The -enable-checking configure code in libcody didn't play well with
us. This just uses libcpp's configurey for that piece.
libcody/
* configure.ac: Use libcpp's enable-checking code.
* configure: Rebuilt.
module.cc has a static initializer that ends up in a circular
dependency when detailed mem stats are enabled. This removes the need
for that initializer to be dynamic, and we punt to the lazy
initializing we already had inside the object in question anyway. At
the cost of an additional indirection.
gcc/cp/
* module.cc (loc_spans): Make spans a pointer, not inline.
Adjust all accesses.
When SECTION_RETAIN is used, issue a warning when a symbol without used
attribute and a symbol with used attribute are placed in the section with
the same name, like
int __attribute__((used,section(".data.foo"))) foo2 = 2;
int __attribute__((section(".data.foo"))) foo1 = 1;
since assembler will put them in different sections with the same section
name.
gcc/
PR target/98146
* varasm.c (switch_to_section): Warn when a symbol without used
attribute and a symbol with used attribute are placed in the
section with the same name.
gcc/testsuite/
PR target/98146
* c-c++-common/attr-used-5.c: Updated.
* c-c++-common/attr-used-6.c: Likewise.
* c-c++-common/attr-used-7.c: Likewise.
* c-c++-common/attr-used-8.c: Likewise.
When definitions marked with used attribute and unmarked definitions are
placed in the section with the same name, switch to a new section if the
SECTION_RETAIN bit doesn't match.
gcc/
PR target/98146
* output.h (switch_to_section): Add a tree argument, default to
nullptr.
* varasm.c (get_section): If the SECTION_RETAIN bit doesn't match,
return and switch to a new section later.
(assemble_start_function): Pass decl to switch_to_section.
(assemble_variable): Likewise.
(switch_to_section): If the SECTION_RETAIN bit doesn't match,
switch to a new section.
gcc/testsuite/
PR target/98146
* c-c++-common/attr-used-5.c: New test.
* c-c++-common/attr-used-6.c: Likewise.
* c-c++-common/attr-used-7.c: Likewise.
* c-c++-common/attr-used-8.c: Likewise.
* c-c++-common/attr-used-9.c: Likewise.
Currently the <experimental/random>, <experimental/source_location> and
<experimental/utility> headers can be included in C++98 and C++11 modes,
but gives errors. With this change they can be included, but define
nothing.
libstdc++-v3/ChangeLog:
PR libstdc++/98319
* include/experimental/random: Only define contents for C++14
and later.
* include/experimental/source_location: Likewise.
* include/experimental/utility: Likewise.
* testsuite/experimental/feat-lib-fund.cc: Include all LFTS
headers that are present. Allow test to run for all modes.
This adds a test to compare the performance of std::atomic_flag with
similar operations on std::atomic_uchar and std::atomic_int.
libstdc++-v3/ChangeLog:
PR libstdc++/46447
* testsuite/performance/29_atomics/atomic_flag.cc: New test.
This fixes a bug caused by a mismatch between the macros defined by
<errno.h> when GCC is built and the macros defined by <errno.h> when
users include <system_error>. If the user code is compiled with
_XOPEN_SOURCE defined to 500 or 600, Darwin suppresses the
ENOTRECOVERABLE and EOWNERDEAD macros, which are not defined by SUSv3
(aka POSIX.1-2001).
Since POSIX requires the errno macros to be macros (and not variables or
enumerators) we can just test for them directly using the preprocessor.
That means that <system_error> will match what is actuallydefined when
it's included, not what was defined when GCC was built. With that change
there is no need for the GLIBCXX_CHECK_SYSTEM_ERROR configure checks and
they can be removed.
libstdc++-v3/ChangeLog:
PR libstdc++/93151
* acinclude.m4 (GLIBCXX_CHECK_SYSTEM_ERROR): Remove.
* configure.ac: Regenerate.
* config/os/generic/error_constants.h: Test POSIX errno macros
directly, instead of corresponding _GLIBCXX_HAVE_EXXX macros.
* testsuite/19_diagnostics/headers/system_error/errc_std_c++0x.cc:
Likewise.
* testsuite/19_diagnostics/headers/system_error/93151.cc: New
test.
Solaris' sys/socket uses the poisoned bcopy identifier, so we must
preemptively copy a bit of cody's inclusion logic to get it earlier.
gcc/cp/
* mapper-client.cc: Include sys/socket.h before system.h.
gcc/ada/
* sem_ch5.adb (Analyze_Iterator_Specification): If iterator
filter is present, preanalyze filter without expansion.
(Analyze_Loop_Parameter_Specification): When
loop_Parameter_Specification is rewritten as
Iterator_Specification, transfer Iterator_Filter if present.
gcc/ada/
* libgnat/s-objrea.ads (Object_Arch): Add ARM enum
* libgnat/s-objrea.adb (Initialize): Add EM_ARM case.
(Read_Address): Add ARM case to 32bit read.
* Makefile.rtl: Add trasym units to the runtime for armhf-linux.
gcc/ada/
* libgnat/s-valrea.adb (Maxexp32): New constant array.
(Maxexp64): Likewise.
