer  <gerald@pfeifer.com>

* doc/xml/manual/documentation_hacking.xml: Convert three links
to doxygen.nl to https.
* doc/html/manual/documentation_hacking.html: Regenerate.

Daily bump.

C++ : Add the -stdlib= option.

This option allows the user to specifiy alternate C++ runtime libraries,
for example when a platform uses libc++ as the installed C++ runtime.

We introduce the command line option: -stdlib= which is the user-facing
mechanism to select the C++ runtime to be used when compiling and linking
code.  This is the same option spelling as that used by clang to allow the
use of libstdc++.

The availability (and thus function) of the option are a configure-time
choice using the configuration control:
--with-gxx-libcxx-include-dir=

Specification of the path for the libc++ headers, enables the -stdlib=
option (using the path as given), default values are set when the path
is unconfigured.

If --with-gxx-libcxx-include-dir is given together with --with-sysroot=,
then we test to see if the include path starts with the sysroot and, if so,
record the sysroot-relative component as the local path.  At runtime, we
prepend the sysroot that is actually active.

At link time, we use the C++ runtime in force and (if that is libc++) also
append the libc++abi ABI library. As for other cases, if a target sets the
name pointer for the ABI library to NULL the G++ driver will omit it from
the link line.

gcc/ChangeLog:

* configure.ac: Add gxx-libcxx-include-dir handled
in the same way as the regular cxx header directory.
* Makefile.in: Regenerated.
* config.in: Likewise.
* configure: Likewise.
* cppdefault.c: Pick up libc++ headers if the option
is enabled.
* cppdefault.h (struct default_include): Amend comments
to reflect the extended use of the cplusplus field.
* incpath.c (add_standard_paths): Allow for multiple
c++ header include path variants.
* doc/invoke.texi: Document the -stdlib= option.

gcc/c-family/ChangeLog:

* c.opt: Add -stdlib= option and enumerations for
libstdc++ and libc++.

gcc/cp/ChangeLog:

* g++spec.c (LIBCXX, LIBCXX_PROFILE, LIBCXX_STATIC): New.
(LIBCXXABI, LIBCXXABI_PROFILE, LIBCXXABI_STATIC): New.
(enum stdcxxlib_kind): New.
(lang_specific_driver): Allow selection amongst multiple
c++ runtime libraries.

RISC-V: Fix python3 compatibility for multilib-generator

The subprocess return string is raw bytes in python3, it must decode
before used as string, verifed with python2 and python3.

gcc/ChangeLog:

* config/riscv/multilib-generator (arch_canonicalize): Call
decode for the subprocess return value.

compiler: parenthesize channel type strings if necessary

Avoid the ambiguity between "chan <- (chan int)" and "chan (<- chan int)".
This parenthesizes the same way as the gc compiler.

Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/279961

Daily bump.

PR middle-end/98160 - ICE in warn_dealloc_offset on member placement new and delete

gcc/ChangeLog:

PR middle-end/98160
* builtins.c (warn_dealloc_offset): Avoid assuming calls are made
through declared functions and not pointers.

gcc/testsuite/ChangeLog:

PR middle-end/98160
* g++.dg/warn/pr98160.C: New test.

PR c++/98413 - ICE on placement new and member pointer

gcc/ChangeLog:

PR c++/98413
* builtins.c (get_offset_range): Avoid non-integers/-pointers.

gcc/testsuite/ChangeLog:

PR c++/98413
* g++.dg/warn/pr98413.C: New test.

Account for VxWorks headers in libstdc++ test on names

Undefine various macros unexpectedly defined by VxWorks headers.

for  libstdc++-v3/ChangeLog

* testsuite/17_intro/names.cc: Account for VxWorks headers.

mark some tests in gcc.target/arm as requiring "fpic" support

Require effective target fpic support in tests using pic flags.

for  gcc/testsuite/ChangeLog

* gcc.target/arm/data-rel-1.c: Require "fpic" support.
* gcc.target/arm/data-rel-2.c: Likewise.
* gcc.target/arm/data-rel-3.c: Ditto.
* gcc.target/arm/pr44788.c: Ditto.
* gcc.target/arm/pr52006.c: Ditto.
* gcc.target/arm/pr59858.c: Ditto.
* gcc.target/arm/tlscall.c: Ditto.
* gcc.target/arm/require-pic-register-loc.c: Ditto.
Adjust line numbers.

c++: EOF location [PR 96045]

Setting the EOF token location to be the start of a line just after
the ending newline is not most helpful.  While that location is
probably the right place to report preprocessing and lexing issues,
when parsing, the location just after the last token is better.  That
way we get to point at some actual text.  Setting the location from
the previous token has the advantage over just setting the location to
be the end of the final line, in that any ending comments do not get
considered, which I think is better.

PR c++/96045
gcc/cp/
* parser.c (cp_lexer_new_main): Adjust EOF token location.
gcc/testsuite/
* g++.dg/diagnostic/pr96045-1.C: New.
* g++.dg/diagnostic/pr96045-2.C: New.
* g++.dg/diagnostic/pr96045-3.C: New.
* c-c++-common/goacc/pr79428-1.c: Adjust EOF diagnostic location.
* c-c++-common/gomp/pr79428-2.c: Likewise
* c-c++-common/raw-string-6.c: Likewise
* g++.dg/cpp0x/decltype63.C: Likewise
* g++.dg/cpp0x/gen-attrs-64.C: Likewise
* g++.dg/cpp0x/pr68726.C: Likewise
* g++.dg/cpp0x/pr78341.C: Likewise
* g++.dg/cpp1y/pr65202.C: Likewise
* g++.dg/cpp1y/pr65340.C: Likewise
* g++.dg/cpp1y/pr68578.C: Likewise
* g++.dg/cpp1z/class-deduction44.C: Likewise
* g++.dg/diagnostic/unclosed-extern-c.C: Likewise
* g++.dg/diagnostic/unclosed-function.C: Likewise
* g++.dg/diagnostic/unclosed-namespace.C: Likewise
* g++.dg/diagnostic/unclosed-struct.C: Likewise
* g++.dg/ext/pr84598.C: Likewise
* g++.dg/other/switch4.C: Likewise
* g++.dg/parse/attr4.C: Likewise
* g++.dg/parse/cond4.C: Likewise
* g++.dg/parse/crash10.C: Likewise
* g++.dg/parse/crash18.C: Likewise
* g++.dg/parse/crash27.C: Likewise
* g++.dg/parse/crash34.C: Likewise
* g++.dg/parse/crash35.C: Likewise
* g++.dg/parse/crash52.C: Likewise
* g++.dg/parse/crash59.C: Likewise
* g++.dg/parse/crash61.C: Likewise
* g++.dg/parse/crash67.C: Likewise
* g++.dg/parse/error14.C: Likewise
* g++.dg/parse/error56.C: Likewise
* g++.dg/parse/invalid1.C: Likewise
* g++.dg/parse/parameter-declaration-1.C: Likewise
* g++.dg/parse/parser-pr28152-2.C: Likewise
* g++.dg/parse/parser-pr28152.C: Likewise
* g++.dg/parse/pr68722.C: Likewise
* g++.dg/parse/pr96258.C: Likewise
* g++.dg/pr46852.C: Likewise
* g++.dg/pr46868.C: Likewise
* g++.dg/template/crash115.C: Likewise
* g++.dg/template/crash43.C: Likewise
* g++.dg/template/crash90.C: Likewise
* g++.dg/template/error-recovery1.C: Likewise
* g++.dg/template/error57.C: Likewise
* g++.old-deja/g++.other/crash31.C: Likewise

c++: Fix initializing empty base from prvalue [PR97597]

unsafe_return_slot_p wasn't recognizing an empty base as
potentially-overlapping, which it definitely is.

The change to build_base_path is to make the virtual conversion also
recognized by is_empty_base_ref; unsafe_return_slot_p doesn't to handle
virtual conversions, but hypothetical future callers might.

gcc/cp/ChangeLog:

PR c++/97597
* class.c (is_empty_base_ref): New.
(build_base_path): Add NOP_EXPR after offset.
* cp-tree.h (is_empty_base_ref): Declare it.
* call.c (unsafe_return_slot_p): Call it.

gcc/testsuite/ChangeLog:

PR c++/97597
* g++.dg/init/empty3.C: New test.

compiler: nil-checked pointers and index expressions can trap

The compiler was treating indirections through pointers that had been
explicitly checked against nil and slice and string index expressions
as non-trapping memory references.  That is true for ordinary Go
programs, but it isn't true if the programs construct their own memory
addresses.  In particular it isn't true for the kinds of programs that
want to use runtime.SetPanicOnFault.

The effect of this will be slightly larger binaries, due to additional
exception information, and perhaps slightly less optimization.

Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/279960

c++: Improve testcase [PR98332]

gcc/testsuite/ChangeLog:

* g++.dg/cpp0x/constexpr-overflow3.C: Use INT_MAX.

c++: Fix up floating point complex handling in build_zero_init_1 [PR98353]

While the gimplifier patch I've just committed fixed an ICE, in some cases
like on the committed testcase cp folding doesn't happen after
build_zero_init_1 because it is called already during gimplification.

For the scalar types, if we want to use convert, the problem with complex floats
is that it returns a COMPLEX_EXPR with FLOAT_EXPR arguments which have
INTEGER_CST 0 as argument.  As fold isn't recursive, it doesn't do anything
in that case, we need to first fold those FLOAT_EXPRs to REAL_CST 0.0 and
only afterwards the COMPLEX_EXPR can be folded into COMPLEX_CST with 0.0
arguments.

This patch instead just uses build_zero_cst that creates the zero constant for
any scalar types (and more) directly, instead of going through multiple hops.

2020-12-23  Jakub Jelinek  <jakub@redhat.com>

PR c++/98353
* init.c (build_zero_init_1): Use build_zero_cst for SCALAR_TYPE_P
zero initializers.

c++: Fix constexpr array ICE [PR98332]

The element initializer was non-constant, so its CONSTRUCTOR element ended
up NULL, so unshare_constructor crashed trying to look at it.  This patch
fixes this in two places: First, by returning when we see a non-constant
initializer; second, by not crashing on NULL.

gcc/cp/ChangeLog:

PR c++/98332
* constexpr.c (unshare_constructor): Check for NULL.
(cxx_eval_vec_init_1): Always exit early if non-constant.

gcc/testsuite/ChangeLog:

PR c++/98332
* g++.dg/cpp0x/constexpr-overflow3.C: New test.

Darwin : Adjust handling of MACOSX_DEPLOYMENT_TARGET for macOS 11.

The shift to macOS version 11 also means that '11' without any
following '.x' is accepted as a valid version number.  This adjusts
the validation code to accept this and map it to 11.0.0 which
matches what the clang toolchain appears to do.

gcc/ChangeLog:

* config/darwin-driver.c (validate_macosx_version_min): Allow
MACOSX_DEPLOYMENT_TARGET=11.
(darwin_default_min_version): Adjust warning spelling to avoid
an apostrophe.

compiler: remove references to slicecopy and slicestringcopy

The only calls to the runtime functions were removed in CL 170005.
The slicestringcopy function will be removed in the Go 1.16beta1 release.

Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/279932

d: Force TYPE_MODE of classes and non-POD structs as BLKmode

Without this being forced, the optimizer could still make decisions that
require objects of the non-POD types to need a temporary, which would
result in an ICE during the expand to RTL passes.

gcc/d/ChangeLog:

PR d/98427
* types.cc (TypeVisitor::visit (TypeStruct *)): Set TYPE_MODE of all
non-trivial types as BLKmode.
(TypeVisitor::visit (TypeClass *)): Likewise.

gcc/testsuite/ChangeLog:

PR d/98427
* gdc.dg/pr98427.d: New test.

libcody:  Remove testsuite [PR 98318]

libcody's testsuite is not very portable, and is testing a host
library, so will do weird things on cross compiling, I think.  Let's
just drop it.

libcody/
* Makefile.in (LDFLAGS): Drop -L..
* tests/01-serialize/connect.cc: Delete.
* tests/01-serialize/decoder.cc: Delete.
* tests/01-serialize/encoder.cc: Delete.
* tests/02-comms/client-1.cc: Delete.
* tests/02-comms/pivot-1.cc: Delete.
* tests/02-comms/server-1.cc: Delete.
* tests/Makesub.in: Delete.
* tests/jouster: Delete.

i386: Fix __builtin_trunc with FE_DOWNWARD rounding direction [PR96793]

x86_expand_truncdf_32 expander uses x86_sse_copysign_to_positive, which
is unable to change the sign from - to +.  When FE_DOWNWARD rounding
direction is in effect, the expanded sequence that involves subtraction
can trigger x - x = -0.0 special rule.  x86_sse_copysign_to_positive
fails to change the sign of the intermediate value, assumed to always
be positive, back to positive.

The patch adds one extra fabs that strips the sign from the intermediate
value when flag_rounding_math is in effect.

2020-12-23  Uroš Bizjak  <ubizjak@gmail.com>

gcc/
PR target/96793
* config/i386/i386-expand.c (ix86_expand_truncdf_32):
Remove the sign of the intermediate value for flag_rounding_math.

gcc/testsuite/
PR target/96793
* gcc.target/i386/pr96793-1.c: New test.

c++tools:  Fix PIE  [PR 98324]

This adds --enable-default-pie support to c++tools, so that the sample
server is build -fPIE if requested.

PR bootstrap/98324
c++tools/
* Makefile.in: Add FLAGPIE.
* configure.ac: Add --enable-default-pie support.
* configure: Rebuilt.

Daily bump.

compiler: initialize saw_send_stmt locals

The C++ compiler wasn't warning because we take their address.

Fixes golang/go#43252

Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/279752

runtime: eliminate scase.kind field

This is the gofrontend version of https://golang.org/cl/245125.

Original CL description:

    Currently, we include a "kind" field on scase to distinguish the three
    kinds of cases in a select statement: sends, receives, and defaults.

    This commit removes by kind field by instead arranging for the
    compiler to always place sends before receives, and to provide their
    counts separately. It also passes an explicit "block bool" parameter
    to avoid needing to include a default case in the array.

    It's safe to shuffle cases like this because the runtime will
    randomize the order they're polled in anyway.

    For golang/go#40410.

This is being brought over to gofrontend as a step toward upgrading to
Go1.16beta1.

Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/279735

testsuite:  C++ module tests

This adds most of the modules tests.  I do not include the tests that
excercise system & C++ library header files.  Those will be later.

gcc/testsuite/
* g++.dg/modules/access-1_a.C: New.: New.
* g++.dg/modules/access-1_b.C: New.
* g++.dg/modules/access-1_c.C: New.
* g++.dg/modules/adhoc-1_a.C: New.
* g++.dg/modules/adhoc-1_b.C: New.
* g++.dg/modules/adl-1_a.C: New.
* g++.dg/modules/adl-1_b.C: New.
* g++.dg/modules/adl-1_c.C: New.
* g++.dg/modules/adl-2_a.C: New.
* g++.dg/modules/adl-2_b.C: New.
* g++.dg/modules/adl-2_c.C: New.
* g++.dg/modules/adl-3_a.C: New.
* g++.dg/modules/adl-3_b.C: New.
* g++.dg/modules/adl-3_c.C: New.
* g++.dg/modules/adl-4_a.C: New.
* g++.dg/modules/adl-4_b.C: New.
* g++.dg/modules/adl-5_a.c: New.
* g++.dg/modules/adl-5_b.C: New.
* g++.dg/modules/adl-5_c.C: New.
* g++.dg/modules/adl-5_d.C: New.
* g++.dg/modules/alias-1_a.H: New.
* g++.dg/modules/alias-1_b.C: New.
* g++.dg/modules/alias-1_c.C: New.
* g++.dg/modules/alias-1_d.C: New.
* g++.dg/modules/alias-1_e.C: New.
* g++.dg/modules/alias-1_f.C: New.
* g++.dg/modules/alias-2_a.H: New.
* g++.dg/modules/alias-2_b.C: New.
* g++.dg/modules/align-type-1_a.C: New.
* g++.dg/modules/align-type-1_b.C: New.
* g++.dg/modules/ambig-1_a.C: New.
* g++.dg/modules/ambig-1_b.C: New.
* g++.dg/modules/anon-1_a.C: New.
* g++.dg/modules/anon-1_b.C: New.
* g++.dg/modules/anon-1_c.C: New.
* g++.dg/modules/anon-2.h: New.
* g++.dg/modules/anon-2_a.H: New.
* g++.dg/modules/anon-2_b.C: New.
* g++.dg/modules/atom-decl-0_a.C: New.
* g++.dg/modules/atom-decl-0_b.C: New.
* g++.dg/modules/atom-decl-0_c.C: New.
* g++.dg/modules/atom-decl-2.C: New.
* g++.dg/modules/atom-decl-3.C: New.
* g++.dg/modules/atom-pragma-1.C: New.
* g++.dg/modules/atom-pragma-3.C: New.
* g++.dg/modules/atom-preamble-1.C: New.
* g++.dg/modules/atom-preamble-2_a.C: New.
* g++.dg/modules/atom-preamble-2_b.C: New.
* g++.dg/modules/atom-preamble-2_c.C: New.
* g++.dg/modules/atom-preamble-2_d.C: New.
* g++.dg/modules/atom-preamble-2_e.C: New.
* g++.dg/modules/atom-preamble-2_f.C: New.
* g++.dg/modules/atom-preamble-3.C: New.
* g++.dg/modules/atom-preamble-4.C: New.
* g++.dg/modules/auto-1.h: New.
* g++.dg/modules/auto-1_a.H: New.
* g++.dg/modules/auto-1_b.C: New.
* g++.dg/modules/auto-2.h: New.
* g++.dg/modules/auto-2_a.H: New.
* g++.dg/modules/auto-2_b.C: New.
* g++.dg/modules/bad-mapper-1.C: New.
* g++.dg/modules/bad-mapper-2.C: New.
* g++.dg/modules/bad-mapper-3.C: New.
* g++.dg/modules/ben-1.map: New.
* g++.dg/modules/ben-1_a.C: New.
* g++.dg/modules/ben-1_b.C: New.
* g++.dg/modules/bfield-1_a.C: New.
* g++.dg/modules/bfield-1_b.C: New.
* g++.dg/modules/bfield-2_a.C: New.
* g++.dg/modules/bfield-2_b.C: New.
* g++.dg/modules/bool-1.h: New.
* g++.dg/modules/bool-1_a.H: New.
* g++.dg/modules/bool-1_b.H: New.
* g++.dg/modules/bool-1_c.C: New.
* g++.dg/modules/bug-1_a.C: New.
* g++.dg/modules/bug-1_b.C: New.
* g++.dg/modules/builtin-1_a.C: New.
* g++.dg/modules/builtin-1_b.C: New.
* g++.dg/modules/builtin-2.C: New.
* g++.dg/modules/builtin-3_b.C: New.
* g++.dg/modules/builtin-4_a.H: New.
* g++.dg/modules/builtin-4_b.C: New.
* g++.dg/modules/builtin-5_a.H: New.
* g++.dg/modules/builtin-5_b.C: New.
* g++.dg/modules/builtin-6_a.H: New.
* g++.dg/modules/builtin-6_b.C: New.
* g++.dg/modules/builtin-7_a.H: New.
* g++.dg/modules/builtin-7_b.C: New.
* g++.dg/modules/by-name-1.C: New.
* g++.dg/modules/cexpr-1_a.C: New.
* g++.dg/modules/cexpr-1_b.C: New.
* g++.dg/modules/cexpr-2_a.C: New.
* g++.dg/modules/cexpr-2_b.C: New.
* g++.dg/modules/circ-1_a.C: New.
* g++.dg/modules/circ-1_b.C: New.
* g++.dg/modules/circ-1_c.C: New.
* g++.dg/modules/circ-1_d.C: New.
* g++.dg/modules/class-1_a.C: New.
* g++.dg/modules/class-1_b.C: New.
* g++.dg/modules/class-1_c.C: New.
* g++.dg/modules/class-2_a.C: New.
* g++.dg/modules/class-2_b.C: New.
* g++.dg/modules/class-3_a.C: New.
* g++.dg/modules/class-3_b.C: New.
* g++.dg/modules/class-3_c.C: New.
* g++.dg/modules/class-3_d.C: New.
* g++.dg/modules/class-4_a.C: New.
* g++.dg/modules/class-4_b.C: New.
* g++.dg/modules/class-5_a.C: New.
* g++.dg/modules/class-5_b.C: New.
* g++.dg/modules/class-5_c.C: New.
* g++.dg/modules/class-6_a.C: New.
* g++.dg/modules/class-6_b.C: New.
* g++.dg/modules/class-6_c.C: New.
* g++.dg/modules/class-7_a.C: New.
* g++.dg/modules/class-7_b.C: New.
* g++.dg/modules/class-7_c.C: New.
* g++.dg/modules/class-8_a.C: New.
* g++.dg/modules/class-8_b.C: New.
* g++.dg/modules/clone-1_a.C: New.
* g++.dg/modules/clone-1_b.C: New.
* g++.dg/modules/concept-1_a.C: New.
* g++.dg/modules/concept-1_b.C: New.
* g++.dg/modules/concept-2_a.C: New.
* g++.dg/modules/concept-2_b.C: New.
* g++.dg/modules/concept-3_a.C: New.
* g++.dg/modules/concept-3_b.C: New.
* g++.dg/modules/concept-4.H: New.
* g++.dg/modules/concept-5.h: New.
* g++.dg/modules/concept-5_a.H: New.
* g++.dg/modules/concept-5_b.C: New.
* g++.dg/modules/concept-6.h: New.
* g++.dg/modules/concept-6_a.H: New.
* g++.dg/modules/concept-6_b.C: New.
* g++.dg/modules/constrained-partial-1_a.C: New.
* g++.dg/modules/constrained-partial-1_b.C: New.
* g++.dg/modules/convop-1_a.C: New.
* g++.dg/modules/convop-1_b.C: New.
* g++.dg/modules/cpp-1.C: New.
* g++.dg/modules/cpp-2_a.H: New.
* g++.dg/modules/cpp-2_b.H: New.
* g++.dg/modules/cpp-2_c.C: New.
* g++.dg/modules/cpp-3.C: New.
* g++.dg/modules/cpp-4.C: New.
* g++.dg/modules/cpp-4.h: New.
* g++.dg/modules/cpp-5_a.H: New.
* g++.dg/modules/cpp-5_b.C: New.
* g++.dg/modules/cpp-5_c.C: New.
* g++.dg/modules/cpp-6_a.H: New.
* g++.dg/modules/cpp-6_b.H: New.
* g++.dg/modules/cpp-6_c.C: New.
* g++.dg/modules/debug-1_a.C: New.
* g++.dg/modules/debug-1_b.C: New.
* g++.dg/modules/decomp-1_a.C: New.
* g++.dg/modules/decomp-1_b.C: New.
* g++.dg/modules/deferred-1.h: New.
* g++.dg/modules/deferred-1_a.H: New.
* g++.dg/modules/deferred-1_b.C: New.
* g++.dg/modules/dep-1_a.C: New.
* g++.dg/modules/dep-1_b.C: New.
* g++.dg/modules/dep-2.C: New.
* g++.dg/modules/dep-3.C: New.
* g++.dg/modules/dir-only-1.C: New.
* g++.dg/modules/dir-only-2_a.H: New.
* g++.dg/modules/dir-only-2_b.C: New.
* g++.dg/modules/dir-only-3.C: New.
* g++.dg/modules/dir-only-4.C: New.
* g++.dg/modules/dir-recovery.C: New.
* g++.dg/modules/enum-1_a.C: New.
* g++.dg/modules/enum-1_b.C: New.
* g++.dg/modules/enum-2_a.C: New.
* g++.dg/modules/enum-2_b.C: New.
* g++.dg/modules/enum-3_a.C: New.
* g++.dg/modules/enum-3_b.C: New.
* g++.dg/modules/enum-4_a.C: New.
* g++.dg/modules/enum-4_b.C: New.
* g++.dg/modules/enum-5_a.H: New.
* g++.dg/modules/enum-5_b.C: New.
* g++.dg/modules/enum-6_a.H: New.
* g++.dg/modules/enum-6_b.C: New.
* g++.dg/modules/enum-7.C: New.
* g++.dg/modules/enum-8_a.H: New.
* g++.dg/modules/enum-8_b.H: New.
* g++.dg/modules/enum-8_c.C: New.
* g++.dg/modules/enum-8_d.C: New.
* g++.dg/modules/enum-bad-1_a.H: New.
* g++.dg/modules/enum-bad-1_b.C: New.
* g++.dg/modules/err-1_a.C: New.
* g++.dg/modules/err-1_b.C: New.
* g++.dg/modules/err-1_c.C: New.
* g++.dg/modules/err-1_d.C: New.
* g++.dg/modules/except-1.C: New.
* g++.dg/modules/except-2.h: New.
* g++.dg/modules/except-2_a.H: New.
* g++.dg/modules/except-2_b.C: New.
* g++.dg/modules/except-3.h: New.
* g++.dg/modules/except-3_a.H: New.
* g++.dg/modules/except-3_b.C: New.
* g++.dg/modules/exp-xlate-1_a.H: New.
* g++.dg/modules/exp-xlate-1_b.C: New.
* g++.dg/modules/export-1.C: New.
* g++.dg/modules/extern-tpl-1_a.H: New.
* g++.dg/modules/extern-tpl-1_b.C: New.
* g++.dg/modules/extern-tpl-1_c.C: New.
* g++.dg/modules/extern-tpl-2_a.H: New.
* g++.dg/modules/extern-tpl-2_b.H: New.
* g++.dg/modules/extern-tpl-2_c.C: New.
* g++.dg/modules/extern-tpl-2_d.C: New.
* g++.dg/modules/flag-1_a.C: New.
* g++.dg/modules/flag-1_b.C: New.
* g++.dg/modules/fn-inline-1_a.C: New.
* g++.dg/modules/fn-inline-1_b.C: New.
* g++.dg/modules/fn-inline-1_c.C: New.
* g++.dg/modules/freeze-1_a.C: New.
* g++.dg/modules/freeze-1_b.C: New.
* g++.dg/modules/freeze-1_c.C: New.
* g++.dg/modules/freeze-1_d.C: New.
* g++.dg/modules/friend-1_a.C: New.
* g++.dg/modules/friend-1_b.C: New.
* g++.dg/modules/friend-1_c.C: New.
* g++.dg/modules/friend-2_a.C: New.
* g++.dg/modules/friend-2_b.C: New.
* g++.dg/modules/friend-3.C: New.
* g++.dg/modules/friend-4_a.C: New.
* g++.dg/modules/friend-4_b.C: New.
* g++.dg/modules/friend-5_a.C: New.
* g++.dg/modules/friend-5_b.C: New.
* g++.dg/modules/gc-1_a.C: New.
* g++.dg/modules/gc-1_b.C: New.
* g++.dg/modules/gc-1_c.C: New.
* g++.dg/modules/gc-1_d.C: New.
* g++.dg/modules/gc-2.map: New.
* g++.dg/modules/gc-2_a.C: New.
* g++.dg/modules/global-1_a.C: New.
* g++.dg/modules/global-1_b.C: New.
* g++.dg/modules/gmf-1_a.C: New.
* g++.dg/modules/gmf-1_b.C: New.
* g++.dg/modules/gmf-2_a.H: New.
* g++.dg/modules/gmf-2_b.C: New.
* g++.dg/modules/gmf-2_c.C: New.
* g++.dg/modules/gmf-2_d.C: New.
* g++.dg/modules/gvar_a.C: New.
* g++.dg/modules/gvar_b.C: New.
* g++.dg/modules/hdr-1_a.H: New.
* g++.dg/modules/hdr-1_b.H: New.
* g++.dg/modules/hdr-1_c.C: New.
* g++.dg/modules/hdr-init-1_a.H: New.
* g++.dg/modules/hdr-init-1_b.H: New.
* g++.dg/modules/hdr-init-1_c.C: New.
* g++.dg/modules/horcrux-1_a.C: New.
* g++.dg/modules/horcrux-1_b.C: New.
* g++.dg/modules/ice-1.C: New.
* g++.dg/modules/imp-inline-1_a.C: New.
* g++.dg/modules/imp-inline-1_b.C: New.
* g++.dg/modules/imp-member-1_a.C: New.
* g++.dg/modules/imp-member-1_b.C: New.
* g++.dg/modules/imp-member-1_c.C: New.
* g++.dg/modules/imp-member-1_d.C: New.
* g++.dg/modules/imp-member-1_e.C: New.
* g++.dg/modules/imp-member-2_a.C: New.
* g++.dg/modules/imp-member-2_b.C: New.
* g++.dg/modules/imp-member-2_c.C: New.
* g++.dg/modules/imp-member-3.H: New.
* g++.dg/modules/import-1_a.C: New.
* g++.dg/modules/import-1_b.C: New.
* g++.dg/modules/import-1_c.C: New.
* g++.dg/modules/import-1_d.C: New.
* g++.dg/modules/import-1_e.C: New.
* g++.dg/modules/import-1_f.C: New.
* g++.dg/modules/import-1_g.C: New.
* g++.dg/modules/import-2.C: New.
* g++.dg/modules/inc-xlate-1.map: New.
* g++.dg/modules/inc-xlate-1_a.H: New.
* g++.dg/modules/inc-xlate-1_b.H: New.
* g++.dg/modules/inc-xlate-1_c.C: New.
* g++.dg/modules/inc-xlate-1_e.C: New.
* g++.dg/modules/indirect-1_a.C: New.
