For rs6000 target, it is not valid to make PC-relative
references to TLS symbols. So addr_to_insn_form() needs
to check if things are TLS before returning PC-rel forms.
2020-04-14 Aaron Sawdey <acsawdey@linux.ibm.com>
PR target/94542
* config/rs6000/rs6000.c (address_to_insn_form): Do not attempt to
use PC-relative addressing for TLS references.
This fixes a garbled concepts diagnostic by moving the handling of DECLTYPE_TYPE
from pp_cxx_type_specifier_seq to cxx_pretty_printer::simple_type_specifier, a
move which also seems to be more consistent with the language grammar.
This patch also fixes pretty printing of rvalue reference types via
cxx_pretty_printer::type_id, which eventually calls pp_c_pointer which currently
doesn't distinguish between lvalue and rvalue references.
gcc/c-family/ChangeLog:
PR c++/85278
* c-pretty-print.c (pp_c_pointer) <case REFERENCE_TYPE>: Print a double
ampersand if it's an rvalue reference type.
gcc/cp/ChangeLog:
PR c++/85278
* cxx-pretty-print.c (cxx_pretty_printer:simple_type_specifier)
<case DECLTYPE_TYPE>: Handle DECLTYPE_TYPE here instead of ...
(pp_cxx_type_specifier_seq) <case DECLTYPE_TYPE>: ... here.
(cxx_pretty_printer::direct_abstract_declarator) <case DECLTYPE_TYPE>:
New no-op case.
gcc/testsuite/ChangeLog:
PR c++/85278
* g++.dg/concepts/diagnostic9.C: New test.
IPA-SRA can segfault when processing a call to an internal function
because such calls do not have corresponding call graphs and should be
treated like memory accesses anyway, which what the patch below does.
The patch makes an attempt to differentiate between reads and writes,
although for things passed by value it does not make any difference.
It treats all arguments of functions denoted with internal_store_fn_p
as written to, even though in practice only some are, but for IPA-SRA
purposes, actions needed to be taken when processing a read are also
always performed when analyzing a write, so the code is just slightly
pessimistic. But not as pessimistic as bailing out on any internal
call immediately.
I have LTO bootstrapped and tested the patch on x86_64-linux and
normally bootstrapped and tested it on aarch64-linux, although one
which is not SVE capable. I would appreciate testing on such machine
too - as well as a small testcase that would follow all relevant
conventions in gcc.target/aarch64/sve.
2020-04-14 Martin Jambor <mjambor@suse.cz>
PR ipa/94434
* ipa-sra.c: Include internal-fn.h.
(enum isra_scan_context): Update comment.
(scan_function): Treat calls to internal_functions like loads or stores.
When evaluating the initializer of 'a' in the following example
struct A {
A() = default; A(const A&);
A *p = this;
};
constexpr A foo() { return {}; }
constexpr A a = foo();
the PLACEHOLDER_EXPR for 'this' in the aggregate initializer returned by foo
gets resolved to the RESULT_DECL of foo. But due to guaranteed RVO, the 'this'
should really be resolved to '&a'.
Fixing this properly by immediately resolving 'this' and PLACEHOLDER_EXPRs to
the ultimate object under construction would in general mean that we would no
longer be able to cache constexpr calls for which RVO possibly applies, because
the result of the call may now depend on the ultimate object under construction.
So as a mostly correct stopgap solution that retains cachability of RVO'd
constexpr calls, this patch fixes this issue by rewriting all occurrences of the
RESULT_DECL in the result of a constexpr function call with the current object
under construction, after the call returns. This means the 'this' pointer
during construction of the temporary will still point to the temporary object
instead of the ultimate object, but besides that this approach seems
functionally equivalent to the proper approach.
gcc/cp/ChangeLog:
PR c++/94034
* constexpr.c (replace_result_decl_data): New struct.
(replace_result_decl_data_r): New function.
(replace_result_decl): New function.
(cxx_eval_call_expression): Use it.
* tree.c (build_aggr_init_expr): Set the location of the AGGR_INIT_EXPR
to that of its initializer.
gcc/testsuite/ChangeLog:
PR c++/94034
* g++.dg/cpp0x/constexpr-empty15.C: New test.
* g++.dg/cpp1y/constexpr-nsdmi6a.C: New test.
* g++.dg/cpp1y/constexpr-nsdmi6b.C: New test.
* g++.dg/cpp1y/constexpr-nsdmi7a.C: New test.
* g++.dg/cpp1y/constexpr-nsdmi7b.C: New test.