(Maxexp80): Likewise.
(Integer_to_Real): New local constants Maxexp and B.
When the exponent is too negative, do the divison in two steps.
gcc/ada/
* sem_ch13.adb (Analyze_Aspect_Specifications): Add a codefix
for extra parentheses around aspect Annotate expression; reject
"(null record)" aggregate and extra parentheses around aspect
Test_Case expression.
* sem_prag.adb (Analyze_Pragma): Reject "null", "(null record)"
and extra parentheses around pragma Contract_Cases; likewise for
pragma Subprogram_Variant.
gcc/ada/
* adaint.h (__gnat_in_child_after_fork): New flag to express
child process side after fork call.
* adaint.c (__gnat_portable_spawn): Set flag
__gnat_in_child_after_fork.
* expect.c (__gnat_expect_fork): Set __gnat_in_child_after_fork
to one on child side.
* libgnat/memtrack.adb
(In_Child_After_Fork): Flag to disable memory tracking.
(Allow_Trace): New routine defining if memory should be tracked.
(Alloc, Realloc, Free): Use Allow_Trace in "if" condition
instead of First_Call.
gcc/ada/
* libgnat/a-tifiio.adb: Mark body not in SPARK.
* libgnat/a-tifiio.ads: Mark spec in SPARK.
* libgnat/a-tifiio__128.adb: Mark body not in SPARK.
The following patch teaches the bswap pass to handle for small (2/4/8 byte
long) vectors a CONSTRUCTOR by determining if the bytes of the constructor
come from non-vector sources and are either nop or bswap and changing the
CONSTRUCTOR in that case to VIEW_CONVERT_EXPR from scalar integer to
the vector type.
Unfortunately, as I found after the patch was written, due to pass ordering
this doesn't really fix the original testcase, just the one I wrote,
because both loop and slp vectorization is done only after the bswap pass.
A possible way out of that would be to perform just this particular bswap
optimization (i.e. for CONSTRUCTOR assignments with integral vector types
call find_bswap_or_nop and bswap_replace if successful) also during the
store merging pass, it isn't really a store, but the store merging pass
already performs bswapping when handling store, so it wouldn't be that big
hack. What do you think?
2020-12-16 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/96239
* gimple-ssa-store-merging.c (find_bswap_or_nop): Handle a vector
CONSTRUCTOR.
(bswap_replace): Likewise.
* gcc.dg/pr96239.c: New test.
Since g:7caa49706316e650fb67719e1a1bf3a35054b685 the option is ignored
as we print used command line for -fverbose-asm output.
gcc/ChangeLog:
* doc/options.texi: Remove Report keyword.
* opt-functions.awk: Print error when Report keyword
is used.
* optc-gen.awk: Do not handle Report keyword.
* opts.h (struct cl_option): Remove cl_report bitfield flag.
This fixes the precision mismatch introduced by the previous change.
gcc/ChangeLog:
PR tree-optimization/98272
* tree-switch-conversion.c (bit_test_cluster::emit): When finding
out whether the entry test can be merged in the bit test, do the
computation using the type of the index expression.
gcc/testsuite/ChangeLog:
* gcc.dg/pr98272.c: New test.
This patch is to use paradoxical subreg instead of
zero_extend for promoting QI/HI to SI/DI when we
want to construct one vector with these modes.
Since we do the gpr->vsx movement and vector merge
or pack later, the high part is useless and safe to
use paradoxical subreg. It can avoid useless rlwinms
generated for signed cases.
Bootstrapped/regtested on powerpc64le-linux-gnu P9.
gcc/ChangeLog:
* config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
paradoxical subreg instead of zero_extend for QI/HI promotion.
gcc/testsuite/ChangeLog:
* gcc.target/powerpc/pr96933-1.c: Adjusted to check no rlwinm.
* gcc.target/powerpc/pr96933-2.c: Likewise.
It turned out that r11-5942 fixed this old PR, and it was detected
by one of the few dg-ice tests -- exactly the point of them! Now
the PR won't be opened until someone notices that it'd been fixed.
The patch failed to remove the dg-ice though, so now it XPASSes. Fixing
this now.
gcc/testsuite/ChangeLog:
* g++.dg/cpp0x/constexpr-52830.C: Remove dg-ice.
When the divisor is bool or has [0, 1] range, as division by
0 is UB, the only remaining option in valid programs is division by 1,
so we can optimize X / bool_range_Y into X.
2020-12-15 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/96094
* match.pd (X / bool_range_Y -> X): New simplification.
* gcc.dg/tree-ssa/pr96094.c: New test.
The DMD front-end shouldn't, but can sometimes leak manifest constants
in the AST passed to the code generator. To prevent this being an
issue, the setting of DECL_INITIAL has been moved to the point where the
CONST_DECL is used, rather than in the declaration handler.
gcc/d/ChangeLog:
PR d/98277
* decl.cc (DeclVisitor::visit (VarDeclaration *)): Move setting of
DECL_INITIAL for manifest constants to ...
(get_symbol_decl): ... here.
gcc/testsuite/ChangeLog:
PR d/98277
* gdc.dg/pr98277.d: New test.