* g++.dg/modules/indirect-1_b.C: New.
* g++.dg/modules/indirect-1_c.C: New.
* g++.dg/modules/indirect-2_a.C: New.
* g++.dg/modules/indirect-2_b.C: New.
* g++.dg/modules/indirect-2_c.C: New.
* g++.dg/modules/indirect-3_a.C: New.
* g++.dg/modules/indirect-3_b.C: New.
* g++.dg/modules/indirect-3_c.C: New.
* g++.dg/modules/indirect-4_a.C: New.
* g++.dg/modules/indirect-4_b.C: New.
* g++.dg/modules/indirect-4_c.C: New.
* g++.dg/modules/inext-1.H: New.
* g++.dg/modules/inh-tmpl-ctor-1.h: New.
* g++.dg/modules/inh-tmpl-ctor-1_a.H: New.
* g++.dg/modules/inh-tmpl-ctor-1_b.C: New.
* g++.dg/modules/init-1_a.C: New.
* g++.dg/modules/init-1_b.C: New.
* g++.dg/modules/init-2_a.C: New.
* g++.dg/modules/init-2_b.C: New.
* g++.dg/modules/init-2_c.C: New.
* g++.dg/modules/inst-1_a.C: New.
* g++.dg/modules/inst-1_b.C: New.
* g++.dg/modules/inst-2_a.C: New.
* g++.dg/modules/inst-2_b.C: New.
* g++.dg/modules/inst-3_a.C: New.
* g++.dg/modules/inst-3_b.C: New.
* g++.dg/modules/inst-4_a.C: New.
* g++.dg/modules/inst-4_b.C: New.
* g++.dg/modules/inst-5_a.H: New.
* g++.dg/modules/inst-5_b.C: New.
* g++.dg/modules/internal-1.C: New.
* g++.dg/modules/internal-2_a.H: New.
* g++.dg/modules/internal-2_b.H: New.
* g++.dg/modules/internal-2_c.C: New.
* g++.dg/modules/isalnum.H: New.
* g++.dg/modules/keyword-1_a.C: New.
* g++.dg/modules/keyword-1_b.C: New.
* g++.dg/modules/lambda-1_a.C: New.
* g++.dg/modules/lambda-1_b.C: New.
* g++.dg/modules/lambda-2.h: New.
* g++.dg/modules/lambda-2_a.H: New.
* g++.dg/modules/lambda-2_b.C: New.
* g++.dg/modules/lambda-2_c.C: New.
* g++.dg/modules/lambda-3.h: New.
* g++.dg/modules/lambda-3_a.H: New.
* g++.dg/modules/lambda-3_b.C: New.
* g++.dg/modules/lambda-3_c.C: New.
* g++.dg/modules/lambda-4.h: New.
* g++.dg/modules/lambda-4_a.H: New.
* g++.dg/modules/lambda-4_b.C: New.
* g++.dg/modules/lang-1_a.H: New.
* g++.dg/modules/lang-1_b.C: New.
* g++.dg/modules/lang-1_c.C: New.
* g++.dg/modules/lang-2_a.C: New.
* g++.dg/modules/lang-2_b.C: New.
* g++.dg/modules/late-ret-1.H: New.
* g++.dg/modules/late-ret-2_a.H: New.
* g++.dg/modules/late-ret-2_b.H: New.
* g++.dg/modules/late-ret-2_c.C: New.
* g++.dg/modules/late-ret-3_a.H: New.
* g++.dg/modules/late-ret-3_b.H: New.
* g++.dg/modules/late-ret-3_c.C: New.
* g++.dg/modules/lazy-1_a.C: New.
* g++.dg/modules/lazy-1_b.C: New.
* g++.dg/modules/leg-merge-1_a.H: New.
* g++.dg/modules/leg-merge-1_b.H: New.
* g++.dg/modules/leg-merge-1_c.C: New.
* g++.dg/modules/leg-merge-1_d.C: New.
* g++.dg/modules/leg-merge-2_a.H: New.
* g++.dg/modules/leg-merge-2_b.H: New.
* g++.dg/modules/leg-merge-2_c.C: New.
* g++.dg/modules/leg-merge-3_a.H: New.
* g++.dg/modules/leg-merge-3_b.H: New.
* g++.dg/modules/leg-merge-3_c.C: New.
* g++.dg/modules/leg-merge-3_d.C: New.
* g++.dg/modules/leg-merge-4_a.H: New.
* g++.dg/modules/leg-merge-4_b.H: New.
* g++.dg/modules/leg-merge-4_c.C: New.
* g++.dg/modules/leg-merge-5_a.H: New.
* g++.dg/modules/leg-merge-5_b.H: New.
* g++.dg/modules/leg-merge-5_c.C: New.
* g++.dg/modules/leg-merge-6_a.H: New.
* g++.dg/modules/leg-merge-6_b.H: New.
* g++.dg/modules/leg-merge-6_c.C: New.
* g++.dg/modules/leg-merge-7_a.H: New.
* g++.dg/modules/leg-merge-7_b.H: New.
* g++.dg/modules/leg-merge-7_c.C: New.
* g++.dg/modules/leg-merge-8_a.H: New.
* g++.dg/modules/leg-merge-8_b.H: New.
* g++.dg/modules/leg-merge-8_c.C: New.
* g++.dg/modules/leg-merge-9_a.H: New.
* g++.dg/modules/leg-merge-9_b.H: New.
* g++.dg/modules/leg-merge-9_c.C: New.
* g++.dg/modules/legacy-1_a.H: New.
* g++.dg/modules/legacy-1_b.C: New.
* g++.dg/modules/legacy-1_c.C: New.
* g++.dg/modules/legacy-2.h: New.
* g++.dg/modules/legacy-2.map: New.
* g++.dg/modules/legacy-2_a.H: New.
* g++.dg/modules/legacy-2_b.H: New.
* g++.dg/modules/legacy-2_c.C: New.
* g++.dg/modules/legacy-2_d.C: New.
* g++.dg/modules/legacy-3.h: New.
* g++.dg/modules/legacy-3_a.H: New.
* g++.dg/modules/legacy-3_b.H: New.
* g++.dg/modules/legacy-3_c.H: New.
* g++.dg/modules/legacy-6.map: New.
* g++.dg/modules/legacy-6_a.H: New.
* g++.dg/modules/legacy-6_b.H: New.
* g++.dg/modules/legacy-6_c.C: New.
* g++.dg/modules/legacy-6_d.C: New.
* g++.dg/modules/legacy-6_e.C: New.
* g++.dg/modules/legacy-6_f.C: New.
* g++.dg/modules/legacy-7_a.H: New.
* g++.dg/modules/legacy-7_b.C: New.
* g++.dg/modules/legacy-8_a.H: New.
* g++.dg/modules/legacy-8_b.H: New.
* g++.dg/modules/legacy-8_c.C: New.
* g++.dg/modules/legacy-8_d.C: New.
* g++.dg/modules/legacy-8_e.C: New.
* g++.dg/modules/libfn-1_a.C: New.
* g++.dg/modules/libfn-1_b.C: New.
* g++.dg/modules/literals-1_a.C: New.
* g++.dg/modules/literals-1_b.C: New.
* g++.dg/modules/loc-1_a.C: New.
* g++.dg/modules/loc-1_b.C: New.
* g++.dg/modules/loc-1_c.C: New.
* g++.dg/modules/loc-2_a.C: New.
* g++.dg/modules/loc-2_b.C: New.
* g++.dg/modules/loc-2_c.C: New.
* g++.dg/modules/loc-2_d.C: New.
* g++.dg/modules/loc-2_e.C: New.
* g++.dg/modules/loc-2_f.C: New.
* g++.dg/modules/loc-wrapper-1.h: New.
* g++.dg/modules/loc-wrapper-1_a.H: New.
* g++.dg/modules/loc-wrapper-1_b.C: New.
* g++.dg/modules/local-1_a.C: New.
* g++.dg/modules/local-1_b.C: New.
* g++.dg/modules/local-extern-1.C: New.
* g++.dg/modules/local-extern-2.H: New.
* g++.dg/modules/local-struct-1_a.C: New.
* g++.dg/modules/local-struct-1_b.C: New.
* g++.dg/modules/macloc-1_a.C: New.
* g++.dg/modules/macloc-1_b.C: New.
* g++.dg/modules/macloc-1_c.C: New.
* g++.dg/modules/macloc-1_d.C: New.
* g++.dg/modules/macloc-2_a.H: New.
* g++.dg/modules/macloc-2_b.C: New.
* g++.dg/modules/macro-1_a.H: New.
* g++.dg/modules/macro-1_b.C: New.
* g++.dg/modules/macro-2_a.H: New.
* g++.dg/modules/macro-2_b.H: New.
* g++.dg/modules/macro-2_c.H: New.
* g++.dg/modules/macro-2_d.C: New.
* g++.dg/modules/macro-3_a.H: New.
* g++.dg/modules/macro-3_b.H: New.
* g++.dg/modules/macro-3_c.C: New.
* g++.dg/modules/macro-4_a.H: New.
* g++.dg/modules/macro-4_b.H: New.
* g++.dg/modules/macro-4_c.H: New.
* g++.dg/modules/macro-4_d.C: New.
* g++.dg/modules/macro-4_e.C: New.
* g++.dg/modules/macro-4_f.C: New.
* g++.dg/modules/macro-4_g.C: New.
* g++.dg/modules/macro-5_a.H: New.
* g++.dg/modules/macro-5_b.H: New.
* g++.dg/modules/macro-5_c.C: New.
* g++.dg/modules/macro-6_a.H: New.
* g++.dg/modules/macro-6_b.C: New.
* g++.dg/modules/macro-6_c.C: New.
* g++.dg/modules/macro-7_a.C: New.
* g++.dg/modules/macro-7_b.C: New.
* g++.dg/modules/macro-7_c.C: New.
* g++.dg/modules/map-1.map: New.
* g++.dg/modules/map-1_a.C: New.
* g++.dg/modules/map-1_b.C: New.
* g++.dg/modules/map-1_b.map: New.
* g++.dg/modules/map-2.C: New.
* g++.dg/modules/map-2.map: New.
* g++.dg/modules/member-def-1_a.C: New.
* g++.dg/modules/member-def-1_b.C: New.
* g++.dg/modules/member-def-1_c.C: New.
* g++.dg/modules/member-def-1_d.C: New.
* g++.dg/modules/member-def-2_a.C: New.
* g++.dg/modules/member-def-2_b.C: New.
* g++.dg/modules/member-def-2_c.C: New.
* g++.dg/modules/member-def-2_d.C: New.
* g++.dg/modules/memref-1_a.C: New.
* g++.dg/modules/memref-1_b.C: New.
* g++.dg/modules/merge-10.h: New.
* g++.dg/modules/merge-10_a.H: New.
* g++.dg/modules/merge-10_b.C: New.
* g++.dg/modules/merge-11.h: New.
* g++.dg/modules/merge-11_a.H: New.
* g++.dg/modules/merge-11_b.C: New.
* g++.dg/modules/merge-12.h: New.
* g++.dg/modules/merge-12_a.H: New.
* g++.dg/modules/merge-12_b.C: New.
* g++.dg/modules/merge-13.h: New.
* g++.dg/modules/merge-13_a.H: New.
* g++.dg/modules/merge-13_b.C: New.
* g++.dg/modules/merge-14.h: New.
* g++.dg/modules/merge-14_a.H: New.
* g++.dg/modules/merge-14_b.C: New.
* g++.dg/modules/merge-15.h: New.
* g++.dg/modules/merge-15_a.H: New.
* g++.dg/modules/merge-15_b.C: New.
* g++.dg/modules/merge-1_a.C: New.
* g++.dg/modules/merge-1_b.C: New.
* g++.dg/modules/merge-2_a.H: New.
* g++.dg/modules/merge-2_b.C: New.
* g++.dg/modules/merge-3_a.H: New.
* g++.dg/modules/merge-3_b.C: New.
* g++.dg/modules/merge-4.h: New.
* g++.dg/modules/merge-4_a.H: New.
* g++.dg/modules/merge-4_b.C: New.
* g++.dg/modules/merge-5.h: New.
* g++.dg/modules/merge-5_a.H: New.
* g++.dg/modules/merge-5_b.C: New.
* g++.dg/modules/merge-6.h: New.
* g++.dg/modules/merge-6_a.H: New.
* g++.dg/modules/merge-6_b.C: New.
* g++.dg/modules/merge-7.h: New.
* g++.dg/modules/merge-7_a.H: New.
* g++.dg/modules/merge-7_b.C: New.
* g++.dg/modules/merge-8.h: New.
* g++.dg/modules/merge-8_a.H: New.
* g++.dg/modules/merge-8_b.C: New.
* g++.dg/modules/merge-9.h: New.
* g++.dg/modules/merge-9_a.H: New.
* g++.dg/modules/merge-9_b.C: New.
* g++.dg/modules/mod-exp-1_a.C: New.
* g++.dg/modules/mod-exp-1_b.C: New.
* g++.dg/modules/mod-imp-1_a.C: New.
* g++.dg/modules/mod-imp-1_b.C: New.
* g++.dg/modules/mod-imp-1_c.C: New.
* g++.dg/modules/mod-imp-1_d.C: New.
* g++.dg/modules/mod-impl-1_a.C: New.
* g++.dg/modules/mod-impl-1_b.C: New.
* g++.dg/modules/mod-impl-1_c.C: New.
* g++.dg/modules/mod-impl-1_d.C: New.
* g++.dg/modules/mod-indirect-1_a.C: New.
* g++.dg/modules/mod-indirect-1_b.C: New.
* g++.dg/modules/mod-indirect-1_c.C: New.
* g++.dg/modules/mod-indirect-1_d.C: New.
* g++.dg/modules/mod-indirect-1_e.C: New.
* g++.dg/modules/mod-stamp-1_a.C: New.
* g++.dg/modules/mod-stamp-1_b.C: New.
* g++.dg/modules/mod-stamp-1_c.C: New.
* g++.dg/modules/mod-stamp-1_d.C: New.
* g++.dg/modules/mod-sym-1.C: New.
* g++.dg/modules/mod-sym-2.C: New.
* g++.dg/modules/mod-sym-3.C: New.
* g++.dg/modules/mod-tpl-1_a.C: New.
* g++.dg/modules/mod-tpl-1_b.C: New.
* g++.dg/modules/mod-tpl-2_a.C: New.
* g++.dg/modules/mod-tpl-2_b.C: New.
* g++.dg/modules/mutual-friend.ii: New.
* g++.dg/modules/namespace-1_a.C: New.
* g++.dg/modules/namespace-1_b.C: New.
* g++.dg/modules/namespace-1_c.C: New.
* g++.dg/modules/namespace-2_a.C: New.
* g++.dg/modules/namespace-2_b.C: New.
* g++.dg/modules/namespace-3_a.C: New.
* g++.dg/modules/namespace-3_b.C: New.
* g++.dg/modules/namespace-4_a.C: New.
* g++.dg/modules/namespace-4_b.C: New.
* g++.dg/modules/namespace-4_c.C: New.
* g++.dg/modules/nest-1_a.C: New.
* g++.dg/modules/nest-1_b.C: New.
* g++.dg/modules/nest-1_c.C: New.
* g++.dg/modules/nested-1_a.C: New.
* g++.dg/modules/nested-1_b.C: New.
* g++.dg/modules/nested-1_c.C: New.
* g++.dg/modules/nested-2_a.C: New.
* g++.dg/modules/nested-2_b.C: New.
* g++.dg/modules/nested-constr-1.h: New.
* g++.dg/modules/nested-constr-1_a.H: New.