In this PR the testcase ICEs because a BIT_INSERT_EXPR whose replaced bits are
not fully inside the container is generated. A size check is added to avoid
this kind of ICE.
gcc/ChangeLog:
PR tree-optimization/94574
* tree-ssa.c (non_rewritable_lvalue_p): Add size check when analyzing
whether a vector-insert is rewritable using a BIT_INSERT_EXPR.
gcc/testsuite/ChangeLog:
PR tree-optimization/94574
* gcc.dg/pr94574.c: New test.
Even though ix86_hard_regno_mode_ok doesn't allow xmm16-xmm31 nor
ymm16-ymm31 in 128/256 bit modes when AVX512VL is disabled, LRA can
still generate reg to reg moves with xmm16-xmm31 and ymm16-ymm31 in
128/256 bit modes. Remove mode size check in ix86_get_ssemov.
gcc/
PR target/94561
* config/i386/i386.c (ix86_get_ssemov): Remove mode size check.
gcc/testsuite/
PR target/94561
* gcc.target/i386/pr94561.c: New test.
* module.c (gfc_match_use): Fix name-conflict check for use-associating
the same symbol again in a submodule.
* gfortran.dg/use_rename_10.f90: New.
* gfortran.dg/use_rename_11.f90: New.
2020-04-14 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/94270
* interface.c (gfc_get_formal_from_actual_arglist): Always
set artificial attribute for symbols.
* trans-decl.c (generate_local_decl): Do not warn if the
symbol is artifical.
2020-04-14 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/94270
* gfortran.dg/warn_unused_dummy_argument_6.f90: New test.
While reviewing [basic.scope.param] I noticed we don't show the location
of the previous declaration when giving an error about "A parameter name
shall not be redeclared in the outermost block of the function definition".
PR c++/94588
* name-lookup.c (check_local_shadow): Add an inform call.
* g++.dg/diagnostic/redeclaration-1.C: Add dg-message.
gcc/ChangeLog:
* doc/extend.texi (-Wall): Mention -Wformat-overflow and
-Wformat-truncation. Move -Wzero-length-bounds last.
(-Wrestrict): Document positive form of option enabled by -Wall.
We are hitting a recursive loop when printing the signature of a function
containing a decltype([]{}) type. The loop is
dump_function_decl -> dump_substitution
-> dump_template_bindings
-> dump_type
-> dump_aggr_type
-> dump_scope -> dump_function_decl
and we loop because dump_template_bindings wants to print the resolved type of
decltype([]{}) (i.e. just a lambda type), so it calls dump_aggr_type, which
wants to print the function scope of the lambda type. But the function scope of
the lambda type is the function which we're in the middle of printing.
This patch breaks the loop by passing TFF_NO_FUNCTION_ARGUMENTS to
dump_function_decl from dump_scope, so that we avoid recursing into
dump_substitution and ultimately looping.
This also means we no longer emit the "[with ...]" clause when printing a
function template scope, and we instead just emit its template argument list in
a more natural way, e.g. instead of
foo(int, char) [with T=bool]::x
we would now print
foo<bool>::x
which seems like an improvement on its own.
The full signature of the function 'spam' in the below testcase is now
void spam(decltype (<lambda>)*) [with T = int; decltype (<lambda>) = spam<int>::<lambda()>]
gcc/cp/ChangeLog:
PR c++/94521
* error.c (dump_scope): Pass TFF_NO_FUNCTION_ARGUMENTS to
dump_function_decl when printing a function template instantiation as a
scope.
gcc/testsuite/ChangeLog:
PR c++/94521
* g++.dg/cpp2a/lambda-uneval12.C: New test.
In this PR we're incorrectly rejecting a self-modifying constexpr initializer as
a consequence of the fix for PR78572.
It looks like however that the fix for PR78572 is obsoleted by the fix for
PR89336: the testcase from the former PR successfully compiles even with its fix
reverted.
But then further testing showed that the analogous testcase of PR78572 where the
array has an aggregate element type is still problematic (i.e. we ICE) even with
the fix for PR78572 applied. The reason is that in cxx_eval_bare_aggregate we
attach a constructor_elt of aggregate type always to the end of the new
CONSTRUCTOR, but that's not necessarily correct if the CONSTRUCTOR is
self-modifying. We should instead be using get_or_insert_ctor_field to insert
the constructor_elt in the right place.
So this patch reverts the PR78572 fix and makes the appropriate changes to
cxx_eval_bare_aggregate. This fixes PR94470, and we now are also able to fully
reduce the initializers of 'arr' and 'arr2' in the new test array57.C to
constant initializers.
gcc/cp/ChangeLog:
PR c++/94470
* constexpr.c (get_or_insert_ctor_field): Set default value of parameter
'pos_hint' to -1.
