gcc.git
5 years agore PR libgomp/88460 ([nvptx] FAIL: libgomp.c++/for-24.C (internal compiler error))
Jakub Jelinek [Thu, 13 Dec 2018 12:53:19 +0000 (13:53 +0100)]
re PR libgomp/88460 ([nvptx] FAIL: libgomp.c++/for-24.C (internal compiler error))

PR libgomp/88460
* testsuite/libgomp.c++/for-24.C (results): Include it in
omp declare target region.
(main): Use map (always, tofrom: results) instead of
map (tofrom: results).

From-SVN: r267093

5 years agore PR testsuite/88454 (test case gcc.dg/tree-ssa/split-path-5.c fails after r266971)
Jakub Jelinek [Thu, 13 Dec 2018 12:52:11 +0000 (13:52 +0100)]
re PR testsuite/88454 (test case gcc.dg/tree-ssa/split-path-5.c fails after r266971)

PR testsuite/88454
* gcc.dg/tree-ssa/split-path-5.c (__ctype_ptr__): Change type from
const char * to const signed char *.
(bmhi_init): Change pattern parameter's type the same.  Use
__builtin_strlen instead of undeclared strlen.

From-SVN: r267092

5 years agoFix [fs.path.gen] tests to use backslashes for mingw
Jonathan Wakely [Thu, 13 Dec 2018 12:26:52 +0000 (12:26 +0000)]
Fix [fs.path.gen] tests to use backslashes for mingw

The normalized paths contain backslashes so fix the expected values to
use backslashes too.

* testsuite/27_io/filesystem/path/generation/proximate.cc: Use
preferred directory separators for normalized paths.
* testsuite/27_io/filesystem/path/generation/relative.cc: Likewise.

From-SVN: r267090

5 years agoFix test to work when path::native() returns wstring
Jonathan Wakely [Thu, 13 Dec 2018 12:09:33 +0000 (12:09 +0000)]
Fix test to work when path::native() returns wstring

* testsuite/27_io/filesystem/path/itr/traversal.cc: Fix test for
mingw.

From-SVN: r267089

5 years agoS/390: Fix zvector vec_double builtin
Andreas Krebbel [Thu, 13 Dec 2018 11:57:16 +0000 (11:57 +0000)]
S/390: Fix zvector vec_double builtin

The (unsigned) long int to double vector conversion instructions
expect 2 immediate parameters. One for the inexact suppression
control and another one for the rounding mode.  However, the
vec_double builtin has just the vector source operand.  The 2
addtional operands need to be added with an intermediate expander.
The expanders were already there but unfortunately not wired up
correctly to the builtin.

gcc/ChangeLog:

2018-12-13  Andreas Krebbel  <krebbel@linux.ibm.com>

* config/s390/s390-builtins.def (s390_vec_double_s64): Map to
s390_vec_double_s64 instead of s390_vcdgb.
(s390_vec_double_u64): Map to s390_vec_double_u64 instead of
s390_vcdlgb.

gcc/testsuite/ChangeLog:

2018-12-13  Andreas Krebbel  <krebbel@linux.ibm.com>

* gcc.target/s390/zvector/vec-double-1.c: New test.
* gcc.target/s390/zvector/vec-double-2.c: New test.

From-SVN: r267084

5 years agoS/390: Use VEC_INEXACT/VEC_NOINEXACT instead of magic numbers.
Andreas Krebbel [Thu, 13 Dec 2018 11:55:43 +0000 (11:55 +0000)]
S/390: Use VEC_INEXACT/VEC_NOINEXACT instead of magic numbers.

2018-12-13  Andreas Krebbel  <krebbel@linux.ibm.com>

* config/s390/vx-builtins.md ("vec_ctd_s64", "vec_ctd_u64")
("vec_ctsl", "vec_ctul"): Replace 0 with VEC_NOINEXACT.
("vec_double_s64", "vec_double_u64"): Replace 4 with VEC_INEXACT.

From-SVN: r267083

5 years agoDisable new tests for configurations with no libstdc++fs.a
Jonathan Wakely [Thu, 13 Dec 2018 11:01:03 +0000 (11:01 +0000)]
Disable new tests for configurations with no libstdc++fs.a

* testsuite/27_io/filesystem/path/construct/80762.cc: Skip test if
the Filesystem TS support is not configured.
* testsuite/experimental/filesystem/path/construct/80762.cc: Likewise.

From-SVN: r267082

5 years ago[AArch64][doc] Clarify -msve-vector-bits=128 behaviour
Kyrylo Tkachov [Thu, 13 Dec 2018 10:37:15 +0000 (10:37 +0000)]
[AArch64][doc] Clarify -msve-vector-bits=128 behaviour

We've received reports about the -msve-vector-bits=128 bits being somewhat ambiguous.
It isn't clear whether -msve-vector-bits=128 forces vector-length-agnostic code or whether
-msve-vector-bits=scalable forces 128-bit vector-lengh-specific code.
The latter is a, perhaps unintuitive, reading that we want to exclude.

This patch makes it more explicit that -msve-vector-bits=128 is special and produces
vector-length *agnostic* code. In the end, I've rewritten the whole option documentation.

Checked make pdf that the output looks reasonable.

    * doc/invoke.texi (-msve-vector-bits): Clarify -msve-vector-bits=128
    behavior.

From-SVN: r267081

5 years agoi386-common.c (processor_names): Add cascadelake.
Wei Xiao [Thu, 13 Dec 2018 08:57:28 +0000 (08:57 +0000)]
i386-common.c (processor_names): Add cascadelake.

gcc/
      * common/config/i386/i386-common.c (processor_names): Add cascadelake.
      (processor_alias_table): Add cascadelake.
      * config.gcc: Add -march=cascadelake.
      * config/i386/i386-c.c (ix86_target_macros_internal): Handle cascadelake.
      * config/i386/i386.c (Add m_CASCADELAKE): New.
      (processor_cost_table): Add cascadelake.
      (get_builtin_code_for_version): Handle cascadelake.
      * config/i386/i386.h (TARGET_CASCADELAKE, PROCESSOR_CASCADELAKE): New.
      (PTA_CASCADELAKE): Ditto.
      * doc/invoke.texi: Add -march=cascadelake.

gcc/testsuite/
      * gcc.target/i386/funcspec-56.inc: Handle new march.

From-SVN: r267080

5 years agoAdd note about abort and ZCX runtime.
Arnaud Charlet [Thu, 13 Dec 2018 08:11:45 +0000 (09:11 +0100)]
Add note about abort and ZCX runtime.

Part of RC12-001

From-SVN: r267079

5 years agore PR target/88465 (AVX512: optimize loading of constant values to kN registers)
Jakub Jelinek [Thu, 13 Dec 2018 08:08:22 +0000 (09:08 +0100)]
re PR target/88465 (AVX512: optimize loading of constant values to kN registers)

PR target/88465
* config/i386/i386.md (*movdi_internal, *movsi_internal,
*movhi_internal, *movqi_internal): Add alternative(s) to load
0 or -1 into k registers using kxor or kxnoq instructions.

* gcc.target/i386/avx512f-pr88465.c: New test.
* gcc.target/i386/avx512dq-pr88465.c: New test.

From-SVN: r267078

5 years agore PR target/88461 (AVX512: gcc should keep value in kN registers if possible)
Jakub Jelinek [Thu, 13 Dec 2018 08:00:42 +0000 (09:00 +0100)]
re PR target/88461 (AVX512: gcc should keep value in kN registers if possible)

PR target/88461
* config/i386/sse.md (VI1248_AVX512VLBW, AVX512ZEXTMASK): New
mode iterators.
(<avx512>_testm<mode>3<mask_scalar_merge_name>,
<avx512>_testnm<mode>3<mask_scalar_merge_name>): Merge patterns
with VI12_AVX512VL and VI48_AVX512VL iterators into ones with
VI1248_AVX512VLBW iterator.
(*<avx512>_testm<VI1248_AVX512VLBW:mode>3_zext,
*<avx512>_testm<VI1248_AVX512VLBW:mode>3_zext_mask,
*<avx512>_testnm<VI1248_AVX512VLBW:mode>3_zext,
*<avx512>_testnm<VI1248_AVX512VLBW:mode>3_zext_mask): New
define_insns.

* gcc.target/i386/pr88461.c: New test.

From-SVN: r267077

5 years agore PR target/88461 (AVX512: gcc should keep value in kN registers if possible)
Jakub Jelinek [Thu, 13 Dec 2018 07:58:42 +0000 (08:58 +0100)]
re PR target/88461 (AVX512: gcc should keep value in kN registers if possible)

PR target/88461
* config/i386/i386.md (*zero_extendsidi2, zero_extend<mode>di2,
*zero_extend<mode>si2, *zero_extendqihi2): Add =*k, *km alternatives.

From-SVN: r267076

5 years agoDaily bump.
GCC Administrator [Thu, 13 Dec 2018 00:16:35 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r267074

5 years agoruntime: handle DW_EH_PE_absptr in type table encoding
Ian Lance Taylor [Wed, 12 Dec 2018 23:26:58 +0000 (23:26 +0000)]
runtime: handle DW_EH_PE_absptr in type table encoding

    The type table encoding can be DW_EH_PE_absptr, but this case
    was missing, which was causing abort on ARM32 EABI. Add the
    missing case.

    Reviewed-on: https://go-review.googlesource.com/c/153857

From-SVN: r267070

5 years agore PR fortran/88463 (Rejects conforming source, OpenMP Parallel region Default(None...
Jakub Jelinek [Wed, 12 Dec 2018 22:49:35 +0000 (23:49 +0100)]
re PR fortran/88463 (Rejects conforming source, OpenMP Parallel region Default(None) reference to module parameter array, separate source)

PR fortran/88463
* trans-openmp.c (gfc_omp_predetermined_sharing): Handle TREE_READONLY
VAR_DECLs with DECL_EXTERNAL like those with TREE_STATIC.

* testsuite/libgomp.fortran/pr88463-1.f90: New test.
* testsuite/libgomp.fortran/pr88463-2.f90: New test.

From-SVN: r267069

5 years agoos/signal: increase deliver time for signal testcase
Ian Lance Taylor [Wed, 12 Dec 2018 22:48:46 +0000 (22:48 +0000)]
os/signal: increase deliver time for signal testcase

    This increases the time to wait for signals to be delivered in the
    TestAtomicStop testcase. When running gccgo tests on ppc64 or ppc64le,
    there are intermittent failures in this test because the wait time is
    too small.

    Updates golang/go#29046

    Reviewed-on: https://go-review.googlesource.com/c/153879

From-SVN: r267068

5 years agoomp-builtins.def (BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START, [...]): Fix up funct...
Jakub Jelinek [Wed, 12 Dec 2018 22:47:55 +0000 (23:47 +0100)]
omp-builtins.def (BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START, [...]): Fix up function types - remove one argument.

* omp-builtins.def (BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME): Fix up
function types - remove one argument.

* testsuite/libgomp.c-c++-common/for-16.c: New test.

From-SVN: r267067

5 years agodecl.c (grokdeclarator): Fix location of error message about static data member defin...
Paolo Carlini [Wed, 12 Dec 2018 21:36:26 +0000 (21:36 +0000)]
decl.c (grokdeclarator): Fix location of error message about static data member definition.