* g++.dg/modules/nested-constr-1_b.C: New.
* g++.dg/modules/nested-constr-2_a.C: New.
* g++.dg/modules/nested-constr-2_b.C: New.
* g++.dg/modules/nested-constr-2_c.C: New.
* g++.dg/modules/nodes-1_a.C: New.
* g++.dg/modules/nodes-1_b.C: New.
* g++.dg/modules/noexcept-1.h: New.
* g++.dg/modules/noexcept-1_a.H: New.
* g++.dg/modules/noexcept-1_b.C: New.
* g++.dg/modules/ns-alias-1_a.C: New.
* g++.dg/modules/ns-alias-1_b.C: New.
* g++.dg/modules/ns-alias-1_c.C: New.
* g++.dg/modules/ns-dir-1_a.C: New.
* g++.dg/modules/ns-dir-1_b.C: New.
* g++.dg/modules/ns-dup-1_a.C: New.
* g++.dg/modules/ns-dup-1_b.C: New.
* g++.dg/modules/ns-imp-1_a.C: New.
* g++.dg/modules/ns-imp-1_b.C: New.
* g++.dg/modules/ns-imp-1_c.C: New.
* g++.dg/modules/ns-part-1_a.C: New.
* g++.dg/modules/ns-part-1_b.C: New.
* g++.dg/modules/ns-part-1_c.C: New.
* g++.dg/modules/nsdmi-1_a.C: New.
* g++.dg/modules/nsdmi-1_b.C: New.
* g++.dg/modules/nsdmi-2.C: New.
* g++.dg/modules/omp-1_a.C: New.
* g++.dg/modules/omp-1_b.C: New.
* g++.dg/modules/omp-1_c.C: New.
* g++.dg/modules/omp-2_a.C: New.
* g++.dg/modules/omp-2_b.C: New.
* g++.dg/modules/only-1.C: New.
* g++.dg/modules/only-2.C: New.
* g++.dg/modules/only-3.C: New.
* g++.dg/modules/operator-1_a.C: New.
* g++.dg/modules/operator-1_b.C: New.
* g++.dg/modules/p0713-1.C: New.
* g++.dg/modules/p0713-2.C: New.
* g++.dg/modules/p0713-3.C: New.
* g++.dg/modules/part-1_a.C: New.
* g++.dg/modules/part-1_b.C: New.
* g++.dg/modules/part-1_c.C: New.
* g++.dg/modules/part-2_a.C: New.
* g++.dg/modules/part-2_b.C: New.
* g++.dg/modules/part-2_c.C: New.
* g++.dg/modules/part-2_d.C: New.
* g++.dg/modules/part-2_e.C: New.
* g++.dg/modules/part-3_a.C: New.
* g++.dg/modules/part-3_b.C: New.
* g++.dg/modules/part-3_c.C: New.
* g++.dg/modules/part-3_d.C: New.
* g++.dg/modules/part-4_a.C: New.
* g++.dg/modules/part-4_b.C: New.
* g++.dg/modules/part-4_c.C: New.
* g++.dg/modules/part-6_a.C: New.
* g++.dg/modules/part-6_b.C: New.
* g++.dg/modules/part-6_c.C: New.
* g++.dg/modules/part-6_d.C: New.
* g++.dg/modules/part-6_e.C: New.
* g++.dg/modules/part-7_a.C: New.
* g++.dg/modules/part-7_b.C: New.
* g++.dg/modules/part-7_c.C: New.
* g++.dg/modules/part-hdr-1_a.H: New.
* g++.dg/modules/part-hdr-1_b.C: New.
* g++.dg/modules/part-hdr-1_c.C: New.
* g++.dg/modules/part-mac-1_a.H: New.
* g++.dg/modules/part-mac-1_b.C: New.
* g++.dg/modules/part-mac-1_c.C: New.
* g++.dg/modules/partial-1.h: New.
* g++.dg/modules/partial-1_a.H: New.
* g++.dg/modules/partial-1_b.C: New.
* g++.dg/modules/pmf-1.h: New.
* g++.dg/modules/pmf-1_a.H: New.
* g++.dg/modules/pmf-1_b.C: New.
* g++.dg/modules/pmf-2.h: New.
* g++.dg/modules/pmf-2_a.H: New.
* g++.dg/modules/pmf-2_b.C: New.
* g++.dg/modules/pmp-1_a.C: New.
* g++.dg/modules/pmp-1_b.C: New.
* g++.dg/modules/pmp-2.C: New.
* g++.dg/modules/pmp-3.C: New.
* g++.dg/modules/pragma-1_a.H: New.
* g++.dg/modules/pragma-1_b.C: New.
* g++.dg/modules/predef-1.C: New.
* g++.dg/modules/predef-1.h: New.
* g++.dg/modules/predef-2.h: New.
* g++.dg/modules/predef-2_a.C: New.
* g++.dg/modules/predef-2_b.C: New.
* g++.dg/modules/preproc-1.C: New.
* g++.dg/modules/preproc-2_a.H: New.
* g++.dg/modules/preproc-2_b.C: New.
* g++.dg/modules/printf-1_a.H: New.
* g++.dg/modules/printf-1_b.C: New.
* g++.dg/modules/reparent-1_a.C: New.
* g++.dg/modules/reparent-1_b.C: New.
* g++.dg/modules/reparent-1_c.C: New.
* g++.dg/modules/scc-1.C: New.
* g++.dg/modules/scc-2.C: New.
* g++.dg/modules/shadow-1_a.C: New.
* g++.dg/modules/shadow-1_b.C: New.
* g++.dg/modules/stat-tpl-1_a.H: New.
* g++.dg/modules/static-1_a.C: New.
* g++.dg/modules/static-1_b.C: New.
* g++.dg/modules/static-1_c.C: New.
* g++.dg/modules/std-1_a.C: New.
* g++.dg/modules/std-1_b.C: New.
* g++.dg/modules/stdns_a.C: New.
* g++.dg/modules/stdns_b.C: New.
* g++.dg/modules/sv-1.h: New.
* g++.dg/modules/sv-1_a.C: New.
* g++.dg/modules/sv-1_b.C: New.
* g++.dg/modules/sym-subst-1.C: New.
* g++.dg/modules/sym-subst-2_a.C: New.
* g++.dg/modules/sym-subst-2_b.C: New.
* g++.dg/modules/sym-subst-3_a.C: New.
* g++.dg/modules/sym-subst-3_b.C: New.
* g++.dg/modules/sym-subst-4.C: New.
* g++.dg/modules/sym-subst-5.C: New.
* g++.dg/modules/sym-subst-6.C: New.
* g++.dg/modules/sys/alias-2_a.H: New.
* g++.dg/modules/sys/inext-1.H: New.
* g++.dg/modules/tdef-1_a.C: New.
* g++.dg/modules/tdef-1_b.C: New.
* g++.dg/modules/tdef-2_a.C: New.
* g++.dg/modules/tdef-2_b.C: New.
* g++.dg/modules/tdef-2_c.C: New.
* g++.dg/modules/tdef-3_a.C: New.
* g++.dg/modules/tdef-3_b.C: New.
* g++.dg/modules/tdef-3_c.C: New.
* g++.dg/modules/tdef-4_a.C: New.
* g++.dg/modules/tdef-4_b.C: New.
* g++.dg/modules/tdef-4_c.C: New.
* g++.dg/modules/tdef-5_a.C: New.
* g++.dg/modules/tdef-5_b.C: New.
* g++.dg/modules/tdef-6_a.H: New.
* g++.dg/modules/tdef-6_b.C: New.
* g++.dg/modules/tdef-7.h: New.
* g++.dg/modules/tdef-7_a.H: New.
* g++.dg/modules/tdef-7_b.C: New.
* g++.dg/modules/tdef-8_a.C: New.
* g++.dg/modules/tdef-8_b.C: New.
* g++.dg/modules/tdef-inst-1.h: New.
* g++.dg/modules/tdef-inst-1_a.C: New.
* g++.dg/modules/tdef-inst-1_b.C: New.
* g++.dg/modules/thunk-1_a.C: New.
* g++.dg/modules/thunk-1_b.C: New.
* g++.dg/modules/tmpl-part-req-1.h: New.
* g++.dg/modules/tmpl-part-req-1_a.H: New.
* g++.dg/modules/tmpl-part-req-1_b.C: New.
* g++.dg/modules/tmpl-part-req-2.h: New.
* g++.dg/modules/tmpl-part-req-2_a.H: New.
* g++.dg/modules/tmpl-part-req-2_b.C: New.
* g++.dg/modules/token-1.C: New.
* g++.dg/modules/token-2_a.C: New.
* g++.dg/modules/token-2_b.C: New.
* g++.dg/modules/token-3.C: New.
* g++.dg/modules/token-4.C: New.
* g++.dg/modules/token-5.C: New.
* g++.dg/modules/tpl-alias-1.h: New.
* g++.dg/modules/tpl-alias-1_a.H: New.
* g++.dg/modules/tpl-alias-1_b.C: New.
* g++.dg/modules/tpl-ary-1.h: New.
* g++.dg/modules/tpl-ary-1_a.H: New.
* g++.dg/modules/tpl-ary-1_b.C: New.
* g++.dg/modules/tpl-extern-fn-1_a.H: New.
* g++.dg/modules/tpl-extern-fn-1_b.C: New.
* g++.dg/modules/tpl-extern-var-1_a.H: New.
* g++.dg/modules/tpl-extern-var-1_b.C: New.
* g++.dg/modules/tpl-friend-1_a.C: New.
* g++.dg/modules/tpl-friend-1_b.C: New.
* g++.dg/modules/tpl-friend-2_a.C: New.
* g++.dg/modules/tpl-friend-2_b.C: New.
* g++.dg/modules/tpl-friend-3_a.C: New.
* g++.dg/modules/tpl-friend-3_b.C: New.
* g++.dg/modules/tpl-friend-4_a.C: New.
* g++.dg/modules/tpl-friend-4_b.C: New.
* g++.dg/modules/tpl-friend-5_a.C: New.
* g++.dg/modules/tpl-friend-5_b.C: New.
* g++.dg/modules/tpl-friend-6_a.C: New.
* g++.dg/modules/tpl-friend-6_b.C: New.
* g++.dg/modules/tpl-friend-7_a.C: New.
* g++.dg/modules/tpl-friend-7_b.C: New.
* g++.dg/modules/tpl-friend-merge-1.cc: New.
* g++.dg/modules/tpl-friend-merge-1.h: New.
* g++.dg/modules/tpl-friend-merge-1_a.H: New.
* g++.dg/modules/tpl-friend-merge-1_b.H: New.
* g++.dg/modules/tpl-friend-merge-1_c.H: New.
* g++.dg/modules/tpl-friend-merge-1_d.C: New.
* g++.dg/modules/tpl-friend-merge-1_e.C: New.
* g++.dg/modules/tpl-friend-merge-1_f.C: New.
* g++.dg/modules/tpl-spec-1_a.C: New.
* g++.dg/modules/tpl-spec-1_b.C: New.
* g++.dg/modules/tpl-spec-2_a.C: New.
* g++.dg/modules/tpl-spec-2_b.C: New.
* g++.dg/modules/tpl-spec-2_c.C: New.
* g++.dg/modules/tpl-spec-2_d.C: New.
* g++.dg/modules/tpl-spec-3_a.C: New.
* g++.dg/modules/tpl-spec-3_b.C: New.
* g++.dg/modules/tpl-spec-4_a.C: New.
* g++.dg/modules/tpl-spec-4_b.C: New.
* g++.dg/modules/tpl-spec-5_a.C: New.
* g++.dg/modules/tpl-spec-5_b.C: New.
* g++.dg/modules/tpl-spec-6_a.C: New.
* g++.dg/modules/tpl-spec-6_b.C: New.
* g++.dg/modules/tpl-spec-7.C: New.
* g++.dg/modules/tpl-tpl-friend-1_a.C: New.
* g++.dg/modules/tpl-tpl-friend-1_b.C: New.
* g++.dg/modules/tpl-tpl-mem-1_a.C: New.
* g++.dg/modules/tpl-tpl-mem-1_b.C: New.
* g++.dg/modules/tpl-tpl-merge-1.h: New.
* g++.dg/modules/tpl-tpl-merge-1_a.H: New.
* g++.dg/modules/tpl-tpl-merge-1_b.C: New.
* g++.dg/modules/tpl-tpl-merge-2.h: New.
* g++.dg/modules/tpl-tpl-merge-2_a.H: New.
* g++.dg/modules/tpl-tpl-merge-2_b.C: New.
* g++.dg/modules/tpl-tpl-parm-1_a.H: New.
* g++.dg/modules/tpl-tpl-parm-1_b.C: New.
* g++.dg/modules/tpl-tpl-parm-2.h: New.
* g++.dg/modules/tpl-tpl-parm-2_a.H: New.
* g++.dg/modules/tpl-tpl-parm-2_b.C: New.
* g++.dg/modules/tplmem-1_a.C: New.
* g++.dg/modules/tplmem-1_b.C: New.
* g++.dg/modules/tplmem-3_a.C: New.
* g++.dg/modules/tplmem-3_b.C: New.
* g++.dg/modules/ttp-1_a.C: New.
* g++.dg/modules/ttp-1_b.C: New.
* g++.dg/modules/ttp-2_a.C: New.
* g++.dg/modules/ttp-2_b.C: New.
* g++.dg/modules/ttp-3_a.C: New.
* g++.dg/modules/ttp-3_b.C: New.
* g++.dg/modules/typename-1_a.C: New.
* g++.dg/modules/typename-1_b.C: New.
* g++.dg/modules/unnamed-1_a.C: New.
* g++.dg/modules/unnamed-1_b.C: New.
* g++.dg/modules/unnamed-2.C: New.
* g++.dg/modules/used-1_a.H: New.
* g++.dg/modules/used-1_b.H: New.
* g++.dg/modules/used-1_c.C: New.
* g++.dg/modules/using-1_a.C: New.
* g++.dg/modules/using-1_b.C: New.
* g++.dg/modules/using-1_c.C: New.
* g++.dg/modules/using-2_a.C: New.
* g++.dg/modules/using-2_b.C: New.
* g++.dg/modules/using-2_c.C: New.
* g++.dg/modules/using-3.C: New.
* g++.dg/modules/using-4_a.C: New.
* g++.dg/modules/using-4_b.C: New.
* g++.dg/modules/using-5_a.C: New.
* g++.dg/modules/using-5_b.C: New.
* g++.dg/modules/using-6_a.C: New.
* g++.dg/modules/using-6_b.C: New.
* g++.dg/modules/using-7.C: New.
* g++.dg/modules/using-8_a.C: New.
* g++.dg/modules/using-8_b.C: New.
* g++.dg/modules/using-enum-1_a.H: New.
* g++.dg/modules/using-enum-1_b.C: New.
* g++.dg/modules/var-1_a.C: New.
* g++.dg/modules/var-1_b.C: New.
* g++.dg/modules/var-tpl-1_a.C: New.
* g++.dg/modules/var-tpl-1_b.C: New.
* g++.dg/modules/var-tpl-concept-1.h: New.
* g++.dg/modules/var-tpl-concept-1_a.C: New.
* g++.dg/modules/var-tpl-concept-1_b.C: New.
* g++.dg/modules/virt-1_a.C: New.
* g++.dg/modules/virt-1_b.C: New.
* g++.dg/modules/virt-2_a.C: New.
* g++.dg/modules/virt-2_b.C: New.
* g++.dg/modules/virt-2_c.C: New.
* g++.dg/modules/vmort-1_a.C: New.
* g++.dg/modules/vmort-1_b.C: New.
* g++.dg/modules/vmort-2_a.C: New.
* g++.dg/modules/vmort-2_b.C: New.
* g++.dg/modules/vmort-2_c.C: New.
* g++.dg/modules/vtt-1_a.C: New.
* g++.dg/modules/vtt-1_b.C: New.
* g++.dg/modules/vtt-1_c.C: New.
* g++.dg/modules/vtt-2.h: New.
* g++.dg/modules/vtt-2_a.H: New.
* g++.dg/modules/vtt-2_b.C: New.

c++: Module ICE fix

A missing check for decl lang specific has made itself apparent.

gcc/cp/
* module.cc (has_definition): Check DECL_LANG_SPECIFIC.

runtime: add "success" field to sudog

This is the gofrontend version of https://golang.org/cl/245019.
Original CL description:

    The current wakeup protocol for channel communications is that the
    second goroutine sets gp.param to the sudog when a value is
    successfully communicated over the channel, and to nil when the wakeup
    is due to closing the channel.

    Setting nil to indicate channel closure works okay for chansend and
    chanrecv, because they're only communicating with one channel, so they
    know it must be the channel that was closed. However, it means
    selectgo has to re-poll all of the channels to figure out which one
    was closed.

    This commit adds a "success" field to sudog, and changes the wakeup
    protocol to always set gp.param to sg, and to use sg.success to
    indicate successful communication vs channel closure.

    While here, this also reorganizes the chansend code slightly so that
    the sudog is still released to the pool if the send blocks and then is
    awoken because the channel closed.

    For golang/go#40410

This is being brought over to gofrontend as a step toward upgrading to
Go1.16beta1, setting up for more compiler changes related to select handling.

Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/279734

runtime: omit nil-channel cases from selectgo's orders

This is the gofrontend version of https://golang.org/cl/245123.
Original CL description:

    Currently, selectgo does an initial pass over the cases array to look
    for entries with nil channels, so they can be easily recognized and
    skipped later on. But this still involves actually visiting the cases.

    This commit changes selectgo to omit cases with nil channels when
    constructing pollorder, so that they'll be skipped over entirely later
    on. It also checks for caseDefault up front, which will facilitate
    changing it to use a "block bool" parameter instead.

    Updates golang/go#40410

This is being brought over to gofrontend as a step toward upgrading to
Go1.16beta1, setting up for more compiler changes related to select handling.

Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/279733

runtime: remove scase.releasetime field

This is the gofrontend version of https://golang.org/cl/245122.
Original CL description:

    selectgo will report at most one block event, so there's no need to
    keep a releasetime for every select case. It suffices to simply track
    the releasetime of the case responsible for the wakeup.

    Updates golang/go#40410.

This is being brought over to gofrontend as a step toward upgrading to
Go1.16beta1.

Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/279732

c++: Handle array members in build_comparison_op [PR93480]

http://eel.is/c++draft/class.compare.default#6 says for the
expanded list of subobjects:
"In that list, any subobject of array type is recursively expanded
to the sequence of its elements, in the order of increasing subscript."
but build_comparison_op just tried to compare the whole arrays, which
failed and therefore the defaulted comparison was deleted.

The following patch instead compares the array elements, and
if info.defining, adds runtime loops around it so that it iterates
over increasing subscripts.

For flexible array members it punts, we don't know how large those will be,
for zero sized arrays it doesn't even try to compare the elements,
because if there are no elements, there is nothing to compare, and
for [1] arrays it will not emit a loop because it is enough to use
[0] array ref to cover everything.

2020-12-21  Jakub Jelinek  <jakub@redhat.com>

PR c++/93480
* method.c (common_comparison_type): If comps[i] is a TREE_LIST,
use its TREE_VALUE instead.
(build_comparison_op): Handle array members.

* g++.dg/cpp2a/spaceship-synth10.C: New test.
* g++.dg/cpp2a/spaceship-synth-neg5.C: New test.

arm&aarch64: subdivide the type attribute "alu_shfit_imm"

The type attribute "alu_shfit_imm" is subdivided into
"alu_shift_imm_lsl_1to4" and "alu_shift_imm_other", to accommodate
optimazations of some microarchitectures.

Here is the detailed discussion.
https://gcc.gnu.org/pipermail/gcc/2020-September/233594.html

gcc/
* config/arm/types.md (define_attr "autodetect_type"): New.
(define_attr "type"): Subdivide alu_shift_imm.
* config/arm/common.md: New file.
* config/aarch64/predicates.md:Include common.md.
* config/arm/predicates.md:Include common.md.
* config/aarch64/aarch64.md (*add_<shift>_<mode>): Set autodetect_type.
(*add_<shift>_si_uxtw): Likewise.
(*sub_<shift>_<mode>): Likewise.
(*sub_<shift>_si_uxtw): Likewise.
(*neg_<shift>_<mode>2): Likewise.
(*neg_<shift>_si2_uxtw): Likewise.
* config/arm/arm.md (*addsi3_carryin_shift): Likewise.
(add_not_shift_cin): Likewise.
(*subsi3_carryin_shift): Likewise.
(*subsi3_carryin_shift_alt): Likewise.
(*rsbsi3_carryin_shift): Likewise.
(*rsbsi3_carryin_shift_alt): Likewise.
(*arm_shiftsi3): Likewise.
(*<arith_shift_insn>_multsi): Likewise.
(*<arith_shift_insn>_shiftsi): Likewise.
(subsi3_carryin): Set new type.
(*if_arith_move): Set new type.
(*if_move_arith): Set new type.
(define_attr "core_cycles"): Use new type.
* config/arm/arm-fixed.md (arm_ssatsihi_shift): Set autodetect_type.
* config/arm/thumb2.md (*orsi_not_shiftsi_si): Likewise.
(*thumb2_shiftsi3_short): Set new type.
* config/aarch64/falkor.md (falkor_alu_1_xyz): Use new type.
* config/aarch64/saphira.md (saphira_alu_1_xyz): Likewise.
* config/aarch64/thunderx.md (thunderx_arith_shift): Likewise.
* config/aarch64/thunderx2t99.md (thunderx2t99_alu_shift): Likewise.
* config/aarch64/thunderx3t110.md (thunderx3t110_alu_shift): Likewise.
(thunderx3t110_alu_shift1): Likewise.
* config/aarch64/tsv110.md (tsv110_alu_shift): Likewise.
* config/arm/arm1020e.md (1020alu_shift_op): Likewise.
* config/arm/arm1026ejs.md (alu_shift_op): Likewise.
* config/arm/arm1136jfs.md (11_alu_shift_op): Likewise.
* config/arm/arm926ejs.md (9_alu_op): Likewise.
* config/arm/cortex-a15.md (cortex_a15_alu_shift): Likewise.
* config/arm/cortex-a17.md (cortex_a17_alu_shiftimm): Likewise.
* config/arm/cortex-a5.md (cortex_a5_alu_shift): Likewise.
* config/arm/cortex-a53.md (cortex_a53_alu_shift): Likewise.
* config/arm/cortex-a57.md (cortex_a57_alu_shift): Likewise.
* config/arm/cortex-a7.md (cortex_a7_alu_shift): Likewise.
* config/arm/cortex-a8.md (cortex_a8_alu_shift): Likewise.
* config/arm/cortex-a9.md (cortex_a9_dp_shift): Likewise.
* config/arm/cortex-m4.md (cortex_m4_alu): Likewise.
* config/arm/cortex-m7.md (cortex_m7_alu_shift): Likewise.
* config/arm/cortex-r4.md (cortex_r4_alu_shift): Likewise.
* config/arm/exynos-m1.md (exynos_m1_alu_shift): Likewise.
* config/arm/fa526.md (526_alu_shift_op): Likewise.
* config/arm/fa606te.md (606te_alu_op): Likewise.
* config/arm/fa626te.md (626te_alu_shift_op): Likewise.
* config/arm/fa726te.md (726te_alu_shift_op): Likewise.
* config/arm/fmp626.md (mp626_alu_shift_op): Likewise.
* config/arm/marvell-pj4.md (pj4_shift): Likewise.
(pj4_shift_conds): Likewise.
(pj4_alu_shift): Likewise.
(pj4_alu_shift_conds): Likewise.
* config/arm/xgene1.md (xgene1_alu): Likewise.
* config/arm/arm.c (xscale_sched_adjust_cost): Likewise.

i386: Fix __builtin_floor with FE_DOWNWARD rounding direction [PR96793]

x86_expand_floorceil expander uses x86_sse_copysign_to_positive, which
is unable to change the sign from - to +.  When FE_DOWNWARD rounding
direction is in effect, the expanded sequence that involves subtraction
can trigger x - x = -0.0 special rule.  x86_sse_copysign_to_positive
fails to change the sign of the intermediate value, assumed to always
be positive, back to positive.

The patch adds one extra fabs that strips the sign from the intermediate
value when flag_rounding_math is in effect.

2020-12-22  Uroš Bizjak  <ubizjak@gmail.com>

gcc/
PR target/96793
* config/i386/i386-expand.c (ix86_expand_floorceil):
Remove the sign of the intermediate value for flag_rounding_math.
(ix86_expand_floorceildf_32): Ditto.

gcc/testsuite/
PR target/96793
* gcc.target/i386/pr96793.c: New test.

c++: Fix build with clang

After the addition of libcody building with clang 10.0.1 would fail
as follows:

  In file included from .../libcody/cody.hh:24:
  In file included from /usr/include/c++/v1/memory:653:
  /usr/include/c++/v1/typeinfo:346:5: error: no member named 'fancy_abort'
  in namespace 'std::__1'; did you mean simply 'fancy_abort'?
    _VSTD::abort();
    ^~~~~~~

mapper-client.cc and mapper-resolver.cc have addressed this already
by getting <string> and <vector> included; do the same for module.cc.

gcc/cp:
2020-12-22  Gerald Pfeifer  <gerald@pfeifer.com>

* module.cc (INCLUDE_STRING): Define.
(INCLUDE_VECTOR): Ditto.

c++: Add some asserts in mangle.c

The added asserts are places I noticed that we aren't adding the "on" to
distinguish an operator name from an expression involving the operator.
Hopefully that's because operator names never get there; if not, crash bug
reports will let us fix the issue.

gcc/cp/ChangeLog:

* mangle.c (write_member_name): Add assert.
(write_expression): Add asserts.

c++: Fix demangling of <unresolved-name>

The ABI for unresolved scoped names on the RHS of . and -> used to be

  sr <type> <unqualified-id>

That changed years ago to something more complex, but G++ was never updated.
This change was particularly incompatible for simple qualified-ids like
A::x, which were previously mangled as sr1A1x, and now sr1AE1x.

This obviously makes life hard for demanglers, which can't know whether to
consume that E or not.  To work around this, we now try demangling with the
newer ABI, and if that fails and we saw an "sr", try again with the older
ABI.

libiberty/ChangeLog:

PR c++/67343
* cp-demangle.h (struct d_info): Add unresolved_name_state.
* cp-demangle.c (d_prefix): Add subst parm.
(d_nested_name): Pass it.
(d_unresolved_name): Split out from...
(d_expression_1): ...here.
(d_demangle_callback): Maybe retry with old sr mangling.
* testsuite/demangle-expected: Add test.

Fix Typo.

gcc/ChangeLog

* config/i386/i386.md (*one_cmpl<mode>2_1): Fix typo, change
alternative from 2 to 1 in attr isa.

-fdump-go-spec: don't crash if DECL_ORIGINAL_TYPE is NULL

* godump.c (go_output_typedef): If DECL_ORIGINAL_TYPE is NULL, use
TREE_TYPE.

Go frontend: ensure mpfr exponent range is large enough for Go

PR go/98402
* go-lang.c (go_langhook_init): Force MPFR exponent range to be
large enough to support Go constants.

Daily bump.

gimplify: Gimplify value in gimplify_init_ctor_eval_range [PR98353]

gimplify_init_ctor_eval_range wasn't gimplifying value, so if it wasn't
a gimple val, verification at the end of gimplification would ICE (or with
release checking some random pass later on would ICE or misbehave).

2020-12-21  Jakub Jelinek  <jakub@redhat.com>

PR c++/98353
* gimplify.c (gimplify_init_ctor_eval_range): Gimplify value before
storing it into cref.

* g++.dg/opt/pr98353.C: New test.

c++: Fix demangling of qualified-id after '.'

The demangler was assuming that only an unqualified-id could appear after
. or ->.

libiberty/ChangeLog:

* cp-demangle.c (d_expression_1): Recognize qualified-id
on RHS of dt/pt.
* testsuite/demangle-expected: Add test.

c++: Fix demangling of x.operator type

d_operator_name decides whether "cv" indicates a cast or a conversion
operator based on is_expression.  "on" specifies that we want the conversion
operator.

libiberty/ChangeLog:

* cp-demangle.c (d_unqualified_name): Clear is_expression.
* testsuite/demangle-expected: Add tests.

Darwin : Update the kernel version to macOS version mapping.

With the change to macOS 11 and Darwin20, the algorithm for mapping
kernel version to macOS version has changed.

We now have darwin 20.X.Y => macOS 11.(X > 0 ? X - 1 : 0).??.
It currently unclear if the Y will be mapped to macOS patch version
and, if so, whether it will be one-based or 0-based.
Likewise, it's unknown if Darwin 21 will map to macOS 12, so these
entries are unchanged for the present.

gcc/ChangeLog:

* config/darwin-driver.c (darwin_find_version_from_kernel):
Compute the minor OS version from the minor kernel version.

libstdc++: Disable floating-point std::to_chars on unsupported targets

This patch conditionally disables the floating-point std::to_chars
implementation on targets whose float and double aren't IEEE binary32
and binary64, until a proper fallback can be added for such targets.
This fixes a bootstrap failure on non-IEEE-754 FP targets such as
vax-netbsdelf.

The new preprocessor tests in c++config that detect the binary32 and
binary64 formats were copied from gcc/testsuite/gcc.dg/float-exact-1.c.

libstdc++-v3/ChangeLog:

* include/bits/c++config (_GLIBCXX_FLOAT_IS_IEEE_BINARY_32):
Define this macro.
(_GLIBCXX_DOUBLE_IS_IEEE_BINARY_64): Likewise.
* include/std/charconv (to_chars): Use these macros to
conditionally hide the overloads for floating-point types.
* src/c++17/floating_to_chars.cc: Use the macros to
conditionally disable this file.
(floating_type_traits<float>): Remove redundant static assert.
(floating_type_traits<double>): Likewise.
* testsuite/20_util/to_chars/double.cc: Run this test only on
ieee-floats effective targets.
* testsuite/20_util/to_chars/float.cc: Likewise.
* testsuite/20_util/to_chars/long_double.cc: Likewise.
* testsuite/lib/libstdc++.exp
(check_effective_target_ieee-floats): Define new proc for
detecting whether float and double have the IEEE binary32 and
binary64 formats.

c++: Windows rename [PR 98412]

Some system's rename(2) fails if the target already exists, so delete it
first.

gcc/cp/
* module.cc (create_dirs): Add logging.
(finish_module_processing): Unlink before rename.

bootstrap: std:stoul non-portable [PR 98412]

Fix some more system-specific issues.  Not everyone's C++11 is the same :(

PR bootstrap/98412
libcody/
* client.cc: Include cstdlib.
* server.cc: Include cstdlib.
gcc/cp/
* mapper-client.cc: INCLUDE_STRING, INCLUDE_VECTOR.
(module_client::open_module_client): Avoid std::stoul.
* mapper-resolver.cc: INCLUDE_STRING, INCLUDE_VECTOR.

libcody: Include <netinet/in.h> as needed

libcody as originally added to GCC fail to build on *-unknown-freebsd11.4
and presumably others:

  c++ -std=c++11 -g -include config.h -I ... -MMD -MP -MF netclient.d -c
    -o netclient.o
  .../libcody/netclient.cc:114:3: error: unknown type sockaddr_in6 addr;
                                                      ^~~~~~~~~~~~

sockaddr_in6 is declared in /usr/include/netinet6/in6.h,
which is included by /usr/include/netinet/in.h.

Indeed, per https://man7.org/linux/man-pages/man7/ipv6.7.html

    #include <sys/socket.h>
    #include <netinet/in.h>

should be included, and our own gcc/ada/gsocket.h also has

    #if !(defined (VMS) || defined (__MINGW32__))
    #include <sys/socket.h>
    #include <sys/un.h>
    #include <netinet/in.h>
    :
    #endif

libcody/ChangeLog:

2020-12-21  Gerald Pfeifer  <gerald@pfeifer.com>

* netserver.cc: Include <netinet/in.h>.

libcody: Add ranlib

Add RANLIB.

libcody/
* Makefile.in (RANLIB): New var.
* Makesub.in (libcody.a): Apply RANLIB.
* configure.ac: Call AC_PROG_RANLIB.
* configure: Rebuilt.

libcody: to_string is not always available [PR 98412]

to_string is not always available, so don't use it.

libcody/
* buffer.cc (MessageBuffer::AppendInteger): Workaround
to_string's non-ubiquity.

libcody: Add netinet.h

We explicitly need to inlude netinet.h, (despite what happened on my test systems)

libcody/
* netclient.cc: Add netinet.h.

c++tools: Fix exe suffix [PR 98409]

I had a thinko about variable case, and, coupled with Make's behaviour
of just consing up variables out of nothing, and linux not having an
executable extension, didn't notice.

PR other/98409
c++tools/
* Makefile.in: Fix exeext variable case.

gcc-changelog: new error for quoted utf8 filenames

contrib/ChangeLog:

* gcc-changelog/git_commit.py: Add new error for quoted
filenames.
* gcc-changelog/test_email.py: Test it.
* gcc-changelog/test_patches.txt: Test it.

gcc-changelog: add Unicode test-case

contrib/ChangeLog:

* gcc-changelog/test_email.py: New test.
* gcc-changelog/test_patches.txt: Likewise.

fold-const: Fix up a buffer overflow in native_encode_initializer [PR98407]

For flexible array members we need to incrementally clear just from
ptr + total_bytes up to new ptr + total_bytes, but memset has been called
with the length from ptr, so was missing - total_bytes.  Additionally,
in this code off is guaranteed to be -1 and thus o 0, so don't bother pretending
we could handle anything else, it would be more complicated than that.

2020-12-21  Jakub Jelinek  <jakub@redhat.com>

PR tree-optimization/98407
* fold-const.c (native_encode_initializer): When handling flexible
array members, fix up computation of length for memset.  Also remove
" - o" as o is always guaranteed to be 0 in this code path.

* gcc.c-torture/compile/pr98407.c: New test.

openmp: Fix up handling of addressable temporaries in simd lb, b and incr expressions [PR98383]

For simd, we have code to artificially add locally defined variables into
private clauses if they are addressable, so that omplower turns them into
"omp simd array" variables.  As the testcase shows, this is undesirable if
those temporaries only show in the lb, b or incr expressions and nowhere else,
if it is just used there, we really want normal scalar temporaries.

This patch implements that by making sure we don't set for those GOVD_LOCAL-ish
temporaries turned into GOVD_PRIVATE the GOVD_SEEN flag during gimplification
of the lb, b and incr expressions, which means that the private clause isn't
added for those.

2020-12-21  Jakub Jelinek  <jakub@redhat.com>

PR c++/98383
* gimplify.c (struct gimplify_omp_ctx): Add in_for_exprs flag.
(gimple_add_tmp_var): For addressable temporaries appearing in
simd lb, b or incr expressions, don't add a private clause unless
it is seen also outside of those expressions in the simd body.
(omp_notice_variable): Likewise.
(gimplify_omp_for): Set and reset in_for_exprs around gimplification
of lb, b or incr expressions.

* g++.dg/gomp/pr98383.C: New test.

libgo: adjust sysinfo scripts for changed -fdump-go-spec

The -fdump-go-spec flag to GCC recently changed to be more fastidious
about handling incomplete types. This caused some breakage in
mk[r]sysinfo.sh on Solaris. This commit adjusts for the new behavior.
Specifically:

  * Types that refer to _in6_addr may be hidden behind a typedef and can
    no longer be filtered out with `grep -v in6_addr`. Instead just
    rewrite the definition of _in6_addr to [16]byte wherever it appears.

  * timestruc_t is now (correctly) emitted as an alias for timespec, so this
    case is handled specially.

  * stdio.h is included in sysinfo.c to avoid emitting an incomplete
    definition of the FILE type.

  * Dummy definitions for _u?pad128_t are now emitted automatically,
    which conflict with the definitions installed by mk[r]sysinfo.sh.
    These definitions were actually dead code, so just remove them.

Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/278672

Daily bump.

gcc: xtensa: implement bswapsi2, bswapdi2 and helpers

2020-12-20  Max Filippov  <jcmvbkbc@gmail.com>
gcc/
* config/xtensa/xtensa.md (bswapsi2, bswapdi2): New patterns.

gcc/testsuite/
* gcc.target/xtensa/bswap.c: New test.

libgcc/
* config/xtensa/lib1funcs.S (__bswapsi2, __bswapdi2): New
functions.
* config/xtensa/t-xtensa (LIB1ASMFUNCS): Add _bswapsi2 and
_bswapdi2.

libstdc++: Fix indentation in <semaphore>

libstdc++-v3/ChangeLog:

* include/std/semaphore: Fix indentation.

doc: Fix a typo [PR98400]

Fix a typo in an option name in documentation.

2020-12-20  Jakub Jelinek  <jakub@redhat.com>

PR other/98400
* doc/invoke.texi (-mbackchain): Fix a typo - -mmo-backchain ->
-mno-backchain.

Set DECL_ARTIFICIAL on gfortran internal variables.

It seems we sometimes use DECL_ARTIFICIAL as choosing between
different code paths.  In order not to make -fdebug-aux-vars
do different things, set DECL_ARTIFICIAL on the variables to
avoid these different code paths (and the corresponding
regressions).

gcc/fortran/ChangeLog:

* trans.c (create_var_debug_raw): Set DECL_ARTIFICIAL
on variables.

Correct coarray indices for test case.

gcc/testsuite/ChangeLog:

* gfortran.dg/coarray/send_char_array_1.f90: Correct coarray
indces.

Daily bump.

expr: Fix up constant_byte_string bitfield handling [PR98366]

constant_byte_string now uses a convert_to_bytes function, which doesn't
handle bitfields at all (don't punt on them, just puts them into wrong bits
or bytes).  Furthermore, I don't see a reason why that function should exist
at all, it duplicates native_encode_initializer functionality.
Except that native_encode_initializer punted on flexible array members and 2
tests in the testsuite relied on constant_byte_string handling those.
So, this patch throws away convert_to_bytes, uses native_encode_initializer
instead, but teaches it to handle flexible array members (only in the
non-mask mode with off == -1 for now), furthermore, it adds various corner
case checks that the old implementation was missing (like that STRING_CSTs
use int as length and therefore we shouldn't try to build larger than that
strings, or that native_encode*/native_interpret* APIs require sane
host and target bytes (8-bit on both).

2020-12-19  Jakub Jelinek  <jakub@redhat.com>

PR middle-end/98366
* fold-const.c (native_encode_initializer): Don't try to
memset more than total_bytes with off == -1 even if len is large.
Handle flexible array member initializers if off == -1 and mask is
NULL.
* expr.c (convert_to_bytes): Remove.
(constant_byte_string): Use native_encode_initializer instead of
convert_to_bytes.  Remove extraneous semicolon.  Punt on various
corner-cases the APIs don't handle, like sizes > INT_MAX,
BITS_PER_UNIT != 8, CHAR_BIT != 8.

* gcc.c-torture/execute/pr98366.c: New test.

d: Fix ICE in in force_decl_die, at dwarf2out.c with -gdwarf-2 -gstrict-dwarf [PR98067]

Manifest constants in D are represented as CONST_DECLs, which can be
imported from one module to another.  However, when compiling on strict
dwarf2 targets such as *-*-darwin10, importing CONST_DECLs cannot be
represented in debug as D did not exist as an AT_language until dwarf3,
and the only available fallback being DW_LANG_C.  As CONST_DECLs are
treated as enumerators in C, and not outputted individually in
gen_decl_die, this causes an internal error in force_decl_die to occur.

To handle this, similar to other places in dwarf2out, if a CONST_DECL is
seen in dwarf2out_imported_module_or_decl_1, then we simply return early
if the language is not one of Ada, D, or Fortran.

gcc/ChangeLog:

PR d/98067
* dwarf2out.c (dwarf2out_imported_module_or_decl_1): Handle
  CONST_DECL only if is_fortran, is_ada, or is_dlang.

gcc/testsuite/ChangeLog:

PR d/98067
* gdc.dg/debug/debug.exp: New test.
* gdc.dg/debug/dwarf2/dwarf2.exp: New test.
* gdc.dg/debug/dwarf2/imports/pr98067.d: New test.
* gdc.dg/debug/dwarf2/langdw2.d: New test.
* gdc.dg/debug/dwarf2/langdw3.d: New test.
* gdc.dg/debug/dwarf2/pr98067.d: New test.
* gdc.dg/debug/trivial.d: New test.

bswap: Fix up a thinko with empty CONSTRUCTORs [PR98378]

The code I've added recently in find_bswap_or_nop for VECTOR CONSTRUCTORs
is missing punt on an important case - namely empty CONSTRUCTORs, because in that
case the loop will not initialize *n and the code after the loop will then
use the uninitialized structure.

2020-12-19  Jakub Jelinek  <jakub@redhat.com>

PR tree-optimization/98378
* gimple-ssa-store-merging.c (find_bswap_or_nop): Punt if CONSTRUCTOR
has no elements.

Daily bump.

compiler: check for floating-point exponent overflow

Adjust mksysinfo and mkrsysinfo to strip out floating-point max numbers,
as they can trigger this error.

Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/278476

Re: [r11-6232 Regression] FAIL: gcc.dg/pr97750.c

Adjust testcase to not actually check for an undefined warning since we
don't really care about it, or what line it is issued on.

gcc/testsuite/
* gcc.dg/pr97750.c: Remove check for warning.

openmp: Don't optimize shared to firstprivate on task with depend clause

The attached testcase is miscompiled, because we optimize shared clauses
to firstprivate when task body can't modify the variable even when the
task has depend clause.  That is wrong, because firstprivate means the
variable will be copied immediately when the task is created, while with
depend clause some other task might change it later before the dependencies
are satisfied and the task should observe the value only after the change.

2020-12-18  Jakub Jelinek  <jakub@redhat.com>

* gimplify.c (struct gimplify_omp_ctx): Add has_depend member.
(gimplify_scan_omp_clauses): Set it to true if OMP_CLAUSE_DEPEND
appears on OMP_TASK.
(gimplify_adjust_omp_clauses_1, gimplify_adjust_omp_clauses): Force
GOVD_WRITTEN on shared variables if task construct has depend clause.

* testsuite/libgomp.c/task-6.c: New test.

openmp: Fix g++.dg/gomp/declare-target-3.C testcase when offloading is disabled

2020-12-18  Kwok Cheung Yeung  <kcy@codesourcery.com>

gcc/testsuite/
* g++.dg/gomp/declare-target-3.C: Only check .offload_var_table
entries if offloading is enabled.

libcc1: Fix up libcc1 configure [PR98330]

Either we should write test x$darwin_dynamic_lookup = xyes
or we should make sure the variable is always defined to non-empty string.

2020-12-18  Jakub Jelinek  <jakub@redhat.com>

PR bootstrap/98380
* configure.ac: Set darwin_dynamic_lookup=no instead to empty
string.
* configure: Regenerated.

aarch64: SVE: ICE in expand_direct_optab_fn [PR98177]

Problem comes from using the wrong interface to get the index type for a
COND_REDUCTION. For fixed-length SVE we get a V2SI (a 64-bit Advanced
SIMD vector) instead of a VNx2SI (an SVE vector that stores SI elements
in DI containers).

Credits to Richard Sandiford for pointing out the issue's root cause.

Original PR snippet proposed to reproduce issue was only causing ICE for C++
compiler (see pr98177-1 test cases). I've slightly modified original
snippet in order to reproduce issue on both C and C++ compilers. These
are pr98177-2 test cases.

gcc/ChangeLog:

PR target/98177
* tree-vect-loop.c (vect_create_epilog_for_reduction): Use
get_same_sized_vectype to obtain index type.
(vectorizable_reduction): Likewise.

gcc/testsuite/ChangeLog:

PR target/98177
* g++.target/aarch64/sve/pr98177-1.C: New test.
* g++.target/aarch64/sve/pr98177-2.C: New test.
* gcc.target/aarch64/sve/pr98177-1.c: New test.
* gcc.target/aarch64/sve/pr98177-2.c: New test.

libstdc++: Fix mistake in PR98374 change [PR98377]

The #ifdef RADIXCHAR directive should be moved one line up so that it
also guards the outer if statement, or else when RADIXCHAR is not
defined the outer if statement will end up nonsensically guarding the
declaration of output_length_upper_bound a few lines below it.

libstdc++-v3/ChangeLog:

PR libstdc++/98377
* src/c++17/floating_to_chars.cc (__floating_to_chars_precision):
Fix mistake.

libstdc++: Fix build failure on AArch64 ILP32 [PR98370]

This should fix a build failure on AArch64 ILP32 due to int32_t mapping
to long int instead of int on this platform, which causes type deduction
to fail in the below call to std::max as reported in the PR.

libstdc++-v3/ChangeLog:

PR libstdc++/98370
* src/c++17/floating_to_chars.cc (__floating_to_chars_shortest):
Provide explicit template arguments to the call to std::max.

libstdc++: Fix build failure due to missing <langinfo.h> [PR98374]

This should fix a build failure on Windows which lacks <langinfo.h>,
from which we use nl_langinfo() to obtain the radix character of the
current locale.  (We can't use the more portable localeconv() from
<clocale> to obtain the radix character of the current locale here
because it's not thread-safe, unfortunately.)

This change means that on Windows and other such platforms, we'll just
always assume the radix character used by printf is '.' when formatting
a long double through it.

libstdc++-v3/ChangeLog:

PR libstdc++/98374
* src/c++17/floating_to_chars.cc: Guard include of <langinfo.h>
with __has_include.
(__floating_to_chars_precision) [!defined(RADIXCHAR)]: Don't
attempt to obtain the radix character of the current locale,
just assume it's '.'.

libstdc++: Check FE_TONEAREST is defined before using it

We need to test that FE_TONEAREST is defined before we may use it along
with fegetround/fesetround to adjust the floating-point rounding mode.
This fixes a build failure with older versions of newlib.

libstdc++-v3/ChangeLog:

* src/c++17/floating_from_chars.cc (from_chars_impl)
[!defined(FE_TONEAREST)]: Don't adjust the rounding mode.
* src/c++17/floating_to_chars.cc (__floating_to_chars_precision):
Likewise.

openmp: Implicitly add 'declare target' directives for dynamic initializers in C++

2020-12-18  Kwok Cheung Yeung  <kcy@codesourcery.com>

gcc/
* langhooks-def.h (lhd_get_decl_init): New.
(lhd_finish_decl_inits): New.
(LANG_HOOKS_GET_DECL_INIT): New.
(LANG_HOOKS_OMP_FINISH_DECL_INITS): New.
(LANG_HOOKS_DECLS): Add LANG_HOOKS_GET_DECL_INIT and
LANG_HOOKS_OMP_FINISH_DECL_INITS.
* langhooks.c (lhd_omp_get_decl_init): New.
(lhd_omp_finish_decl_inits): New.
* langhooks.h (struct lang_hooks_for_decls): Add omp_get_decl_init
and omp_finish_decl_inits.
* omp-offload.c (omp_discover_declare_target_var_r): Use
get_decl_init langhook in place of DECL_INITIAL.  Call
omp_finish_decl_inits langhook at end of function.

gcc/cp/
* cp-lang.c (cxx_get_decl_init): New.
(cxx_omp_finish_decl_inits): New.
(LANG_HOOKS_GET_DECL_INIT): New.
(LANG_HOOKS_OMP_FINISH_DECL_INITS): New.
* cp-tree.h (dynamic_initializers): New.
* decl.c (dynamic_initializers): New.
* decl2.c (c_parse_final_cleanups): Add initializer entries
from vars to dynamic_initializers.

gcc/testsuite/
* g++.dg/gomp/declare-target-3.C: New.

aarch64: Extend aarch64-autovec-preference==2 to 128-bit SVE

When compiling with -msve-vector-bits=128, aarch64_preferred_simd_mode
would pass the same vector width to aarch64_simd_container_mode for
both SVE and Advanced SIMD, and so Advanced SIMD would always “win”.
This patch instead makes it choose directly between SVE and Advanced
SIMD modes, so that aarch64-autovec-preference==2 and
aarch64-autovec-preference==4 work for this configuration.

(aarch64-autovec-preference shouldn't affect aarch64_simd_container_mode
because that would have an ABI impact for things like GNU vectors.)

gcc/
* config/aarch64/aarch64.c (aarch64_preferred_simd_mode): Use
aarch64_full_sve_mode and aarch64_vq_mode directly, instead of
going via aarch64_simd_container_mode.

Arm: MVE: Add missing complex mul iterators

Seems when I split the patch I forgot to include these into the rot iterator..
The uncommitted hunks were still in my local tree so didn't notice.

gcc/ChangeLog:

* config/arm/iterators.md (rot): Add UNSPEC_VCMUL, UNSPEC_VCMUL90,
UNSPEC_VCMUL180, UNSPEC_VCMUL270.

c++: Fix windows binary files [PR 98362]

Windows has unique and special needs for open(2).

gcc/cp/
* module.cc (O_CLOEXEC, O_BINARY): Add window's support.
(elf_in::defrost, module_state::do_import)
(finish_module_processing): Use O_BINARY.

As well as the PR this patch fixes problems in handling class objects

2020-12-18  Paul Thomas  <pault@gcc.gnu.org>

gcc/fortran
PR fortran/83118
PR fortran/96012
* resolve.c (resolve_ordinary_assign): Generate a vtable if
necessary for scalar non-polymorphic rhs's to unlimited lhs's.
* trans-array.c (get_class_info_from_ss): New function.
(gfc_trans_allocate_array_storage): Defer obtaining class
element type until all sources of class exprs are tried. Use
class API rather than TREE_OPERAND. Look for class expressions
in ss->info by calling get_class_info_from_ss. After, obtain
the element size for class descriptors. Where the element type
is unknown, cast the data as character(len=size) to overcome
unlimited polymorphic problems.
(gfc_conv_ss_descriptor): Do not fix class variable refs.
(build_class_array_ref, structure_alloc_comps): Replace code
replicating the new function gfc_resize_class_size_with_len.
(gfc_alloc_allocatable_for_assignment): Obtain element size
for lhs in cases of deferred characters and class enitities.
Move code for the element size of rhs to start of block. Clean
up extraction of class parameters throughout this function.
After the shape check test whether or not the lhs and rhs
element sizes are the same. Use earlier evaluation of
'cond_null'. Reallocation of lhs only to happen if size changes
or element size changes.
* trans-expr.c (gfc_resize_class_size_with_len): New function.
(gfc_get_class_from_expr): If a constant expression is
encountered, return NULL_TREE;
(trans_scalar_class_assign): New function.
(gfc_conv_procedure_call): Ensure the vtable is present for
passing a non-class actual to an unlimited formal.
(trans_class_vptr_len_assignment): For expressions of type
BT_CLASS, extract the class expression if necessary. Use a
statement block outside the loop body. Ensure that 'rhs' is
of the correct type. Obtain rhs vptr in all circumstances.
(gfc_trans_scalar_assign): Call trans_scalar_class_assign to
make maximum use of the vptr copy in place of assignment.
(trans_class_assignment): Actually do reallocation if needed.
(gfc_trans_assignment_1): Simplify some of the logic with
'realloc_flag'. Set 'vptr_copy' for all array assignments to
unlimited polymorphic lhs.
* trans.c (gfc_build_array_ref): Call gfc_resize_class_size_
with_len to correct span for unlimited polymorphic decls.
* trans.h : Add prototype for gfc_resize_class_size_with_len.

gcc/testsuite/
PR fortran/83118
PR fortran/96012
* gfortran.dg/dependency_60.f90: New test.
* gfortran.dg/class_allocate_25.f90: New test.
* gfortran.dg/class_assign_4.f90: New test.
* gfortran.dg/unlimited_polymorphic_32.f03: New test.

c++: Fix PCH ICE with __builtin_source_location [PR98343]

Seems the ggc_remove ppc_nx 3 operand member relies on the hash tables to
contain pointers in the first element, which is not the case for
source_location_table* hash table, which has location_t and unsigned as
first two members and pointer somewhere else.
I've tried to change:
   static void
   pch_nx (T &p, gt_pointer_operator op, void *cookie)
   {
-    op (&p, cookie);
+    extern void gt_pch_nx (T *, gt_pointer_operator, void *);
+    gt_pch_nx (&p, op, cookie);
   }
in hash-traits.h, but that failed miserably.
So, this patch instead overrides the two pch_nx overloads (only the second
one is needed, the former one is identical to the ggc_remove one) but I need
to override both.

2020-12-18  Jakub Jelinek  <jakub@redhat.com>

PR c++/98343
* cp-gimplify.c (source_location_table_entry_hash::pch_nx): Override
static member functions from ggc_remove.

* g++.dg/pch/pr98343.C: New test.
* g++.dg/pch/pr98343.Hs: New file.

Go testsuite: handle +build lines correctly

Update the Go testsuite driver to handle +build lines as is done in
the upstream repo, and update some tests to the upstream repo copy
using +build lines with "gc" and "!gccgo" as appropriate.

* go.test/go-test.exp (go-set-goos): New procedure.
(go-gc-match): New procedure.
(go-gc-tests): Call go-set-goos.  Use go-gc-match to handle +build
lines.  Look for +build lines beyond first line of file.

libstdc++: Import MSVC floating-point std::to_chars testcases

The testcases are imported almost verbatim, with the only change being
to the -double_nan and -float_nan testcases.  We expect these values to
be formatted as "-nan" instead of "-nan(ind)".

libstdc++-v3/ChangeLog:

* testsuite/20_util/to_chars/double.cc: New test, consisting of
testcases imported from the MSVC STL testsuite.
* testsuite/20_util/to_chars/float.cc: Likewise.

libstdc++: Add floating-point std::to_chars implementation

This implements the floating-point std::to_chars overloads for float,
double and long double.  We use the Ryu library to compute the shortest
round-trippable fixed and scientific forms for float, double and long
double.  We also use Ryu for performing explicit-precision fixed and
scientific formatting for float and double. For explicit-precision
formatting for long double we fall back to using printf.  Hexadecimal
formatting for float, double and long double is implemented from
scratch.

The supported long double binary formats are binary64, binary80 (x86
80-bit extended precision), binary128 and ibm128.

Much of the complexity of the implementation is in computing the exact
output length before handing it off to Ryu (which doesn't do bounds
checking).  In some cases it's hard to compute the output length
beforehand, so in these cases we instead compute an upper bound on the
output length and use a sufficiently-sized intermediate buffer only if
necessary.

Another source of complexity is in the general-with-precision formatting
mode, where we need to do zero-trimming of the string returned by Ryu,
and where we also take care to avoid having to format the number through
Ryu a second time when the general formatting mode resolves to fixed
(which we determine by doing a scientific formatting first and
inspecting the scientific exponent).  We avoid going through Ryu twice
by instead transforming the scientific form to the corresponding fixed
form via in-place string manipulation.

This implementation is non-conforming in a couple of ways:

1. For the shortest hexadecimal formatting, we currently follow the
   Microsoft implementation's decision to be consistent with the
   output of printf's '%a' specifier at the expense of sometimes not
   printing the shortest representation.  For example, the shortest hex
   form for the number 1.08p+0 is 2.1p-1, but we output the former
   instead of the latter, as does printf.

2. The Ryu routine generic_binary_to_decimal that we use for performing
   shortest formatting for large floating point types is implemented
   using the __int128 type, but some targets with a large long double
   type lack __int128 (e.g. i686), so we can't perform shortest
   formatting of long double on such targets through Ryu.  As a
   temporary stopgap this patch makes the long double to_chars overloads
   just dispatch to the double overloads on these targets, which means
   we lose precision in the output.  (We could potentially fix this by
   writing a specialized version of Ryu's generic_binary_to_decimal
   routine that uses uint64_t instead of __int128.)  [Though I wonder if
   there's a better way to work around the lack of __int128 on i686
   specifically?]

3. Our shortest formatting for __ibm128 doesn't guarantee the round-trip
   property if the difference between the high- and low-order exponent
   is large.  This is because we treat __ibm128 as if it has a
   contiguous 105-bit mantissa by merging the mantissas of the high-
   and low-order parts (using code extracted from glibc), so we
   potentially lose precision from the low-order part.  This seems to be
   consistent with how glibc printf formats __ibm128.

libstdc++-v3/ChangeLog:

* config/abi/pre/gnu.ver: Add new exports.
* include/std/charconv (to_chars): Declare the floating-point
overloads for float, double and long double.
* src/c++17/Makefile.am (sources): Add floating_to_chars.cc.
* src/c++17/Makefile.in: Regenerate.
* src/c++17/floating_to_chars.cc: New file.
(to_chars): Define for float, double and long double.
* testsuite/20_util/to_chars/long_double.cc: New test.

libstdc++: Apply modifications to our local copy of Ryu

This performs the following modifications to our local copy of Ryu in
order to make it more readily usable for our std::to_chars
implementation:

  * Remove all #includes
  * Remove copy_special_str routines
  * Adjust the exponent formatting to match printf
  * Remove some functions we're not going to use
  * Add an out-parameter to d2exp_buffered_n for the scientific exponent
  * Store the sign bit inside struct floating_decimal_[32|64]
  * Rename [df]2s_buffered_n and change their return type
  * Make generic_binary_to_decimal take the bit representation in parts

libstdc++-v3/ChangeLog:

* src/c++17/ryu/common.h, src/c++17/ryu/d2fixed.c,
src/c++17/ryu/d2fixed_full_table.h, src/c++17/ryu/d2s.c,
src/c++17/ryu/d2s_intrinsics.h, src/c++17/ryu/f2s.c,
src/c++17/ryu/f2s_intrinsics.h, src/c++17/ryu/generic_128.c:
Apply local modifications.

libstdc++: Import parts of the Ryu library

This imports the source files from the Ryu library that define
d2s_buffered_n, f2s_buffered_n, d2fixed_buffered_n, d2exp_buffered_n and
generic_binary_to_decimal, which we're going to use as the base of our
std::to_chars implementation.

libstdc++-v3/ChangeLog:

* src/c++17/ryu/MERGE: New file.
* src/c++17/ryu/common.h, src/c++17/ryu/d2fixed.c,
src/c++17/ryu/d2fixed_full_table.h, src/c++17/ryu/d2s.c,
src/c++17/ryu/d2s_full_table.h, src/c++17/ryu/d2s_intrinsics.h,
src/c++17/ryu/digit_table.h, src/c++17/ryu/f2s.c,
src/c++17/ryu/f2s_intrinsics.h, src/c++17/ryu/generic_128.c,
src/c++17/ryu/generic_128.h, src/c++17/ryu/ryu_generic_128.h:
Import these files from the Ryu library.

c++: More precise tracking of potentially unstable satisfaction

This makes tracking of potentially unstable satisfaction results more
precise by recording the specific types for which completion failed
during satisfaction.  We now recompute a satisfaction result only if one
of these types has been completed since the last time we computed the
satisfaction result.  Thus the number of times that we recompute a
satisfaction result is now bounded by the number of such incomplete
types, rather than being effectively unbounded.  This allows us to
remove the invalid assumption in note_ftc_for_satisfaction that was
added to avoid a compile time performance regression in cmcstl2 due to
repeated recomputation of a satisfaction result that depended on
completion of a permanently incomplete class template specialization.

In order to continue to detect the instability in concepts-complete3.C,
we also need to explicitly keep track of return type deduction failure
alongside type completion failure.  So this patch also adds a call to
note_ftc_for_satisfaction in require_deduced_type.

gcc/cp/ChangeLog:

* constraint.cc (satisfying_constraint): Move up definition
and give it bool type.
(failed_type_completion_count): Replace with ...
(failed_type_completions): ... this.
(note_failed_type_completion_for_satisfaction): Append the
supplied argument to failed_type_completions.
(some_type_complete_p): Define.
(sat_entry::maybe_unstable): Replace with ...
(sat_entry::ftc_begin, sat_entry::ftc_end): ... these.
(satisfaction_cache::ftc_count): Replace with ...
(satisfaction_cache::ftc_begin): ... this.
(satisfaction_cache::satisfaction_cache): Adjust accordingly.
(satisfaction_cache::get): Adjust accordingly, using
some_type_complete_p.
(satisfaction_cache::save): Adjust accordingly.
(satisfying_constraint_p): Remove unused function.
(satisfy_constraint): Set satisfying_constraint.
(satisfy_declaration_constraints): Likewise.
* decl.c (require_deduced_type): Call
note_failed_type_completion_for_satisfaction.

c++: Diagnose self-recursive satisfaction

This patch further extends the satisfaction_cache class to diagnose
self-recursive satisfaction.

gcc/cp/ChangeLog:

* constraint.cc (sat_entry::evaluating): New member.
(satisfaction_cache::get): If entry->evaluating, diagnose
self-recursive satisfaction.  Otherwise, set entry->evaluating
if we're not reusing a cached satisfaction result.
(satisfaction_cache::save): Clear entry->evaluating.
(satisfy_atom): Set up diagnosing_failed_constraint before the
first call to get().

gcc/testsuite/ChangeLog:

PR c++/96840
* g++.dg/cpp2a/concepts-pr88395.C: Adjust to expect the
self-recursive satisfaction to get directly diagnosed.
* g++.dg/cpp2a/concepts-recursive-sat2.C: Likewise.
* g++.dg/cpp2a/concepts-recursive-sat4.C: New test.

c++: Diagnose unstable satisfaction

This implements lightweight heuristical detection and diagnosing of
satisfaction whose result changes at different points in the program,
which renders the program ill-formed NDR as of P2104.  We've recently
started to more aggressively cache satisfaction results, and so the goal
with this patch is to make this caching behavior more transparent to
the user.

A satisfaction result is flagged as "potentially unstable" (at the atom
granularity) if during its computation, some type completion failure
occurs.  This is detected by making complete_type_or_maybe_complain
increment a counter upon failure and comparing the value of the counter
before and after satisfaction.  (We don't instrument complete_type
directly because it's used "opportunistically" in many spots where type
completion failure doesn't necessary lead to substitution failure.)

Such flagged satisfaction results are always recomputed from scratch,
even when performing satisfaction quietly.  When saving a satisfaction
result, we now compare the computed result with the cached result, and
if they differ, proceed with diagnosing the instability.

Most of the implementation is confined to the satisfaction_cache class,
which has been completely rewritten.

gcc/cp/ChangeLog:

* constraint.cc (failed_type_completion_count): New.
(note_failed_type_completion_for_satisfaction): New.
(sat_entry::constr): Rename to ...
(sat_entry::atom): ... this.
(sat_entry::location): New member.
(sat_entry::maybe_unstable): New member.
(sat_entry::diagnose_instability): New member.
(struct sat_hasher): Adjust after the above renaming.
(get_satisfaction, save_satisfaction): Remove.
(satisfaction_cache): Rewrite completely.
(satisfy_atom): When instantiation of the parameter mapping
fails, set diagnose_instability.  Propagate location from
inst_cache.entry to cache.entry if the secondary lookup
succeeded.
(satisfy_declaration_constraints): When
failed_type_completion_count differs before and after
satisfaction, then don't cache the satisfaction result.
* cp-tree.h (note_failed_type_completion_for_satisfaction):
Declare.
* pt.c (tsubst) <case TYPENAME_TYPE>: Use
complete_type_or_maybe_complain instead of open-coding it.
* typeck.c (complete_type_or_maybe_complain): Call
note_failed_type_completion_for_satisfaction when type
completion fails.

gcc/testsuite/ChangeLog:

* g++.dg/cpp2a/concepts-complete1.C: New test.
* g++.dg/cpp2a/concepts-complete2.C: New test.
* g++.dg/cpp2a/concepts-complete3.C: New test.

Daily bump.

arm: Add support for Cortex-A78C

This patch adds support for -mcpu=cortex-a78c command line option.
For more information about this processor, see [0]:

[0] https://developer.arm.com/ip-products/processors/cortex-a/cortex-a78c

gcc/ChangeLog:

* config/arm/arm-cpus.in: Add Cortex-A78C core.
* config/arm/arm-tables.opt: Regenerate.
* config/arm/arm-tune.md: Regenerate.
* doc/invoke.texi: Update docs.

rtl-ssa: Fix reg_raw_mode thinko [PR98347]

I'd used reg_raw_mode[regno] for general registers, even though
the array is only valid for hard registers.  This patch uses
regno_reg_rtx instead.

gcc/
PR rtl-optimization/98347
* rtl-ssa/access-utils.h (full_register): Use regno_reg_rtx
instead of reg_raw_mode.

Update default_estimated_poly_value prototype in targhooks.h

commit 64432b680eab0bddbe9a4ad4798457cf6a14ad60
Author: Kyrylo Tkachov <kyrylo.tkachov@arm.com>
Date:   Thu Dec 17 18:02:37 2020 +0000

    vect, aarch64: Extend SVE vs Advanced SIMD costing decisions in vect_better_loop_vinfo_p

changed default_estimated_poly_value to

HOST_WIDE_INT
default_estimated_poly_value (poly_int64 x, poly_value_estimate_kind)
{
  return x.coeffs[0];
}

Update default_estimated_poly_value prototype in targhooks.h to match it.

* targhooks.h (default_estimated_poly_value): Updated.

doc: Standard library header units

It seems users are confused by the lack of standard library header
units.

gcc/
* doc/invoke.texi (C++ Modules): Document lack of std
library header units.

vect, aarch64: Extend SVE vs Advanced SIMD costing decisions in vect_better_loop_vinfo_p

While experimenting with some backend costs for Advanced SIMD and SVE I
hit many cases where GCC would pick SVE for VLA auto-vectorisation even when
the backend very clearly presented cheaper costs for Advanced SIMD.
For a simple float addition loop the SVE costs were:

vec.c:9:21: note:  Cost model analysis:
  Vector inside of loop cost: 28
  Vector prologue cost: 2
  Vector epilogue cost: 0
  Scalar iteration cost: 10
  Scalar outside cost: 0
  Vector outside cost: 2
  prologue iterations: 0
  epilogue iterations: 0
  Minimum number of vector iterations: 1
  Calculated minimum iters for profitability: 4

and for Advanced SIMD (Neon) they're:

vec.c:9:21: note:  Cost model analysis:
  Vector inside of loop cost: 11
  Vector prologue cost: 0
  Vector epilogue cost: 0
  Scalar iteration cost: 10
  Scalar outside cost: 0
  Vector outside cost: 0
  prologue iterations: 0
  epilogue iterations: 0
  Calculated minimum iters for profitability: 0
vec.c:9:21: note:    Runtime profitability threshold = 4

yet the SVE one was always picked. With guidance from Richard this seems
to be due to the vinfo comparisons in vect_better_loop_vinfo_p, in
particular the part with the big comment explaining the
estimated_rel_new * 2 <= estimated_rel_old heuristic.

This patch extends the comparisons by introducing a three-way estimate
kind for poly_int values that the backend can distinguish.
This allows vect_better_loop_vinfo_p to ask for minimum, maximum and
likely estimates and pick Advanced SIMD overs SVE when it is clearly cheaper.

gcc/
* target.h (enum poly_value_estimate_kind): Define.
(estimated_poly_value): Take an estimate kind argument.
* target.def (estimated_poly_value): Update definition for the
above.
* doc/tm.texi: Regenerate.
* targhooks.c (estimated_poly_value): Update prototype.
* tree-vect-loop.c (vect_better_loop_vinfo_p): Use min, max and
likely estimates of VF to pick between vinfos.
* config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Use
estimated_poly_value instead of aarch64_estimated_poly_value.
(aarch64_estimated_poly_value): Take a kind argument and handle
it.

c++: Fix clang problem [PR 98340]

Clang didn't like sizeot (uintset::value) in a templated context.  Not sure
where the problem lies -- ambiguous std, gcc erroneous accept or clang erroneous
reject.  Anyway, this avoids that construct.

PR c++/98340
gcc/cp/
* module.cc (uintset<T>::hash::add): Use uintset (0u).MEMBER,
rather than uintset::MEMBER.