(cxx_eval_bare_aggregate): Use get_or_insert_ctor_field instead of
assuming the the next index belongs at the end of the new CONSTRUCTOR.
(cxx_eval_store_expression): Revert PR c++/78572 fix.
gcc/testsuite/ChangeLog:
PR c++/94470
* g++.dg/cpp1y/constexpr-nsdmi8.C: New test.
* g++.dg/cpp1y/constexpr-nsdmi9.C: New test.
* g++.dg/init/array57.C: New test.
From XCode 11.4 on 10.14/15 use of 10.6 and 10.7 is deprecated.
The tools issue diagnostics if -mmacosx-version-min= < 10.8
Adjust the testcase to avoid that usage on 10.14, 10.15 for now.
gcc/testsuite/ChangeLog:
2020-04-13 Iain Sandoe <iain@sandoe.co.uk>
* gcc.dg/darwin-version-1.c: Use -mmacosx-version-min= 10.8
for system versions 10.14 and 10.15.
The idea is not have another resolution of a pointer if an error has
occurred previously.
2020-04-13 Linus Koenig <link@sig-st.de>
PR fortran/94192
* resolve.c (resolve_fl_var_and_proc): Set flag "error" to 1 if
pointer is found to not have an assumed rank or a deferred shape.
* simplify.c (simplify_bound): If an error has been issued for a
given pointer, one should not attempt to find its bounds.
2020-04-13 Linus Koenig <link@sig-st.de>
PR fortran/94192
* gfortran.dg/bound_resolve_after_error_1.f90: New test.
My fix for 94147 was confusing no-linkage with internal linkage, at
the language level. That's wrong. (the std is confusing here, because
it describes linkage of names (which is wrong), and lambdas have no
names)
Lambdas with extra-scope, have linkage. However, at the
implementation-level that linkage is at least as restricted as the
linkage of the extra-scope decl.
Further, when instantiating a variable initialized by a lambda, we
must determine the visibility of the variable itself, before
instantiating its initializer. If the template arguments are internal
(or no-linkage), the variable will have internal linkage, regardless
of the linkage of the template it is instantiated from. We need to
know that before instantiating the lambda, so we can restrict its
linkage correctly.
* decl2.c (determine_visibility): A lambda's visibility is
affected by its extra scope.
* pt.c (instantiate_decl): Determine var's visibility before
instantiating its initializer.
* tree.c (no_linkage_check): Revert code looking at visibility of
lambda's extra scope.
` gcc/cp/
* g++.dg/cpp0x/lambda/pr94426-[12].C: New.
* g++.dg/abi/lambda-vis.C: Drop a warning.
* g++.dg/cpp0x/lambda/lambda-mangle.C: Lambda visibility on
variable changes.
* g++.dg/opt/dump1.C: Drop warnings of no import.
We must restore the frame pointer in word_mode for eh_return epilogues
since the upper 32 bits of RBP register can have any values.
Tested on Linux/x32 and Linux/x86-64.
PR target/94556
* config/i386/i386.c (ix86_expand_epilogue): Restore the frame
pointer in word_mode for eh_return epilogues.
Some insns, which operate on SImode operands, output assembler template
that comprise of multiple instructions using HImode operands. To access
the high word of an SImode operand, an operand selector '%H' is used to
offset the operand value by a constant amount.
When one of these HImode operands is a memory reference to a post_inc,
the address does not need to be offset, since the preceding instruction
has already offset the address to the correct value.
This fixes an ICE in change_address_1, at emit-rtl.c:2318 for
gcc.c-torture/execute/pr20527-1.c in the "-mlarge -O2 -flto
-fuse-linker-plugin -fno-fat-lto-objects" configuration.
This test generated the following insn, and the attempt to output the
high part of the post_inc address caused the ICE.
(set (reg:SI 6 R6)
(minus:SI (reg:SI 6 R6)
(mem:SI (post_inc:PSI (reg:PSI 10 R10)) {subsi3}
gcc/ChangeLog:
2020-04-13 Jozef Lawrynowicz <jozef.l@mittosystems.com>
* config/msp430/msp430.c (msp430_print_operand): Don't add offsets to
memory references in %B, %C and %D operand selectors when the inner
operand is a post increment address.