/cp
2018-12-12  Paolo Carlini  <paolo.carlini@oracle.com>

* decl.c (grokdeclarator): Fix location of error message about
static data member definition.

/testsuite
2018-12-12  Paolo Carlini  <paolo.carlini@oracle.com>

* g++.dg/other/static5.C: New.

From-SVN: r267066

5 years agoextend.texi (attribute copy): Fix typos.
Martin Sebor [Wed, 12 Dec 2018 21:02:41 +0000 (21:02 +0000)]
extend.texi (attribute copy): Fix typos.

gcc/ChangeLog:

* doc/extend.texi (attribute copy): Fix typos.

From-SVN: r267065

5 years agoextend.texi (Function Attributes): Clarify C++ aspects.
Martin Sebor [Wed, 12 Dec 2018 20:48:00 +0000 (20:48 +0000)]
extend.texi (Function Attributes): Clarify C++ aspects.

gcc/ChangeLog:

* doc/extend.texi (Function Attributes): Clarify C++ aspects.
(Variable Attributes): Same.
(Type Attributes): Same.

From-SVN: r267064

5 years agoFix independent-cloneids-1.c testcase (PR88318)
Segher Boessenkool [Wed, 12 Dec 2018 19:45:45 +0000 (20:45 +0100)]
Fix independent-cloneids-1.c testcase (PR88318)

The testcase uses REs like {(?n)\m_*bar[.$_]constprop[.$_]0:} to find
what functions are defined.  But, this also matches lines like
.L.bar.constprop.0:(which are used on powerpc64-linux).
The "(?n)" doesn't do anything here either.  We should use "^" here
instead of just "\m".

PR testsuite/88318
* gcc.dg/independent-cloneids-1.c: Use ^ not \m.

From-SVN: r267063

5 years agoAdd support for SUBTARGET_OVERRIDE_OPTIONS on aarch64
Olivier Hainque [Wed, 12 Dec 2018 17:12:25 +0000 (17:12 +0000)]
Add support for SUBTARGET_OVERRIDE_OPTIONS on aarch64

* config/aarch64/aarch64.c (aarch64_override_options): Once arch,
cpu and tune were validated, insert SUBTARGET_OVERRIDE_OPTIONS if
defined.

From-SVN: r267060

5 years agoUse REGNUM macros in the definitions of aarch64 PROBE_STACK regs
Olivier Hainque [Wed, 12 Dec 2018 17:10:44 +0000 (17:10 +0000)]
Use REGNUM macros in the definitions of aarch64 PROBE_STACK regs

* config/aarch64/aarch64.c (PROBE_STACK_FIRST_REG) : Redefine as
R9_REGNUM instead of 9.
(PROBE_STACK_SECOND_REG): Redefine as R10_REGNUM instead of 10.

From-SVN: r267059

5 years agoFix PR number in my last commit
Richard Earnshaw [Wed, 12 Dec 2018 16:31:12 +0000 (16:31 +0000)]
Fix PR number in my last commit

From-SVN: r267058

5 years agoOverload std::distance and std::advance for path::iterator
Jonathan Wakely [Wed, 12 Dec 2018 16:13:49 +0000 (16:13 +0000)]
Overload std::distance and std::advance for path::iterator

Although filesystem::path::iterator is only a bidirectional iterator,
the underlying sequence has random access iterators (specifically, raw
pointers). This means std::distance and std::advance can be implemented
more efficiently than the generic versions which apply ++ and --
repeatedly.

PR libstdc++/71044 (partial)
* include/bits/fs_path.h (__path_iter_distance, __path_iter_advance):
New friend functions to implement std::distance and std::advance more
efficiently.
(distance, advance): Add overloads for path::iterator.
* testsuite/27_io/filesystem/path/itr/components.cc: Test new
overload.

From-SVN: r267057

5 years agoPR libstdc++/80762 avoid ambiguous __constructible_from<void, void>
Jonathan Wakely [Wed, 12 Dec 2018 16:13:43 +0000 (16:13 +0000)]
PR libstdc++/80762 avoid ambiguous __constructible_from<void, void>

Ensure we don't try to instantiate __is_constructible_from<void, void>,
because there are two partial specializations that are equally good
matches.

PR libstdc++/80762
* include/bits/fs_path.h (path::_Path): Use remove_cv_t and is_void.
* include/experimental/bits/fs_path.h (path::_Path): Likewise.
* testsuite/27_io/filesystem/path/construct/80762.cc: New test.
* testsuite/experimental/filesystem/path/construct/80762.cc: New test.

From-SVN: r267056

5 years agorequire profiling support for gcc.dg/lto/20100430-1_0.c test
Joel Brobecker [Wed, 12 Dec 2018 14:22:37 +0000 (14:22 +0000)]
require profiling support for gcc.dg/lto/20100430-1_0.c test

This test currently fails unexpectedly if GCC is configured with
--disable-gcov, because it requires -fprofile-arcs. This patch
fixes the issue by requiring profiling support in order to run
this test.

Tested with two compilers, one built with --disable-gcov, resulting
in the test reporting an UNSUPPORTED result; and one built with gcov
support, resulting in 2 PASS tests.

gcc/testsuite/ChangeLog:

        * gcc.dg/lto/20100430-1_0.c: Add dg-require-profiling requirement.

From-SVN: r267055

5 years ago[libbacktrace] Add allocfail.sh test-case
Tom de Vries [Wed, 12 Dec 2018 13:50:51 +0000 (13:50 +0000)]
[libbacktrace] Add allocfail.sh test-case

Add test-case that forces alloc.c functions to fail, and check whether fail
handling is robust.

This is the test-case for "[libbacktrace] Fix segfault upon allocation
failure".  Without that patch, this test-case fails like this:
...
allocfail.sh: line 71: 26041 Segmentation fault      (core dumped) \
  ./allocfail $i > /dev/null 2>&1
Unallowed fail found: 13
FAIL allocfail.sh (exit status: 1)
...

This is a seperate patch because the test-case is nontrivial.

Bootstrapped and reg-tested on x86_64.

2018-12-12  Tom de Vries  <tdevries@suse.de>

* Makefile.am (TESTS): Add allocfail.sh.
(check_PROGRAMS): Add allocfail.
* Makefile.in: Regenerate.
* instrumented_alloc.c: New file.  Redefine malloc and realloc.
Include alloc.c.
* allocfail.c: New file.
* allocfail.sh: New file.

From-SVN: r267054

5 years agore PR target/86806 (SPARC port needs updating for CVE-2017-5753)
Eric Botcazou [Wed, 12 Dec 2018 11:40:28 +0000 (11:40 +0000)]
re PR target/86806 (SPARC port needs updating for CVE-2017-5753)

PR target/86806
* config/sparc/sparc.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
(speculation_barrier): New instruction for V9.

From-SVN: r267053

5 years agoDon't mark stack pointer as clobbered in asm
Andreas Schwab [Wed, 12 Dec 2018 11:22:42 +0000 (11:22 +0000)]
Don't mark stack pointer as clobbered in asm

* config/linux/ia64/futex.h (sys_futex0): Don't mark r12 as
clobbered.

From-SVN: r267052

5 years agolibgcc: rs6000: tramp.S: fix placement of .cfi_endproc for __trampoline_setup
Rasmus Villemoes [Wed, 12 Dec 2018 10:12:36 +0000 (11:12 +0100)]
libgcc: rs6000: tramp.S: fix placement of .cfi_endproc for __trampoline_setup

Currently, .cfi_endproc and FUNC_END(__trampoline_setup) are placed
inside the #else branch of an "#if defined (__VXWORKS__) ...", so
non-pic vxworks does not get proper CFI nor a .size directive for
__trampoline_setup. I assume there's no magic reason for that (which
would warrant a comment), so move them outside.

From-SVN: r267051

5 years agors6000 - add missed hunk
Iain Sandoe [Wed, 12 Dec 2018 09:25:13 +0000 (09:25 +0000)]
rs6000 - add missed hunk

This add the hunk
(macho_call_template): Remove.

missed from r267049

From-SVN: r267050

5 years agors6000 - Fix PR target/88343
Segher Boessenkool [Wed, 12 Dec 2018 08:53:49 +0000 (09:53 +0100)]
rs6000 - Fix PR target/88343

The PR is about unnecessary saves of the pic base register, it shows on m32 Linux and m32/m64 Darwin.

The fix is to check that we are in a pic mode and that the picbase has actually been used.
As a bonus, some #ifdef’d TARGET_MACHO code is no longer required.

2018-12-12 Segher Boessenkool  <segher@kernel.crashing.org>
   Iain Sandoe  <iain@sandoe.co.uk>

PR target/88343
* config/rs6000/rs6000.c (save_reg_p): Do not save the picbase reg
unless it has been used.
(first_reg_to_save): Remove dead code.

Co-Authored-By: Iain Sandoe <iain@sandoe.co.uk>
From-SVN: r267049

5 years agors6000, update Darwin call handling to latest scheme.
Iain Sandoe [Wed, 12 Dec 2018 08:47:16 +0000 (08:47 +0000)]
rs6000, update Darwin call handling to latest scheme.

This patch mormalises Darwin's call handling to match the scheme
recently introduced and gives it target-specific call expanders.

2018-12-12  Iain Sandoe  <iain@sandoe.co.uk>

* config/rs6000/rs6000-protos.h (rs6000_call_darwin): New.
(rs6000_sibcall_darwin): New. (macho_call_template): Remove.
* config/rs6000/rs6000.c (get_prev_label): Forward declaration.
(rs6000_call_template_1): Handle Darwin.
(macho_call_template): Remove.
(rs6000_call_sysv): Remove handling for Darwin.
(rs6000_call_darwin_1, rs6000_call_darwin): New
(rs6000_sibcall_darwin): New.
* config/rs6000/rs6000.md (define_expand “call”): Handle Darwin
with its own expander. (define_expand “call_value”): Likewise.
(define_expand “sibcall”): Likewise.
(define_expand “sibcall_value”): Likewise.
(call_nonlocal_sysv): Remove Darwin special-casing.
(call_value_nonlocal_sysv): Likewise.

From-SVN: r267048

5 years agore PR c++/88446 (__builtin_is_constant_evaluated rejects some converted constant...
Jakub Jelinek [Wed, 12 Dec 2018 08:44:27 +0000 (09:44 +0100)]
re PR c++/88446 (__builtin_is_constant_evaluated rejects some converted constant expressions.)

PR c++/88446
* cp-tree.h (maybe_constant_value): Add manifestly_const_eval
argument.
* constexpr.c (struct constexpr_call): Rename pretend_const_required
member to manifestly_const_eval.
(struct constexpr_ctx): Likewise.
(constexpr_call_hasher::equal): Adjust users.
(cxx_eval_builtin_function_call): Likewise.  Formatting fix.
(cxx_eval_call_expression): Adjust users.
(cxx_eval_outermost_constant_expr, maybe_constant_init_1,
maybe_constant_init): Rename pretend_const_required argument to
manifestly_const_eval, adjust function comments.
(maybe_constant_value): Add manifestly_const_eval argument.  If true,
don't cache and call cxx_eval_outermost_constant_expr with true as
manifestly_const_eval.
* decl.c (compute_array_index_type_loc): Call maybe_constant_value
with true as manifestly_const_eval.

* g++.dg/cpp2a/is-constant-evaluated3.C: New test.

From-SVN: r267047

5 years agors6000, darwin - fix bootstrap.
Iain Sandoe [Wed, 12 Dec 2018 08:42:28 +0000 (08:42 +0000)]
rs6000, darwin - fix bootstrap.

Recent changes to the lowerings for call handling in rs6000
have broken bootstrap for powerpc-darwin.  This patch modifies
the calls to handle Darwin asa subset, a subsequent patch will
split Darwin out.

2018-12-12  Iain Sandoe  <iain@sandoe.co.uk>

* config/rs6000/darwin.md (call_indirect_nonlocal_darwin64): Remove.
(call_nonlocal_darwin64): Remove.
(call_value_indirect_nonlocal_darwin64): Remove.
(call_value_nonlocal_darwin64): Remove.
* config/rs6000/rs6000.c (rs6000_call_template_1): Handle Darwin with
the same asm output as AIX/ELFv2.  (rs6000_call_sysv): Preserve the
CALL_LONG flag when needed for Darwin.
* config/rs6000/rs6000.md (define expand “call”): Expand Darwin as
per sysv.
(define_expand “call_value”): Likewise.
(define_expand “sibcall”): Likewise.
(define_expand “sibcall_value”): Likewise.
(call_indirect_nonlocal_sysv): Mark the clobber mode P.
(call_nonlocal_sysv): Likewise.
(call_nonlocal_sysv_secure): Likewise.
(call_value_indirect_nonlocal_sysv): Likewise.
(call_value_nonlocal_sysv): Likewise.
(call_value_nonlocal_sysv_secure): Likewise.
(call_local64): Mark the clobber DI.
(call_value_local64): Likewise.

From-SVN: r267046

5 years agoP0595R2 - is_constant_evaluated
Jakub Jelinek [Wed, 12 Dec 2018 08:31:01 +0000 (09:31 +0100)]
P0595R2 - is_constant_evaluated

P0595R2 - is_constant_evaluated
* include/bits/c++config (_GLIBCXX_HAVE_BUILTIN_IS_CONSTANT_EVALUATED):
Define if __builtin_is_constant_evaluated is available.
* include/std/type_traits (std::is_constant_evaluated): New constexpr
inline function.
* testsuite/20_util/is_constant_evaluated/1.cc: New test.
* testsuite/20_util/is_constant_evaluated/noexcept.cc: New test.

From-SVN: r267045

5 years agore PR c++/88449 (__builtin_is_constant_evaluated misbehaves due to constexpr call...
Jakub Jelinek [Wed, 12 Dec 2018 08:28:43 +0000 (09:28 +0100)]
re PR c++/88449 (__builtin_is_constant_evaluated misbehaves due to constexpr call caching)

PR c++/88449
* constexpr.c (struct constexpr_call): Add pretend_const_required
member.
(constexpr_call_hasher::equal): Return false if pretend_const_required
members differ.
(cxx_eval_call_expression): Adjust new_call initialization.  Hash in
ctx->pretend_const_required.

* g++.dg/cpp2a/is-constant-evaluated1.C: Change from dg-do compile
to dg-do run.
(e): Adjust comment with correct expected value.
(main): Expect e == 1.
* g++.dg/cpp2a/is-constant-evaluated2.C: New test.

From-SVN: r267044

5 years agore PR fortran/88155 (ICE in gfc_format_decoder, at fortran/error.c:947)
Steven G. Kargl [Wed, 12 Dec 2018 00:53:08 +0000 (00:53 +0000)]
re PR fortran/88155 (ICE in gfc_format_decoder, at fortran/error.c:947)

2018-12-11  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/88155
* primary.c (gfc_match_structure_constructor):  Set the locus of
an expression to avoid a NULL pointer dereference.

2018-12-11  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/88155
* gfortran.dg/pr70870_1.f90: Update testcase to use -std=gnu.
* gfortran.dg/pr88155.f90: New test.

From-SVN: r267041

5 years agoDaily bump.
GCC Administrator [Wed, 12 Dec 2018 00:16:18 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r267040

5 years agore PR fortran/88249 (ICE in gfc_resolve_filepos, at fortran/io.c:2853)
Steven G. Kargl [Tue, 11 Dec 2018 23:13:19 +0000 (23:13 +0000)]
re PR fortran/88249 (ICE in gfc_resolve_filepos, at fortran/io.c:2853)

2018-12-11  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/88249
* gfortran.h: Update prototype for gfc_resolve_filepos().
* io.c (gfc_resolve_filepos): Check for UNIT number if ERR= is present.
Use passed in locus for error message.
* resolve.c (gfc_resolve_code): Pass locus in gfc_resolve_filepos()
call.

2018-12-11  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/88249
* gfortran.dg/pr88249.f90: New test.

From-SVN: r267035

5 years agore PR ada/88429 (libada build fails with --disable-shared)
Eric Botcazou [Tue, 11 Dec 2018 23:04:39 +0000 (23:04 +0000)]
re PR ada/88429 (libada build fails with --disable-shared)

libada/
PR ada/88429
* configure.ac (default_gnatlib_target): Set to gnatlib instead of
gnatlib-plain if --disable-shared.
* configure: Regenerate.
* Makefile.in (all): Replace gnatlib prerequisite with libada.
(ADA_RTS_SUBDIR): Delete.
(libada): New target, renamed from...
(gnatlib): ...this.  Merge with other library targets.
(gnatlib-plain): Delete.
(install-gnatlib): Rename to...
(install-libada): ...this.
(install): Replace install-gnatlib prerequisite with install-libada.
gcc/ada/
PR ada/88429
* gcc-interface/Makefile.in (./stamp-gnatlib1-$(RTSDIR)): Also pass
MULTISUBDIR to sub-make and add quotes around $(THREAD_KIND).
(gnatlib-shared-dual): Also pass PICFLAG_FOR_TARGET to sub-make.
(gnatlib-sjlj): Also pass MULTISUBDIR to sub-make, but do not pass
PICFLAG_FOR_TARGET.
(gnatlib-zcx): Likewise.

From-SVN: r267034

5 years agoruntime: use _URC_FAILURE on ARM32
Ian Lance Taylor [Tue, 11 Dec 2018 20:50:59 +0000 (20:50 +0000)]
runtime: use _URC_FAILURE on ARM32

    ARM32 EABI unwinder does not define _URC_NORMAL_STOP. Instead,
    it has _URC_FAILURE. Use _URC_FAILURE there.

    Should fix ARM32 build.

    Reviewed-on: https://go-review.googlesource.com/c/153417

From-SVN: r267033

5 years agore PR c++/87861 (ICE in output_constructor_regular_field, at varasm.c:5165)
Jakub Jelinek [Tue, 11 Dec 2018 20:37:53 +0000 (21:37 +0100)]
re PR c++/87861 (ICE in output_constructor_regular_field, at varasm.c:5165)

PR c++/87861
* class.c (build_vtbl_initializer): For TARGET_VTABLE_USES_DESCRIPTORS
bump index for each added word.
* constexpr.c (find_array_ctor_elt): Add forward declaration.
(cxx_eval_call_expression): Handle TARGET_VTABLE_USES_DESCRIPTORS
vtable calls.
(cxx_eval_constant_expression) <case OBJ_TYPE_REF>: Divide token
by TARGET_VTABLE_USES_DESCRIPTORS if non-zero.

From-SVN: r267032

5 years agore PR tree-optimization/80520 (Performance regression from missing if-conversion)
Jakub Jelinek [Tue, 11 Dec 2018 20:28:35 +0000 (21:28 +0100)]
re PR tree-optimization/80520 (Performance regression from missing if-conversion)

PR tree-optimization/80520
* gcc.dg/tree-ssa/split-path-11.c (foo): Make the test ilp32 target
clean.

From-SVN: r267031

5 years agoPR c++/86608 - reading constexpr volatile variable.
Marek Polacek [Tue, 11 Dec 2018 18:53:03 +0000 (18:53 +0000)]
PR c++/86608 - reading constexpr volatile variable.

* constexpr.c (potential_constant_expression_1): Check want_rval
instead of checking if we have a decl.
* decl2.c (decl_maybe_constant_var_p): Don't consider volatile
constexpr variables as maybe constant.

* g++.dg/cpp0x/constexpr-volatile2.C: New test.
* g++.dg/cpp0x/pr65327.C: Add dg-error.

From-SVN: r267030

5 years agodecl.c (grokvardecl): Add location_t parameter and use it in build_lang_decl_loc...
Paolo Carlini [Tue, 11 Dec 2018 17:47:56 +0000 (17:47 +0000)]
decl.c (grokvardecl): Add location_t parameter and use it in build_lang_decl_loc and build_decl calls.

/cp
2018-12-11  Paolo Carlini  <paolo.carlini@oracle.com>

* decl.c (grokvardecl): Add location_t parameter and use it
in build_lang_decl_loc and build_decl calls.
(grokdeclarator): Move up loc declaration and use it in the
grokvardecl call too.

/testsuite
2018-12-11  Paolo Carlini  <paolo.carlini@oracle.com>

* g++.dg/pr53037-4.C: Test the first two locations too.

From-SVN: r267029

5 years agodecl.c (grokvardecl): Add location_t parameter and use it in build_lang_decl_loc...
Paolo Carlini [Tue, 11 Dec 2018 17:47:47 +0000 (17:47 +0000)]
decl.c (grokvardecl): Add location_t parameter and use it in build_lang_decl_loc and build_decl calls.

/cp
2018-12-11  Paolo Carlini  <paolo.carlini@oracle.com>

* decl.c (grokvardecl): Add location_t parameter and use it
in build_lang_decl_loc and build_decl calls.
(grokdeclarator): Move up loc declaration and use it in the
grokvardecl call too.

/testsuite
2018-12-11  Paolo Carlini  <paolo.carlini@oracle.com>

* g++.dg/pr53037-4.C: Test the first two locations too.

From-SVN: r267028

5 years agodecl2.c (grokvardecl): Add location_t parameter and use it in build_lang_decl_loc...
Paolo Carlini [Tue, 11 Dec 2018 17:46:42 +0000 (17:46 +0000)]
decl2.c (grokvardecl): Add location_t parameter and use it in build_lang_decl_loc and build_decl calls.

/cp
2018-12-11  Paolo Carlini  <paolo.carlini@oracle.com>

* decl2.c (grokvardecl): Add location_t parameter and use it
in build_lang_decl_loc and build_decl calls.
(grokdeclarator): Move up loc declaration and use it in the
grokvardecl call too.

/testsuite
2018-12-11  Paolo Carlini  <paolo.carlini@oracle.com>

* g++.dg/pr53037-4.C: Test the first two locations too.

From-SVN: r267027

5 years agore PR tree-optimization/88444 (ICE: tree check: expected ssa_name, have integer_cst...
Jakub Jelinek [Tue, 11 Dec 2018 17:16:48 +0000 (18:16 +0100)]
re PR tree-optimization/88444 (ICE: tree check: expected ssa_name, have integer_cst in live_on_edge, at tree-vrp.c:468; or ICE: tree check: expected ssa_name, have integer_cst in get_value_range, at vr-values.c:84)

PR tree-optimization/88444
* tree-vrp.c (register_edge_assert_for_2): Only register assertions
for conversions if rhs1 is a SSA_NAME.

* gcc.dg/pr88444.c: New test.

From-SVN: r267026

5 years agocfgexpand.c (asm_clobber_reg_is_valid): Also produce error when stack pointer is...
Dimitar Dimitrov [Tue, 11 Dec 2018 15:50:51 +0000 (15:50 +0000)]
cfgexpand.c (asm_clobber_reg_is_valid): Also produce error when stack pointer is clobbered.

gcc/
2018-12-11  Dimitar Dimitrov  <dimitar@dinux.eu>

* cfgexpand.c (asm_clobber_reg_is_valid): Also produce
error when stack pointer is clobbered.
(expand_asm_stmt): Refactor clobber check in separate function.

gcc/testsuite/
2018-12-11  Dimitar Dimitrov  <dimitar@dinux.eu>

* gcc.target/i386/pr52813.c: New test.

From-SVN: r267025

5 years agoFix stack offsets and alloca alignments on ppc-vxworks
Eric Botcazou [Tue, 11 Dec 2018 15:03:29 +0000 (15:03 +0000)]
Fix stack offsets and alloca alignments on ppc-vxworks

* config/rs6000/vxworks.h (RS6000_STARTING_FRAME_OFFSET): Define,
accounting for STACK_BOUNDARY 128 regardless of TARGET_ALTIVEC/VSX.
(STACK_DYNAMIC_OFFSET): Likewise.

From-SVN: r267024

5 years agore PR target/88425 (suboptimal code for a<imm?-1:0)
Jakub Jelinek [Tue, 11 Dec 2018 14:50:22 +0000 (15:50 +0100)]
re PR target/88425 (suboptimal code for a<imm?-1:0)

PR target/88425
* config/i386/i386.md (*x86_mov<SWI48:mode>cc_0_m1_neg_leu<SWI:mode>):
New define_insn_and_split.

* gcc.target/i386/pr88425.c: New test.

From-SVN: r267023

5 years agore PR sanitizer/88426 (Compiler crash if use special code with command line switch...
Jakub Jelinek [Tue, 11 Dec 2018 13:51:44 +0000 (14:51 +0100)]
re PR sanitizer/88426 (Compiler crash if use special code with command line switch -fsanitize=float-cast-overflow)

PR sanitizer/88426
* c-convert.c (convert): Call c_fully_fold before calling
ubsan_instrument_float_cast.

* c-c++-common/ubsan/float-cast-overflow-11.c: New test.

From-SVN: r267022

5 years agore PR middle-end/88448 (gnat.dg/opt66.adb etc. FAIL)
Richard Biener [Tue, 11 Dec 2018 13:00:49 +0000 (13:00 +0000)]
re PR middle-end/88448 (gnat.dg/opt66.adb etc. FAIL)

2018-12-11  Richard Biener  <rguenther@suse.de>

PR middle-end/88448
PR middle-end/88415
* gimple.c (gimple_assign_set_rhs_with_ops): Revert previous
change.
* tree-complex.c (update_complex_assignment): Properly transfer
or clean EH info around gimple_assign_set_rhs_with_ops.

From-SVN: r267021

5 years agoFix a failure in the libiberty testsuite by increasing the demangle recursion limit...
Nick Clifton [Tue, 11 Dec 2018 11:59:53 +0000 (11:59 +0000)]
Fix a failure in the libiberty testsuite by increasing the demangle recursion limit to 2048.

PR 88409
* demangle.h (DEMANGLE_RECURSION_LIMIT): Increase to 2048.

From-SVN: r267020

5 years ago[aarch64] PR target/87369 Prefer bsl/bit/bif for copysign
Richard Earnshaw [Tue, 11 Dec 2018 11:26:15 +0000 (11:26 +0000)]
[aarch64] PR target/87369 Prefer bsl/bit/bif for copysign

The copysign operations will almost always be performed on values in
floating-point registers.  As such, we do not want the compiler to
simplify the operations into code sequences that can only be done
using the general-purpose register set.  Unfortunately, this is what
is currently happening.

Fortunately, it seems quite unlikely that copysign() will be
subsequently followed by other logical operations on the values
involved, so I think it is acceptable to use an unspec here.  This
allows us to preserve the operation in a form that allows the register
allocator to make the right choice later on, without limitation on the
final form of the operation (well, if we do end up using the gp
register bank, we get a dead constant load that we cannot easily
eliminate at a late stage).

PR target/37369
* config/aarch64/iterators.md (sizem1): Add sizes for SFmode and DFmode.
(Vbtype): Add SFmode mapping.
* config/aarch64/aarch64.md (copysigndf3, copysignsf3): Delete.
(copysign<GPF:mode>3): New expand pattern.
(copysign<GPF:mode>3_insn): New insn pattern.

From-SVN: r267019

5 years ago[Ada] Add "Global => null" contracts to Ada.Calendar routines
Piotr Trojanek [Tue, 11 Dec 2018 11:12:41 +0000 (11:12 +0000)]
[Ada] Add "Global => null" contracts to Ada.Calendar routines

Routines in Ada.Real_Time are already annotated with Global => null
contracts to suppress spurious warnings from the flow analysis in
GNATprove. This patch adds such contracts to Ada.Calendar. No change in
runtime behavior expected.

2018-12-11  Piotr Trojanek  <trojanek@adacore.com>

gcc/ada/

* libgnat/a-calend.ads: Add "Global => null" contracts to pure
routines.

From-SVN: r267018

5 years ago[Ada] Spurious error with pragma Thread_Local_Storage
Hristian Kirtchev [Tue, 11 Dec 2018 11:12:37 +0000 (11:12 +0000)]
[Ada] Spurious error with pragma Thread_Local_Storage

The following patch modifies the checks related to pragma
Thread_Local_Storage to correct a confusion in semantics which led to
spurious errors.

------------
-- Source --
------------

--  pack.ads

package Pack is
   type Arr is array (1 .. 5) of Boolean;

   type Arr_With_Default is array (1 .. 5) of Boolean
     with Default_Component_Value => False;

   type Int is new Integer range 1 .. 5;

   type Int_With_Default is new Integer range 1 .. 5
     with Default_Value => 1;

   protected type Prot_Typ is
      entry E;
   end Prot_Typ;

   type Rec_1 is record
      Comp : Integer;
   end record;

   type Rec_2 is record
      Comp : Int;
   end record;

   type Rec_3 is record
      Comp : Int_With_Default;
   end record;

   task type Task_Typ is
      entry E;
   end Task_Typ;
end Pack;

--  pack.adb

package body Pack is
   function F (Val : Int) return Int is
   begin
      if Val <= 1 then
         return 1;
      else
         return F (Val - 1) * Val;
      end if;
   end F;

   function F (Val : Int_With_Default) return Int_With_Default is
   begin
      if Val <= 1 then
         return 1;
      else
         return F (Val - 1) * Val;
      end if;
   end F;

   function F (Val : Integer) return Integer is
   begin
      if Val <= 1 then
         return 1;
      else
         return F (Val - 1) * Val;
      end if;
   end F;

   protected body Prot_Typ is
      entry E when True is begin null; end E;
   end Prot_Typ;

   task body Task_Typ is
   begin
      accept E;
   end Task_Typ;

   Obj_1 : Arr;                                                      --  OK
   pragma Thread_Local_Storage (Obj_1);

   Obj_2 : Arr := (others => True);                                  --  OK
   pragma Thread_Local_Storage (Obj_2);

   Obj_3 : Arr := (others => F (2) = Integer (3));                   --  ERROR
   pragma Thread_Local_Storage (Obj_3);

   Obj_4 : Arr_With_Default;                                         --  ERROR
   pragma Thread_Local_Storage (Obj_4);

   Obj_5 : Arr_With_Default := (others => True);                     --  OK
   pragma Thread_Local_Storage (Obj_5);

   Obj_6 : Arr_With_Default := (others => F (2) = Integer (3));      --  ERROR
   pragma Thread_Local_Storage (Obj_6);

   Obj_7 : Integer;                                                  --  OK
   pragma Thread_Local_Storage (Obj_7);

   Obj_8 : Integer := 1;                                             --  OK
   pragma Thread_Local_Storage (Obj_8);

   Obj_9 : Integer := F (2);                                         --  ERROR
   pragma Thread_Local_Storage (Obj_9);

   Obj_10 : Int;                                                     --  OK
   pragma Thread_Local_Storage (Obj_10);

   Obj_11 : Int := 1;                                                --  OK
   pragma Thread_Local_Storage (Obj_11);

   Obj_12 : Int := F (2);                                            --  ERROR
   pragma Thread_Local_Storage (Obj_12);

   Obj_13 : Int_With_Default;                                        --  ERROR
   pragma Thread_Local_Storage (Obj_13);

   Obj_14 : Int_With_Default := 1;                                   --  OK
   pragma Thread_Local_Storage (Obj_14);

   Obj_15 : Int_With_Default := F (2);                               --  ERROR
   pragma Thread_Local_Storage (Obj_15);

   Obj_16 : Prot_Typ;                                                --  ERROR
   pragma Thread_Local_Storage (Obj_16);

   Obj_17 : Rec_1;                                                   --  OK
   pragma Thread_Local_Storage (Obj_17);

   Obj_18 : Rec_1 := (others => 1);                                  --  OK
   pragma Thread_Local_Storage (Obj_18);

   Obj_19 : Rec_1 := (others => F (2));                              --  ERROR
   pragma Thread_Local_Storage (Obj_19);

   Obj_20 : Rec_2;                                                   --  OK
   pragma Thread_Local_Storage (Obj_20);

   Obj_21 : Rec_2 := (others => 1);                                  --  OK
   pragma Thread_Local_Storage (Obj_21);

   Obj_22 : Rec_2 := (others => F (2));                              --  ERROR
   pragma Thread_Local_Storage (Obj_22);

   Obj_23 : Rec_3;                                                   --  ERROR
   pragma Thread_Local_Storage (Obj_23);

   Obj_24 : Rec_3 := (others => 1);                                  --  OK
   pragma Thread_Local_Storage (Obj_24);

   Obj_25 : Rec_3 := (others => F (2));                              --  ERROR
   pragma Thread_Local_Storage (Obj_25);

   Obj_26 : Task_Typ;                                                --  ERROR
   pragma Thread_Local_Storage (Obj_26);
end Pack;

----------------------------
-- Compilation and output --
----------------------------

$ gcc -c pack.adb
pack.adb:47:04: Thread_Local_Storage variable "Obj_4" is improperly
  initialized
pack.adb:47:04: only allowed initialization is explicit "null", static
  expression or static aggregate
pack.adb:62:04: Thread_Local_Storage variable "Obj_9" is improperly
  initialized
pack.adb:62:04: only allowed initialization is explicit "null", static
  expression or static aggregate
pack.adb:71:04: Thread_Local_Storage variable "Obj_12" is improperly
  initialized
pack.adb:71:04: only allowed initialization is explicit "null", static
  expression or static aggregate
pack.adb:74:04: Thread_Local_Storage variable "Obj_13" is improperly
  initialized
pack.adb:74:04: only allowed initialization is explicit "null", static
  expression or static aggregate
pack.adb:80:04: Thread_Local_Storage variable "Obj_15" is improperly
  initialized
pack.adb:80:04: only allowed initialization is explicit "null", static
  expression or static aggregate
pack.adb:83:04: Thread_Local_Storage variable "Obj_16" is improperly
  initialized
pack.adb:83:04: only allowed initialization is explicit "null", static
  expression or static aggregate
pack.adb:92:04: Thread_Local_Storage variable "Obj_19" is improperly
  initialized
pack.adb:92:04: only allowed initialization is explicit "null", static
  expression or static aggregate
pack.adb:101:04: Thread_Local_Storage variable "Obj_22" is improperly
  initialized
pack.adb:101:04: only allowed initialization is explicit "null", static
  expression or static aggregate
pack.adb:104:04: Thread_Local_Storage variable "Obj_23" is improperly
  initialized
pack.adb:104:04: only allowed initialization is explicit "null", static
  expression or static aggregate
pack.adb:110:04: Thread_Local_Storage variable "Obj_25" is improperly
  initialized
pack.adb:110:04: only allowed initialization is explicit "null", static
  expression or static aggregate
pack.adb:113:04: Thread_Local_Storage variable "Obj_26" is improperly
  initialized
pack.adb:113:04: only allowed initialization is explicit "null", static
  expression or static aggregate

2018-12-11  Hristian Kirtchev  <kirtchev@adacore.com>

gcc/ada/

* freeze.adb (Check_Pragma_Thread_Local_Storage): Use the
violating set to diagnose detect an illegal initialization,
rather than the complement of the OK set.
(Freeze_Object_Declaration): Factorize code in
Has_Default_Initialization.
(Has_Default_Initialization, Has_Incompatible_Initialization):
New routines.

From-SVN: r267017

5 years ago[Ada] GNAT.Sockets: add IPv6 support
Dmitriy Anisimkov [Tue, 11 Dec 2018 11:12:32 +0000 (11:12 +0000)]
[Ada] GNAT.Sockets: add IPv6 support

2018-12-11  Dmitriy Anisimkov  <anisimko@adacore.com>

gcc/ada/

* libgnat/g-socket.ads (Family_Type): Add new enumerated value
Family_Unspec to be able to use it in Get_Address_Info parameter
and find IPv4 together with IPv6 addresses.
(Inet_Addr_Bytes_Length): Zero length for Family_Unspec.  New
IPv6 predefined constant addresses.
(IPv4_To_IPv6_Prefix): IPv4 mapped to IPv6 address prefix.
(Is_IPv4_Address): Rename from Is_IP_Address and published.
(Is_IPv6_Address): New routine.
(Image of Inet_Addr_Type): Fix description about IPv6 address
text representation.
(Level_Type): New propocol level IP_Protocol_For_IPv6_Level.
(Add_Membership_V4): New socket option equal to Add_Membership.
(Drop_Membership_V4): New socket option equal to
Drop_Membership.
(Multicast_If_V4): New socket option equal to Multicast_If.
(Multicast_Loop_V4, Add_Membership_V6, Drop_Membership_V6,
Multicast_If_V6, Multicast_Loop_V6, Multicast_Hops, IPv6_Only):
New socket option for IPv6.
(Address_Info): New record to keep address info.
(Address_Info_Array): Array to keep address info records.
(Get_Address_Info): Routine to get address info records by host
and service names.
(Host_Service): Record to keep host and service names.
(Get_Name_Info): New routine to get host and service names by
address.
(Create_Socket): Add Level parameter, IP_Protocol_For_IP_Level
default.
(Name_Array, Inet_Addr_Array): Change array index to Positive.
* libgnat/g-socket.adb (IPV6_Mreq): New record definition for
IPv6.
(Hex_To_Char): Remove.
(Short_To_Network, Network_To_Short): Move to package
GNAT.Sockets.Thin_Common.
(Is_IP_Address): Remove.
(To_In_Addr, To_Inet_Addr): Move to package
GNAT.Sockets.Thin_Common.
(Get_Socket_Option): Get value of Multicast_Loop option as
integer boolean, process IPv6 options. Don't try to get
Add_Membership_V4, Add_Membership_V6, Drop_Membership_V4, and
Drop_Membership_V6 as not supported by the socket API.
(Set_Socket_Option): Set value of Multicast_Loop option as
integer boolean, process IPv6 options.
* gsocket.h
(IPV6_ADD_MEMBERSHIP): Define from IPV6_JOIN_GROUP if necessary
for VxWorks.
(IPV6_DROP_MEMBERSHIP): Define from IPV6_LEAVE_GROUP if
necessary for VxWorks
(HAVE_INET_NTOP): New definition.
(HAVE_INET_PTON): Includes VxWorks now.
* socket.c (__gnat_getaddrinfo, __gnat_getnameinfo,
__gnat_freeaddrinfo, __gnat_gai_strerror, __gnat_inet_ntop): New
routines.
* libgnat/g-sothco.ads, libgnat/g-sothco.adb
(socklen_t, In6_Addr, To_In6_Addr): New.
(To_In_Addr, To_Inet_Addr): Move from package body GNAT.Sockets.
(To_Inet_Addr): New overload with In6_Addr type parmeter.
(In_Addr_Access_Array): Remove.
(Sockaddr): Unchecked_Union instead of Sockaddr_In and old
defined generic Sockaddr.
(Set_Address): Use it to set family, port and address into
Sockaddr.
(Get_Address): New routine to get Socket_Addr_Type from
Sockaddr.
(Addrinfo): Structure to use with getaddrinfo.
(C_Getaddrinfo, C_Freeaddrinfo, C_Getnameinfo, C_GAI_Strerror,
Inet_Ntop): New routine import.
(Short_To_Network, Network_To_Short): Move from package body
GNAT.Sockets.
* libgnat/g-stsifd__sockets.adb: Use Sockaddr instead of
Sockaddr_In.
* s-oscons-tmplt.c (AF_UNSPEC, EAI_SYSTEM, SOCK_RAW,
IPPROTO_IPV6, IP_RECVERR, SIZEOF_socklen_t, IF_NAMESIZE): New
constants.
(AI_xxxx_OFFSET): Constants to consider platform differences in
field positions and sizes for addrinfo structure.
(AI_xxxxx): Flags for getaddrinfo.
(NI_xxxxx): Flags for getnameinfo.
(IPV6_xxxxx): Socket options for IPv6.
(Inet_Ntop_Linkname): New routine.

From-SVN: r267016

5 years ago[Ada] Do not expand code inside ignored ghost bodies
Yannick Moy [Tue, 11 Dec 2018 11:12:26 +0000 (11:12 +0000)]
[Ada] Do not expand code inside ignored ghost bodies

While ignored ghost code is not compiled into the executable, it may
lead to compilation errors when it makes use of language features
requiring runtime support that is not available in the available runtime
library.  These errors are spurious, as the executable will never call
in these runtime units.

This patch deactivates the expansion of code inside ignored ghost bodies
of subprograms and packages, so that this code is still checked for
possible semantic errors, but it does not force the presence of useless
runtime units.

There is no impact on the executable produced.

2018-12-11  Yannick Moy  <moy@adacore.com>

gcc/ada/

* sem_ch6.adb (Analyze_Subprogram_Body_Helper): Deactivate
expansion in ignored ghost subprogram body.
* sem_ch7.adb (Analyze_Package_Body_Helper): Deactivate
expansion in ignored ghost package body.

gcc/testsuite/

* gnat.dg/ghost4.adb: New testcase.

From-SVN: r267015

5 years ago[Ada] Unnesting: do not eliminate subprogram whose address is taken
Ed Schonberg [Tue, 11 Dec 2018 11:12:21 +0000 (11:12 +0000)]
[Ada] Unnesting: do not eliminate subprogram whose address is taken

2018-12-11  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

* exp_unst.adb (Register_Subprogram): A subprogram whose address
is taken (through attribute Address or Access) is reachable and
cannot be eliminated, even if there is no explicit call to it:
the address may be used in an indirect call or in some
address-related operation.

From-SVN: r267014

5 years ago[Ada] Crash on misplaced First operation for GNAT iterable type
Ed Schonberg [Tue, 11 Dec 2018 11:12:16 +0000 (11:12 +0000)]
[Ada] Crash on misplaced First operation for GNAT iterable type

This patch improves the handling of an improper declaaration of aspect
First for a GNAT-defined iterable type,

2018-12-11  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

* sem_util.adb (Get_Actual_Subtype): Function can return type
mark.
(Get_Cursor_Type): Improve recovery and error message on a
misplaced First aspect for an iterable type.

gcc/testsuite/

* gnat.dg/iter4.adb: New testcase.

From-SVN: r267013

5 years ago[Ada] Volatility, validity checks, and System.Aux_DEC
Hristian Kirtchev [Tue, 11 Dec 2018 11:12:11 +0000 (11:12 +0000)]
[Ada] Volatility, validity checks, and System.Aux_DEC

This patch updates validity checks to prevent the validation of an
by-reference formal parameter because the parameter is not being read in
the process.

2018-12-11  Hristian Kirtchev  <kirtchev@adacore.com>

gcc/ada/

* checks.adb: Add with and use clauses for Sem_Mech.
(Ensure_Valid): Update the "annoying special case" to include
entry and function calls. Use Get_Called_Entity to obtain the
entry or subprogram being invoked, rather than retrieving it
manually. Parameters passed by reference do not need a validity
check.

gcc/testsuite/

* gnat.dg/valid4.adb, gnat.dg/valid4_pkg.adb,
gnat.dg/valid4_pkg.ads: New testcase.

From-SVN: r267012

5 years ago[Ada] Improve error message when named number passed as global item
Yannick Moy [Tue, 11 Dec 2018 11:12:03 +0000 (11:12 +0000)]
[Ada] Improve error message when named number passed as global item

When a named number is used in Global/Depends contracts as global item,
the error message could be confusing to users new to Ada. Now the
message explains that named numbers are not objects.

For instance on the following illegal code snippet:

     $ gcc -c bad_global.ads

     1. package Bad_Global is
     2.
     3.    X : constant := 1;
     4.    Y : constant := 1.0;
     5.
     6.    procedure P with
     7.      Global => (Input => X,
                                 |
        >>> global item must denote object, state or current instance
            of concurrent type
        >>> named number "X" is not an object

     8.                 In_Out => Y);
                                  |
        >>> global item must denote object, state or current instance
            of concurrent type
        >>> named number "Y" is not an object

     9.
    10. end Bad_Global;

2018-12-11  Yannick Moy  <moy@adacore.com>

gcc/ada/

* sem_prag.adb (Analyze_Global_Item): Refine error message.

From-SVN: r267011

5 years ago[Ada] Unnesting: fix a missing activation record
Ed Schonberg [Tue, 11 Dec 2018 11:11:58 +0000 (11:11 +0000)]
[Ada] Unnesting: fix a missing activation record

2018-12-11  Ed Schonberg  <schonberg@adacore.com>

gcc/ada

* exp_ch7.adb (Check_Unnesting_In_Declarations): Extend
subprogram so that it is usable for visible and private
declarations of a package declaration, not just for declarations
in the pakage body.
* exp_ch13.adb (Expand_Freeze_Entity): Handle properly the
freezing of a finalizer routine generated for a controlled objet
declaration. Special processing already applies to finalizers
because they are usually displaced into another construct.

From-SVN: r267010

5 years ago[Ada] Unnesting: ensure Current_Subprogram is always initialized
Arnaud Charlet [Tue, 11 Dec 2018 11:11:53 +0000 (11:11 +0000)]
[Ada] Unnesting: ensure Current_Subprogram is always initialized

2018-12-11  Arnaud Charlet  <charlet@adacore.com>

gcc/ada/

* exp_unst.adb (Unnest_Subprogram): Ensure Current_Subprogram is
always initialized.

From-SVN: r267009

5 years ago[Ada] Fix -gnatR3 output for dynamically constrained record
Eric Botcazou [Tue, 11 Dec 2018 11:11:47 +0000 (11:11 +0000)]
[Ada] Fix -gnatR3 output for dynamically constrained record

2018-12-11  Eric Botcazou  <ebotcazou@adacore.com>

gcc/ada/

* gcc-interface/decl.c (gnat_to_gnu_entity): Add
gnat_annotate_type local variable initialized to Empty.
<E_Record_Subtype>: Set it to the Cloned_Subtype, if any.  For
types, back-annotate alignment and size values earlier and only
if the DECL was created here; otherwise, if gnat_annotate_type
is present, take the values from it.
(gnat_to_gnu_field): Add gnat_clause local variable.  If a
component clause is present, call validate_size only once on the
Esize of the component.  Otherwise, in the packed case, do not
call validate_size again on the type of the component but
retrieve directly its RM size.
(components_to_record): Minor tweak.
(set_rm_size): Remove useless test.
* gcc-interface/trans.c (gnat_to_gnu): Do wrap the instance of a
boolean discriminant attached to a variant part.

From-SVN: r267008

5 years ago[Ada] Fix setting of Has_Predicate flag for aggregate subtypes
Ed Schonberg [Tue, 11 Dec 2018 11:11:42 +0000 (11:11 +0000)]
[Ada] Fix setting of Has_Predicate flag for aggregate subtypes

This patch enures that the subtype of an aggregate has the Has_Predicate
flag properly set if the array component has a predicate, including the
case the predicate function for the component has not been constructed
yet.

2018-12-11  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

* sem_aggr.adb (Array_Aggr_Subtype. Resolve_Aggr_Expr): Indicate
that aggregate subtype has a predicate if the component type has
a predicate; do not rely on exisatence of predicate function for
component, in case component is a type no yet frozen, for which
predicate function has not been created yet.

From-SVN: r267007

5 years ago[Ada] Plug small loophole with pathological packed array type
Eric Botcazou [Tue, 11 Dec 2018 11:11:37 +0000 (11:11 +0000)]
[Ada] Plug small loophole with pathological packed array type

This fixes a crash in gigi on a pathological packed array type, whose
component type is a record type without representation clause or packing
but with a clause that bumps its size to a non-multiple value of the
storage unit.  In this case, the front-end fails to detect that calls
to the packing manpulation routines of the run time are necessary.

The fix doesn't change anything for non-pathological cases, i.e. when
the component type has a representation clause or is packed.

2018-12-11  Eric Botcazou  <ebotcazou@adacore.com>

gcc/ada/

* exp_aggr.adb (Packed_Array_Aggregate_Handled): Bail out for
any non-scalar type as component type of the array.

gcc/testsuite/

* gnat.dg/packed_array.adb, gnat.dg/packed_array.ads,
gnat.dg/packed_array_pkg.ads: New testcase.

From-SVN: r267006

5 years ago[Ada] Fix elaboration of record with 1-element dynamic array
Eric Botcazou [Tue, 11 Dec 2018 11:11:31 +0000 (11:11 +0000)]
[Ada] Fix elaboration of record with 1-element dynamic array

2018-12-11  Eric Botcazou  <ebotcazou@adacore.com>

gcc/ada/

* einfo.ads (Is_Bit_Packed_Array): Fix values of component size.
(Is_Packed): Likewise.
* gcc-interface/utils.c (convert): Do not extract the value of a
justified modular type if the destination type is a regular
array.
* gcc-interface/decl.c (gnat_to_gnu_component_type): Pass
TYPE_DECL to validate_size wherever the array has packed
components.

From-SVN: r267005

5 years ago[Ada] gnatbind: ghost code with -gnatQ
Bob Duff [Tue, 11 Dec 2018 11:11:26 +0000 (11:11 +0000)]
[Ada] gnatbind: ghost code with -gnatQ

This patch fixes a bug where if a library unit is compiled with -gnatQ,
and that library unit is an ignored Ghost unit, then gnatbind silently
fails.

2018-12-11  Bob Duff  <duff@adacore.com>

gcc/ada/

* gnat1drv.adb (gnat1drv): Pass the correct Object value when
calling Write_ALI in the case of -gnatQ.
* gnatbind.adb (Gnatbind): Avoid silent failure; give an error
message.

From-SVN: r267004

5 years ago[Ada] Unnesting: fix for constrained arrays and improve static constants
Ed Schonberg [Tue, 11 Dec 2018 11:11:22 +0000 (11:11 +0000)]
[Ada] Unnesting: fix for constrained arrays and improve static constants

2018-12-11  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

* exp_unst.adb (Needs_Fat_Pointer): A fat pointer is required if
the entity has a private type whose full view is an
unconstrained array type.
(Rewrite_One_Ref): If the reference is to a static constant, use
its value rather than create a reference through the activation
record. This is more efficient, and furthermore indispensable if
the context requires a static constant, such as in a branch of a
case statement.

From-SVN: r267003

5 years ago[Ada] Missing predicate check on declaration with aggregate expression
Ed Schonberg [Tue, 11 Dec 2018 11:11:17 +0000 (11:11 +0000)]
[Ada] Missing predicate check on declaration with aggregate expression

This patch adds a missing predicate check on the initial value of an
object whose declaration initializes the object with an aggregate. Such
a declaration is marked No_Initialization to prevent a call to a default
initialization procedure, but the check is needed on the value of the
aggregate.

Executing the following:

   gnatmake -q -gnata pred
   ./pred

must yield:

   raised SYSTEM.ASSERTIONS.ASSERT_FAILURE :
             Dynamic_Predicate failed at root.ads:30

----
with Root;

procedure Pred is
begin
   null;
end Pred;
----
package Root with SPARK_Mode is
   type Index_Type is range 1 .. Natural'Last;

   type Element_Type is private;
   Null_Element : constant Element_Type;

   type Foobar_Type is array (Index_Type range <>) of Element_Type
     with Dynamic_Predicate =>
            Foobar_Type'First > 0 and Foobar_Type'Length > 0;

   Null_Foobar : constant Foobar_Type;

private
   type Kind_Type is (Kind_Invalid, Kind_Valid);

   type Element_Type (Kind : Kind_Type := Kind_Invalid) is record
      Index1 : Index_Type;

      case Kind is
         when Kind_Valid =>
            Index2 : Index_Type;
         when Kind_Invalid =>
            null;
      end case;
   end record;

   Null_Element : constant Element_Type := (Kind   => Kind_Invalid,
                                            Index1 => Index_Type'First);

   Null_Foobar : constant Foobar_Type := (1 .. 0 => Null_Element);
end Root;

2018-12-11  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

* sem_ch3.adb (Analyze_Object_Declaration): Apply
Dynamic_Predicate check to an object of an array type
initialized with an aggregate.

From-SVN: r267002

5 years ago[Ada] Crash on generic instantiation in ignored Ghost context
Hristian Kirtchev [Tue, 11 Dec 2018 11:11:11 +0000 (11:11 +0000)]
[Ada] Crash on generic instantiation in ignored Ghost context

The following patch corrects the freezing of entities to properly
preserve all freeze nodes in case of recursive freezing when the context
is ignored Ghost, and the construct frozen is non-Ghost.

2018-12-11  Hristian Kirtchev  <kirtchev@adacore.com>

gcc/ada/

* freeze.adb (Add_To_Result): Move the ignored Ghost-specific
handling of freeze nodes to...
(Freeze_Entity): ...here. This ensures that the freeze nodes of
constructs that have recursive freezing are preserved when the
context is ignored Ghost, and the top level construct being
frozen is non-Ghost.

gcc/testsuite/

* gnat.dg/ghost3.adb, gnat.dg/ghost3.ads: New testcase.

From-SVN: r267001

5 years ago[Ada] Uintp: add a new UI_From_Integral generic constructor
Ed Schonberg [Tue, 11 Dec 2018 11:11:06 +0000 (11:11 +0000)]
[Ada] Uintp: add a new UI_From_Integral generic constructor

2018-12-11  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

* uintp.ads, uintp.adb (UI_From_Integral): New generic function,
to simplify construction of Universal_Integer representations
from any Integer type. If type is small enough the code is
equivalent to a call to UI_To_Int with appropriate conversions;
otherwise the routine uses the existing mechanism of building a
vector of suitable integer values and calling Vector_To_Uint.
The routine must not be applied to a biased type.

From-SVN: r267000

5 years ago[Ada] Crash on compilation unit function that builds in place
Ed Schonberg [Tue, 11 Dec 2018 11:11:00 +0000 (11:11 +0000)]
[Ada] Crash on compilation unit function that builds in place

This patch fixes a crash on a function that builds its limited result in
place. Previously this was handled properly only if the function was a
child unit.

2018-12-11  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

* sem_ch3.adb (Build_Itype_Reference): Handle properly an itype
reference created for a function that is a compilation unit, for
example if the function builds in place an object of a limited
type.

gcc/testsuite/

* gnat.dg/bip_cu.adb, gnat.dg/bip_cu_constructor.adb,
gnat.dg/bip_cu_constructor.ads, gnat.dg/bip_cu_t.adb,
gnat.dg/bip_cu_t.ads: New testcase.

From-SVN: r266999

5 years ago[Ada] GNAT.Sockets: fix timeout computations for sockets
Dmitriy Anisimkov [Tue, 11 Dec 2018 11:10:53 +0000 (11:10 +0000)]
[Ada] GNAT.Sockets: fix timeout computations for sockets

2018-12-11  Dmitriy Anisimkov  <anisimko@adacore.com>

gcc/ada/

* libgnat/g-socket.ads, libgnat/g-socket.adb: Fix duration
computations to be compatible with the type for socket timeouts
on Windows.

From-SVN: r266998

5 years ago[Ada] Minor reformattings
Gary Dismukes [Tue, 11 Dec 2018 11:10:48 +0000 (11:10 +0000)]
[Ada] Minor reformattings

2018-12-11  Gary Dismukes  <dismukes@adacore.com>

gcc/ada/

* exp_util.ads: Use preferred U.S. spelling of "honored".
* sem_spark.adb: Remove extra whitespace.

From-SVN: r266997

5 years ago[Ada] Crash on ignored Ghost expression function
Hristian Kirtchev [Tue, 11 Dec 2018 11:10:42 +0000 (11:10 +0000)]
[Ada] Crash on ignored Ghost expression function

This patch updates freezing to ensure that freeze nodes are inserted
into the tree when the entity being frozen is non-Ghost, and the context
is an ignored Ghost spec expression.

2018-12-11  Hristian Kirtchev  <kirtchev@adacore.com>

gcc/ada/

* exp_util.adb (Insert_Action): Add new formal parameter
Spec_Expr_OK.
(Insert_Actions): Add new formal parameter Spec_Expr_OK. Update
all calls to Insert_Actions where relevant. Honour an insertion
from a spec expression context when requested by the caller.
* exp_util.ads (Insert_Action): Add new formal parameter
Spec_Expr_OK.
(Insert_Actions): Add new formal parameter Spec_Expr_OK.
* freeze.adb (Add_To_Result): Force the insertion of the freeze
node even when the context is a spec expression.

gcc/testsuite/

* gnat.dg/ghost2.adb, gnat.dg/ghost2.ads: New testcase.

From-SVN: r266996

5 years ago[Ada] Remove vxlink and vxaddr2line from this repository
Jerome Lambourg [Tue, 11 Dec 2018 11:10:37 +0000 (11:10 +0000)]
[Ada] Remove vxlink and vxaddr2line from this repository

Those tools need a dedicated repository as they're VxWorks specific and
not related with the Ada front-end.

2018-12-11  Jerome Lambourg  <lambourg@adacore.com>

gcc/ada/

* vxaddr2line.adb, vxlink-bind.adb, vxlink-bind.ads,
vxlink-link.adb, vxlink-link.ads, vxlink-main.adb, vxlink.adb,
vxlink.ads: Remove.
* gcc-interface/Make-lang.in, gcc-interface/Makefile.in: Remove
bits for vxaddr2line.

From-SVN: r266995

5 years ago[Ada] Minor reformatting
Hristian Kirtchev [Tue, 11 Dec 2018 11:10:32 +0000 (11:10 +0000)]
[Ada] Minor reformatting

2018-12-11  Hristian Kirtchev  <kirtchev@adacore.com>

gcc/ada/

* exp_aggr.adb, exp_ch7.adb, gnat1drv.adb, sem_ch10.adb,
sem_ch13.adb, sem_ch6.adb, sem_ch7.adb, sem_util.adb: Minor
reformatting.

From-SVN: r266994

5 years ago[Ada] Complete implementation of RM C.6(19) clause
Eric Botcazou [Tue, 11 Dec 2018 11:10:27 +0000 (11:10 +0000)]
[Ada] Complete implementation of RM C.6(19) clause

This ensures that the compiler fully implements the C.6(19) clause of
the Ada Reference Manual and gives a warning when the clause does change
the passing mechanism of the affected parameter.

2018-12-11  Eric Botcazou  <ebotcazou@adacore.com>

gcc/ada/

* fe.h (Is_Atomic_Object): Declare.
(Is_Volatile_Object): Likewise.
* gcc-interface/trans.c (atomic_or_volatile_copy_required_p):
New.
(Call_to_gnu): Generate a copy for an actual parameter passed by
reference if the conditions set forth by RM C.6(19) are met and
specificially deal with an atomic actual parameter.

gcc/testsuite/

* gnat.dg/atomic11.adb, gnat.dg/atomic11_pkg1.ads,
gnat.dg/atomic11_pkg2.ads: New testcase.

From-SVN: r266993

5 years ago[Ada] Stubs that complete generic subprogram do have a "prior declaration"
Piotr Trojanek [Tue, 11 Dec 2018 11:10:22 +0000 (11:10 +0000)]
[Ada] Stubs that complete generic subprogram do have a "prior declaration"

The intuition behind the Is_Subprogram_Stub_Without_Prior_Declaration
utility routine is to detect stubs that act as subprogram declarations
and False on stubs that act as completions. This behaviour is now fixed
for stubs that correspond to generic subprogram declarations.

This patch affects a routine that is only used in GNATprove, so no
frontend test provided. An example where the result changed from True to
False is:

-----------
-- p.ads --
-----------

package P is
   generic
   procedure Proc;
end P;

-----------
-- p.adb --
-----------

package body P is
   procedure Proc is separate; -- now we return False for this stub
end P;

----------------
-- p-proc.adb --
----------------

separate (P)
procedure Proc is
begin
   null;
end;

2018-12-11  Piotr Trojanek  <trojanek@adacore.com>

gcc/ada/

* sem_util.adb (Is_Subprogram_Stub_Without_Prior_Declaration):
Return False on stubs that complete a generic subprogram.
* sem_util.ads: Update corresponding comment.

From-SVN: r266992

5 years ago[Ada] GNATprove: improve proofs for uninitialized constrained objects
Ed Schonberg [Tue, 11 Dec 2018 11:10:17 +0000 (11:10 +0000)]
[Ada] GNATprove: improve proofs for uninitialized constrained objects

2018-12-11  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

* sem_ch4.adb (Analyze_Allocator): In GNATprove mode build a
subtype declaration if the allocator has a subtype indication
with a constraint. This allows additional proofs to be applied
to allocators that designate uninitialized constrained objects.

From-SVN: r266991

5 years ago[Ada] Support access types in GNATprove
Yannick Moy [Tue, 11 Dec 2018 11:10:12 +0000 (11:10 +0000)]
[Ada] Support access types in GNATprove

SPARK RM has been updated to support access types in SPARK. Part of this
support is that now SPARK RM 3.1 lists access types as having full
default initialization. Now updated.

There is no impact on compilation.

2018-12-11  Yannick Moy  <moy@adacore.com>

gcc/ada/

* sem_util.adb (Has_Full_Default_Initialization): Consider
access types as having full default initialization.

From-SVN: r266990

5 years ago[Ada] Better error message from GNATprove on illegal switch
Yannick Moy [Tue, 11 Dec 2018 11:10:07 +0000 (11:10 +0000)]
[Ada] Better error message from GNATprove on illegal switch

When a compilation switch is wrongly passed to GNATprove without the
leading hyphen, this patch issues a clear error message instead of the
obscure 'usage' message previously displayed.

There is no impact on compilation.

2018-12-11  Yannick Moy  <moy@adacore.com>

gcc/ada/

* gnat1drv.adb (Gnat1drv): Issue specific error message in
GNATprove mode when multiple file names on the command line.
* osint.adb, osint.ads (Dump_Command_Line_Source_File_Names):
New procedure to print file names on the command line.

From-SVN: r266989

5 years ago[Ada] Fix runtime stack overflow for out/in-out actuals without discr.
Javier Miranda [Tue, 11 Dec 2018 11:10:02 +0000 (11:10 +0000)]
[Ada] Fix runtime stack overflow for out/in-out actuals without discr.

2018-12-11  Javier Miranda  <miranda@adacore.com>

gcc/ada/

* exp_aggr.adb (Convert_To_Assignments): When gnerating C, do
not declare a temporary to initialize an aggregate assigned to
Out or In_Out parameters whose type has no discriminants. This
avoids stack overflow errors at runtime.

From-SVN: r266988

5 years ago[Ada] Additionsal trasformations for unnesting in package bodies
Ed Schonberg [Tue, 11 Dec 2018 11:09:57 +0000 (11:09 +0000)]
[Ada] Additionsal trasformations for unnesting in package bodies

This patch extends the previous algorithm for creating an explicit
elaboration procedure for a package body when expansion generates
subprograms in the statement part of the body. For unnesting to work
properly, these subprograms must appear within an explicit subprogram
body so that uplevel references can be placed in the proper activation
record.

Ongoing work for LLVM generation.

2018-12-11  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

* exp_ch7.adb (Check_Unnesting_Elaboration_Code): Extend
algorithm to cover subprograms generated in nested loops and in
exception handlers, in order to build an explicit elaboration
procedure in more complex cases.

From-SVN: r266987

5 years ago[Ada] Remove vxworks*-crtbe-link.spec
Jerome Lambourg [Tue, 11 Dec 2018 11:09:52 +0000 (11:09 +0000)]
[Ada] Remove vxworks*-crtbe-link.spec

The VxWorks crtbe are now part of libgcc, needed to support C++ on this
target.

2018-12-11  Jerome Lambourg  <lambourg@adacore.com>

gcc/ada/

* Makefile.rtl, gcc-interface/Makefile.in: Remove crtbe bits for
VxWorks.
* libgnat/system-vxworks-arm-rtp-smp.ads,
libgnat/system-vxworks-arm-rtp.ads,
libgnat/system-vxworks-arm.ads,
libgnat/system-vxworks-e500-kernel.ads,
libgnat/system-vxworks-e500-rtp-smp.ads,
libgnat/system-vxworks-e500-rtp.ads,
libgnat/system-vxworks-ppc-kernel.ads,
libgnat/system-vxworks-ppc-rtp-smp.ads,
libgnat/system-vxworks-ppc-rtp.ads,
libgnat/system-vxworks-ppc.ads,
libgnat/system-vxworks-x86-kernel.ads,
libgnat/system-vxworks-x86-rtp-smp.ads,
libgnat/system-vxworks-x86-rtp.ads,
libgnat/system-vxworks7-aarch64-rtp-smp.ads,
libgnat/system-vxworks7-aarch64.ads,
libgnat/system-vxworks7-arm-rtp-smp.ads,
libgnat/system-vxworks7-e500-kernel.ads,
libgnat/system-vxworks7-e500-rtp-smp.ads,
libgnat/system-vxworks7-e500-rtp.ads,
libgnat/system-vxworks7-ppc-kernel.ads,
libgnat/system-vxworks7-ppc-rtp-smp.ads,
libgnat/system-vxworks7-ppc-rtp.ads,
libgnat/system-vxworks7-ppc64-kernel.ads,
libgnat/system-vxworks7-ppc64-rtp-smp.ads,
libgnat/system-vxworks7-x86-kernel.ads,
libgnat/system-vxworks7-x86-rtp-smp.ads,
libgnat/system-vxworks7-x86-rtp.ads,
libgnat/system-vxworks7-x86_64-kernel.ads,
libgnat/system-vxworks7-x86_64-rtp-smp.ads: Remove pragma
Linker_Options for --specs.
* vx_crtbegin.c, vx_crtbegin.inc, vx_crtbegin_array.c,
vx_crtbegin_attr.c, vx_crtend.c, vxworks-gnat-crtbe-link.spec:
Remove.

From-SVN: r266986

5 years ago[Ada] Spurious visibility error on aspect Predicate
Ed Schonberg [Tue, 11 Dec 2018 11:09:46 +0000 (11:09 +0000)]
[Ada] Spurious visibility error on aspect Predicate

The GNAT-defined aspect Predicate has the same semantics as the Ada
aspect Dynamic_Predicate, including direct visibility to the components
of a record type to which the aspect applies.

The following must compile quietly:

   gcc -c integer_stacks.ads

----
pragma SPARK_Mode (On);

with Bounded_Stacks;
package Integer_Stacks is
     new Bounded_Stacks (Element => Integer, Default_Element => 0);
----
generic
   type Element is private;
   Default_Element : Element;
package Bounded_Stacks is

   type Stack (Capacity : Positive) is tagged private
      with Default_Initial_Condition => Empty (Stack);

   function "=" (Left, Right : Stack) return Boolean;

   function Extent (This : Stack) return Natural;

   function Empty (This : Stack) return Boolean;

   function Full (This : Stack) return Boolean;

   procedure Reset (This : out Stack) with
     Post'Class => Empty (This) and
                   not Full (This),
     Global     => null,
     Depends    => (This =>+ null);

   procedure Push (This : in out Stack;  Item : Element) with
     Pre'Class  => not Full (This) and
                   Extent (This) < This.Capacity,
     Post'Class => not Empty (This) and
                   Extent (This) = Extent (This'Old) + 1,
     Global     => null,
     Depends    => (This =>+ Item);

   procedure Pop (This : in out Stack;  Item : out Element) with
     Pre'Class  => not Empty (This),
     Post'Class => not Full (This) and
                   Extent (This) = Extent (This'Old) - 1,
     Global     => null,
     Depends    => (This =>+ null, Item => This);

   function Peek (This : Stack) return Element with
     Pre'Class => not Empty (This),
     Global    => null,
     Depends   => (Peek'Result => This);

private

   type Contents is array (Positive range <>) of Element;

   type Stack (Capacity : Positive) is tagged record
      Values : Contents (1 .. Capacity); -- := (others => Default_Element);
--        Top    : Natural;
      Top    : Natural := 0;
   end record with Predicate => Top <= Capacity,
     Annotate => (GNATprove,
                  Intentional,
                  "type ""Stack"" is not fully initialized",
                  "Because zeroing Top is sufficient");

end Bounded_Stacks;
----
package body Bounded_Stacks is

   ------------
   -- Extent --
   ------------

   function Extent (This : Stack) return Natural is
     (This.Top);

   -----------
   -- Empty --
   -----------

   function Empty (This : Stack) return Boolean is
     (This.Top = 0);

   ----------
   -- Full --
   ----------

   function Full (This : Stack) return Boolean is
     (This.Top = This.Capacity);

   -----------
   -- Reset --
   -----------

   procedure Reset (This : out Stack) is
   begin
      This := (This.Capacity, Top => 0, others => <>);
--        This.Top := 0;
   end Reset;

   ----------
   -- Push --
   ----------

   procedure Push (This : in out Stack;  Item : Element) is
   begin
      This.Top := This.Top + 1;
      This.Values (This.Top) := Item;
   end Push;

   ---------
   -- Pop --
   ---------

   procedure Pop (This : in out Stack;  Item : out Element) is
   begin
      Item := This.Values (This.Top);
      This.Top := This.Top - 1;
   end Pop;

   ----------
   -- Peek --
   ----------

   function Peek (This : Stack) return Element is
     (This.Values (This.Top));

   ---------
   -- "=" --
   ---------

   function "=" (Left, Right : Stack) return Boolean is
   begin
      if Left.Top /= Right.Top then
         return False;
      else
         for K in 1 .. Left.Top loop
            if Left.Values (K) /= Right.Values (K) then
               return False;
            end if;
         end loop;
         return True;
      end if;
   end "=";

end Bounded_Stacks;
----

2018-12-11  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

* sem_ch13.adb (Check_Aspect_At_End_Of_Declarations,
Freeze_Entity_Checks): Process aspect Predicate in the same
fashion as aspect Dynamic_Predicate.

From-SVN: r266985

5 years ago[Ada] Fix an ICE on instantiated subprogram with -gnatc
Eric Botcazou [Tue, 11 Dec 2018 11:09:41 +0000 (11:09 +0000)]
[Ada] Fix an ICE on instantiated subprogram with -gnatc

The following should compile quietly:

$ gcc -c p-proc.ads -gnatc

procedure P.Proc is new G;

with Q;

package P is

  generic procedure G;

end P;

with System;
with Unchecked_Conversion;

package Q is

  generic package Inner_G is

    type T is access all Integer;

    function Cnv is new Unchecked_Conversion (System.Address, T);

  end Inner_G;

end Q;

2018-12-11  Eric Botcazou  <ebotcazou@adacore.com>

gcc/ada/

* gcc-interface/trans.c (elaborate_all_entities_for_package):
Skip aliasing subprograms.

From-SVN: r266984

5 years ago[Ada] Update gnatmetric documentation for the Libadalang-based version
Bob Duff [Tue, 11 Dec 2018 11:09:36 +0000 (11:09 +0000)]
[Ada] Update gnatmetric documentation for the Libadalang-based version

2018-12-11  Bob Duff  <duff@adacore.com>

gcc/ada/

* doc/gnat_ugn/gnat_utility_programs.rst: Update gnatmetric
documentation for the Libadalang-based version.

From-SVN: r266983

5 years ago[Ada] Fix an ACATS regression with CCG
Javier Miranda [Tue, 11 Dec 2018 11:09:29 +0000 (11:09 +0000)]
[Ada] Fix an ACATS regression with CCG

2018-12-11  Javier Miranda  <miranda@adacore.com>

gcc/ada/

* exp_aggr.adb (In_Object_Declaration): Removed.
(Is_CCG_Supported_Aggregate): New subprogram that replaces
In_Object_Declaration extending its functionality to indicate if
an aggregate is in a context supported by the CCG backend.

From-SVN: r266982

5 years ago[Ada] Update gnatstub documentation for the Libadalang-based version
Bob Duff [Tue, 11 Dec 2018 11:09:24 +0000 (11:09 +0000)]
[Ada] Update gnatstub documentation for the Libadalang-based version

2018-12-11  Bob Duff  <duff@adacore.com>

gcc/ada/

* doc/gnat_ugn/gnat_utility_programs.rst: Update gnatstub
documentation for the Libadalang-based version.
* gnat_ugn.texi: Regenerate.

From-SVN: r266981

5 years ago[Ada] Spurious errors on aspect specifications in generic units
Ed Schonberg [Tue, 11 Dec 2018 11:09:19 +0000 (11:09 +0000)]
[Ada] Spurious errors on aspect specifications in generic units

This patch fixes spurious errors on aspect specifications on record
types when the aspect expression references a component of the type that
is not a discriminant. The patch also cleans up the legality checks on
aspect specifications, and improves error message on illegal aspect
specifications whose expressions are not conformant between
specification and freeze point, because of changes in visibility.

2018-12-11  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

* sem_ch13.adb (Push_Type, Pop_Type): New procedures, used for
analysis of aspect expressions for record types, whose
components (not only discriminants) may be referenced in aspect
expressions.
(Analyze_Aspect_Specifications, Analyze_Aspects_At_Freeze_Point,
Analyze_Aspect_At_End-Of_Declarations,
Resolve_Aspect_Expressions): Use the new subprograms.
(Check_Aspect_At_End_Of_Declarations): Improve error message.
(Build_Predicate_Functions): Do not build their bodies in a
generic unit.
(Is_Derived_Type_With_Constraint): New subprogram to uncover and
reject aspect specificationss on types that appear after the
type is frozen.
* sem_ch13.ads (Push_Scope_And_Install_Discriminants,
Uninstall_Discriminants_And_Pop_Scope): Remove.
* sem_ch6.adb, sem_ch6.ads (Fully_Conformant_Expressions):
Additional parameter to improve error message on illegal aspect
specifications whose resolution differ between aspect
specification and freeze point.
* freeze.adb: Remove references to
Install/Uninstall_Discriminants.

gcc/testsuite/

* gnat.dg/aspect1.adb, gnat.dg/aspect1_horizontal.adb,
gnat.dg/aspect1_horizontal.ads, gnat.dg/aspect1_vectors_2d.ads:
New testcase.
* gnat.dg/static_pred1.adb: Expect an error message.

From-SVN: r266980

5 years ago[Ada] Document the -fgnat-encodings switch
Pierre-Marie de Rodat [Tue, 11 Dec 2018 11:09:13 +0000 (11:09 +0000)]
[Ada] Document the -fgnat-encodings switch

2018-12-11  Pierre-Marie de Rodat  <derodat@adacore.com>

gcc/ada/

* doc/gnat_ugn/building_executable_programs_with_gnat.rst:
Document the -fgnat-encodings switch.
* gnat_ugn.texi: Regenerate.

From-SVN: r266979

5 years ago[Ada] Crash on nesting of subunits with bodies acting as specs
Justin Squirek [Tue, 11 Dec 2018 11:09:08 +0000 (11:09 +0000)]
[Ada] Crash on nesting of subunits with bodies acting as specs

This patch corrects an issue whereby a set of nested subunits including
subprogram subunits acting as bodies would cause a crash when a child
subunit "withs" an ansestor in certain instances due to a mismanagment
of the scope stack.

------------
-- Source --
------------

--  w.ads

package W is
end;

--  w-b.ads

package W.B is
  pragma Elaborate_Body;
end;

--  w-b.adb

with W.B.C;

package body w.B is end;

--  w-b-c.adb

with W;
procedure W.B.C is
  package D is
    procedure E;
  end;
  package body D is separate;
begin
  null;
end;

--  w-b-c-d.adb

separate (W.B.C)
package body D is
  procedure E is separate;
end;

--  w-b-c-d-e.adb

with W;

separate (W.B.C.D)
procedure E is
begin
  null;
end;

-----------------
-- Compilation --
-----------------

$ gnatmake -q w-b.adb

2018-12-11  Justin Squirek  <squirek@adacore.com>

gcc/ada/

* sem_ch10.adb (Analyze_Subunit): Modify conditional to fully
remove parent contexts from library-level subprogram bodies in
addition to package bodies.

From-SVN: r266978

5 years ago[Ada] Suppress call to Initial_Condition when the annotation is ignored
Hristian Kirtchev [Tue, 11 Dec 2018 11:09:04 +0000 (11:09 +0000)]
[Ada] Suppress call to Initial_Condition when the annotation is ignored

This patch suppresses the generation of the Initial_Condition procedure
tasked with verifying the run-time semantics of the pragma when the
pragma is ignored by means of -gnata, pragma Assertion_Policy, etc.

------------
-- Source --
------------

--  all_asserts_off.adc

pragma Assertion_Policy (Ignore);

--  all_asserts_on.adc

pragma Assertion_Policy (Check);

--  ic_off.adc

pragma Assertion_Policy (Initial_Condition => Ignore);

--  ic_on.adc

pragma Assertion_Policy (Initial_Condition => Check);

--  init_cond.ads

package Init_Cond
  with SPARK_Mode,
       Initial_Condition => Flag = False
is
   Flag : Boolean := True;

   procedure Set_Flag;
end Init_Cond;

--  init_cond.adb

package body Init_Cond
  with SPARK_Mode
is
   procedure Set_Flag is
   begin
      Flag := True;
   end Set_Flag;
end Init_Cond;

----------------------------
-- Compilation and output --
----------------------------

& gcc  -c -S -gnatDG init_cond.adb -gnatec=all_asserts_on.adc
& grep -c "Initial_Condition;" init_cond.adb.dg
& grep -c "_elabb" init_cond.s
& gcc  -c -S -gnatDG init_cond.adb -gnatec=ic_on.adc
& grep -c "Initial_Condition;" init_cond.adb.dg
& grep -c "_elabb" init_cond.s
& gcc  -c -S -gnatDG init_cond.adb -gnatec=all_asserts_off.adc
& grep -c "Initial_Condition;" init_cond.adb.dg
& grep -c "_elabb" init_cond.s
& gcc  -c -S -gnatDG init_cond.adb -gnatec=ic_off.adc
& grep -c "Initial_Condition;" init_cond.adb.dg
& grep -c "_elabb" init_cond.s
2
4
2
4
0
0
0
0

2018-12-11  Hristian Kirtchev  <kirtchev@adacore.com>

gcc/ada/

* exp_prag.adb (Expand_Pragma_Initial_Condition): Do not
generate an Initial_Condition procedure and a call to it when
the associated pragma is ignored.
* sem_ch10.adb (Analyze_Compilation_Unit): Minor cleanup.

From-SVN: r266977