The %C and %D operand modifiers are supposed to access the 3rd and 4th
words of a 64-bit value, so for memory references they need to offset
the given address by 4 and 6 bytes respectively.
gcc/ChangeLog:
2020-04-13 Jozef Lawrynowicz <jozef.l@mittosystems.com>
* config/msp430/msp430.c (msp430_print_operand): Offset a %C memory
reference by 4 bytes, and %D memory reference by 6 bytes.
gcc/testsuite/ChangeLog:
2020-04-13 Jozef Lawrynowicz <jozef.l@mittosystems.com>
* gcc.target/msp430/operand-modifiers.c: New test.
Removes the implementation of __traits(argTypes), which only supported
x86_64 targets. The only use of this trait is an unused va_arg()
function, this has been removed as well.
Reviewed-on: https://github.com/dlang/dmd/pull/11022
gcc/d/ChangeLog:
2020-04-13 Iain Buclaw <ibuclaw@gdcproject.org>
* Make-lang.in (D_FRONTEND_OBJS): Remove d/argtypes.o.
* d-target.cc (Target::toArgTypes): New function.
libphobos/ChangeLog:
2020-04-13 Iain Buclaw <ibuclaw@gdcproject.org>
* libdruntime/core/stdc/stdarg.d: Remove run-time va_list template.
Darwin mandates an indirection for variables in the commmon
section. Since the change to -fno-common, variables in some
of the thunk tests are now in the .data section where they
may be accessed directly. Remove the indirections from the
scan-assembler matches.
gcc/testsuite/ChangeLog:
2020-04-12 Iain Sandoe <iain@sandoe.co.uk>
* gcc.target/i386/indirect-thunk-1.c: Adjust for fno-common
change, removing indirections for vars in .data.
* gcc.target/i386/indirect-thunk-2.c: Likewise.
* gcc.target/i386/indirect-thunk-3.c: Likewise.
* gcc.target/i386/indirect-thunk-4.c: Likewise.
* gcc.target/i386/indirect-thunk-attr-1.c: Likewise.
* gcc.target/i386/indirect-thunk-attr-2.c: Likewise.
* gcc.target/i386/indirect-thunk-attr-3.c: Likewise.
* gcc.target/i386/indirect-thunk-attr-4.c: Likewise.
* gcc.target/i386/indirect-thunk-attr-5.c: Likewise.
* gcc.target/i386/indirect-thunk-attr-6.c: Likewise.
* gcc.target/i386/indirect-thunk-extern-1.c: Likewise.
* gcc.target/i386/indirect-thunk-extern-2.c: Likewise.
* gcc.target/i386/indirect-thunk-extern-3.c: Likewise.
* gcc.target/i386/indirect-thunk-extern-4.c: Likewise.
* gcc.target/i386/indirect-thunk-inline-1.c: Likewise.
* gcc.target/i386/indirect-thunk-inline-2.c: Likewise.
* gcc.target/i386/indirect-thunk-inline-3.c: Likewise.
* gcc.target/i386/indirect-thunk-inline-4.c: Likewise.
V4SI, V8HI and V16QI modes of redux_<code>_scal_<mode> expander
expand with SSE2 instructions (PSRLDQ and PCMPGTx) so use
TARGET_SSE2 as relevant mode iterator codition.
PR target/94494
* config/i386/sse.md (REDUC_SSE_SMINMAX_MODE): Use TARGET_SSE2
condition for V4SI, V8HI and V16QI modes.
testsuite/ChangeLog:
PR target/94494
* gcc.target/i386/pr94494.c: New test.
The test FAILs on powerpc64-linux with -m32 due to psabi warnings.
Furthermore, the test needs really -msse2 to reproduce on x86 -m32 at -O2.
2020-04-11 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/94482
* gcc.dg/torture/pr94482.c: Add -Wno-psabi -w. Don't add -msse
and sse_runtime effective target on x86, instead only add -msse2
if target is sse2_runtime.
Sometimes the cselib_record_sp_cfa_base_equiv makes it into the var-tracking
used VALUEs and needs to be preserved.
2020-04-11 Jakub Jelinek <jakub@redhat.com>
PR debug/94495
PR target/94551
* cselib.c (cselib_record_sp_cfa_base_equiv): Set PRESERVED_VALUE_P on
val->val_rtx.
The expansions for await expressions were specific to particular
cases, this revises it to be more generic.
a: Revise co_await statement walkers.
We want to process the co_awaits one statement at a time.
We also want to be able to determine the insertion points for
new bind scopes needed to cater for temporaries that are
captured by reference and have lifetimes that need extension
to the end of the full expression. Likewise, the handling of
captured references in the evaluation of conditions might
result in the need to make a frame copy.
This reorganises the statement walking code to make it easier to
extend for these purposes.
b: Factor reference-captured temp code.
We want to be able to use the code that writes a new bind expr
with vars (and their initializers) from several places, so split
that out of the maybe_promote_captured_temps() function into a
new replace_statement_captures (). Update some comments.
c: Generalize await statement expansion.
This revises the expansion to avoid the need to expand conditionally
on the tree type. It resolves PR 94528.
gcc/cp/ChangeLog:
2020-04-10 Iain Sandoe <iain@sandoe.co.uk>
PR c++/94538
* coroutines.cc (co_await_expander): Remove.
(expand_one_await_expression): New.
(process_one_statement): New.
(await_statement_expander): New.
(build_actor_fn): Revise to use per-statement expander.
(struct susp_frame_data): Reorder and comment.
(register_awaits): Factor code.
(replace_statement_captures): New, factored from...
(maybe_promote_captured_temps):.. here.
(await_statement_walker): Revise to process per statement.
(morph_fn_to_coro): Use revised susp_frame_data layout.
gcc/testsuite/ChangeLog:
2020-04-10 Iain Sandoe <iain@sandoe.co.uk>
PR c++/94538
* g++.dg/coroutines/pr94528.C: New test.
In C++20 this is well-formed:
using T = int[2];
T t(1, 2);
which means that std::is_constructible_v<int[2], int, int> should be true.
But constructible_expr immediately returned the error_mark_node when it
saw a list with more than one element. To give accurate results in
C++20, we have to try initializing the aggregate from a parenthesized list of
values.
To not repeat the same mistake as in c++/93790, if there's only one
element, I'm trying {} only when () didn't succeed. is_constructible5.C
verifies this.
In paren-init24.C std::is_nothrow_constructible_v doesn't work due to
error: invalid 'static_cast' from type 'int' to type 'int [1]'
and
error: functional cast to array type 'int [2]'
This needs to be fixed in libstdc++.
PR c++/94149
* method.c (constructible_expr): In C++20, try using parenthesized
initialization of aggregates to determine the result of
__is_constructible.
* g++.dg/cpp2a/paren-init24.C: New test.
* g++.dg/cpp2a/paren-init25.C: New test.
* g++.dg/ext/is_constructible5.C: New test.
gcc/testsuite/ChangeLog:
2020-04-10 Fritz Reese <foreese@gcc.gnu.org>
* gfortran.dg/asynchronous_5.f03: Add -fdump-tree-original and fix
patterns for scan-tree-dump.
... which as of PR89433 commit b48f44bf77a39fefc238a16cf1225c6464c82406 causes
an ICE. Not sure if this is actually supposed to be valid or invalid code.
Until the interactions between OpenACC and OpenMP 'target' get defined
properly, make this a compile-time error.
gcc/
PR middle-end/89433
PR middle-end/93465
* omp-general.c (oacc_verify_routine_clauses): Diagnose if
"#pragma omp declare target" has also been applied.
gcc/testsuite/
PR middle-end/89433
PR middle-end/93465
* c-c++-common/goacc-gomp/pr93465-1.c: New file.
As a prerequesite for PR94304, it becomes easier to manage selectively
compiling sublibraries when there's only one library to link to.
So a druntime convenience library is built to be part of phobos, however
separate druntime library is still built and installed, to allow linking
only to the core runtime explicitly, rather than pulling in the entire
standard library with it.
The gdc driver no longer generates an '-lgdruntime' option, and the
inclusion of the libdruntime library path has been removed from the
testsuite.
gcc/d/ChangeLog:
* d-spec.cc (LIBDRUNTIME): Remove.
(LIBDRUNTIME_PROFILE): Remove.
(lang_specific_driver): Don't link in libgdruntime.
gcc/testsuite/ChangeLog:
* lib/gdc.exp (gdc_link_flags): Remove libdruntime library path.
libphobos/ChangeLog:
* d_rules.am (libdgruntime_la_LINK): Move to libdruntime/Makefile.am.
(libgphobos_la_LINK): Move to src/Makefile.am
* libdruntime/Makefile.am: Add libgdruntime_convenience library.
* libdruntime/Makefile.in: Regenerate.
* src/Makefile.am (libgphobos_la_LIBADD): Add libgdruntime_convenience
library.
(libgphobos_la_DEPENDENCIES): Likewise.
* src/Makefile.in: Regenerate.
* testsuite/lib/libphobos.exp: Remove libdruntime library paths.
* testsuite/testsuite_flags.in: Likewise.
