Ivan Maidanski [Fri, 5 Aug 2016 19:03:12 +0000 (22:03 +0300)]
Use GCC atomic intrinsics for PowerPC 32/64 (GCC 4.8+ and clang 3.8+)
(Cherry-pick commit 8e62b6f from 'ppc-gcc-atomics' branch.)
* src/atomic_ops/sysdeps/gcc/powerpc.h: Include generic.h and do not
include all_aligned_atomic_load_store.h, test_and_set_t_is_ao_t.h if
GCC 4.8+ or clang 3.8+ unless AO_DISABLE_GCC_ATOMICS.
* src/atomic_ops/sysdeps/gcc/powerpc.h (AO_nop_full, AO_lwsync,
AO_nop_write, AO_nop_read, AO_load_acquire, AO_store_release,
AO_test_and_set, AO_test_and_set_release, AO_test_and_set_full,
AO_compare_and_swap, AO_compare_and_swap_acquire,
AO_compare_and_swap_release, AO_compare_and_swap_full,
AO_fetch_compare_and_swap, AO_fetch_compare_and_swap_acquire,
AO_fetch_compare_and_swap_release, AO_fetch_compare_and_swap_full,
AO_fetch_and_add, AO_fetch_and_add_acquire, AO_fetch_and_add_release,
AO_fetch_and_add_full, AO_T_IS_INT): Do not define if generic.h is
included.
* src/atomic_ops_malloc.c [SIZE_MAX && !CPPCHECK] (AO_SIZE_MAX): Cast
SIZE_MAX to size_t explicitly (to workaround incorrect definition
of SIZE_MAX - e.g. it is defined as ~0ULL in musl-tools which is not
correct for a 32-bit target).
Ivan Maidanski [Wed, 15 Mar 2017 08:22:26 +0000 (11:22 +0300)]
Refine configure messages when checking for compiler options
* configure.ac (AC_MSG_CHECKING, AC_MSG_RESULT): Refine message when
checking for a compiler option support (replace "gcc" with "compiler"),
and when reporting PIC option required; put string in [].
* configure.ac (AC_MSG_RESULT): Do not put string in quotes.
Ivan Maidanski [Wed, 1 Mar 2017 22:50:24 +0000 (01:50 +0300)]
Fix missing .exe for test filenames in Makefile (MinGW)
(fix commits ad82960, 13a17dd, ecd43ca)
* tests/Makefile.am (TESTS, check-without-test-driver): Add $(EXEEXT)
suffix to each item (test).
* tests/Makefile.am (TEST_OBJS): Reorder items (to match the that of
TESTS).
* tests/Makefile.am (check_PROGRAMS): Specify all items explicitly (so
that to have all items without the suffix as before this change).
Ivan Maidanski [Tue, 21 Feb 2017 08:43:42 +0000 (11:43 +0300)]
Update shared libraries version info to differentiate against v7.4.x
Version info "current" and "age" numbers are incremented, "revision"
number is reset to zero (indicating that programs using the previous
libraries version may use the new version as drop-in replacement).
Thus, version info is 2:0:1 now (both for libatomic_ops.so and
libatomic_ops_gpl.so).
* src/Makefile.am (LIBATOMIC_OPS_VER_INFO, LIBATOMIC_OPS_GPL_VER_INFO):
Change version info from 1:3:0 (used for libatomic_ops v7.4.x releases)
to 2:0:1.
Ivan Maidanski [Mon, 20 Feb 2017 20:30:55 +0000 (23:30 +0300)]
Fix 'unknown attribute no_sanitize' compiler warning (clang prior to v3.8)
* src/atomic_ops.h [!AO_ATTR_NO_SANITIZE_MEMORY && AO_MEMORY_SANITIZER]
(AO_ATTR_NO_SANITIZE_MEMORY): Do not define to no_sanitize attribute
unless clang-3.8+ (or not clang).
* tests/test_atomic_include.template [AO_HAVE_storeXX
&& AO_MEMORY_SANITIZER && AO_PREFER_GENERALIZED] (test_atomicXX):
Initialize x local variable (before AO_storeXX which can be defined
using CAS); add comment.
* tests/test_atomic_include.template [AO_HAVE_short_storeXX
&& AO_MEMORY_SANITIZER && AO_PREFER_GENERALIZED] (test_atomicXX):
Initialize s local variable.
* tests/test_atomic_include.template [AO_HAVE_char_storeXX
&& AO_MEMORY_SANITIZER && AO_PREFER_GENERALIZED] (test_atomicXX):
Initialize b local variable.
* tests/test_atomic_include.template [AO_HAVE_int_storeXX
&& AO_MEMORY_SANITIZER && AO_PREFER_GENERALIZED] (test_atomicXX):
Initialize zz local variable.
Ivan Maidanski [Mon, 20 Feb 2017 19:16:54 +0000 (22:16 +0300)]
Eliminate 'condition sizeof(long)>4 is always true' cppcheck style warning
* src/atomic_ops_malloc.c (msb): Cast (s >> 32) to unsigned int type
before assigning the result to v; check __SIZEOF_SIZE_T__ value (if
defined) to decide whether (s >> 32) is needed; move condition
(sizeof(size_t) > 4) to SIZEOF_SIZE_T_GT_4 macro (to eliminate cppcheck
false report that the condition is always true).
Ivan Maidanski [Mon, 20 Feb 2017 08:10:07 +0000 (11:10 +0300)]
Workaround 'AM_EXTRA_RECURSIVE_TARGETS not found in library' autoconf error
(fix commit 3b7e732)
* Makefile.am (.PHONY): Add check-nolink, check-nolink-local targets.
* Makefile.am (check-nolink): New target (which invokes existing
check-nolink-local in this Makefile and in tests/Makefile).
* configure.ac (AM_EXTRA_RECURSIVE_TARGETS): Remove (this directive
is not supported by autoconf-1.12 and earlier).
Ivan Maidanski [Sat, 18 Feb 2017 07:23:33 +0000 (10:23 +0300)]
Add Makefile target to run all tests without test-driver
This could be used to catch errors reported by MemorySanitizer
(the traditional "make check" shows the tests with MSan errors as
skipped, so it does not lead to a non-zero exit code of make itself).
* tests/Makefile.am (check-without-test-driver): New phony target (to
compile, link and execute all tests not using test-driver).
Ivan Maidanski [Sat, 18 Feb 2017 06:21:58 +0000 (09:21 +0300)]
Add Makefile target (check-nolink) to compile all source without linking
Note: it is assumed --enable-shared is not passed to configure.
* Makefile.am (check-nolink-local): New Makefile phony target (just
redirects to "all").
* configure.ac: Define check-nolink by AM_EXTRA_RECURSIVE_TARGETS.
* tests/Makefile.am (TEST_OBJS): New variable (with a list of the test
.o files).
* tests/Makefile.am (check-nolink-local): New Makefile phony target
(forces all tests to be compiled but not linked).
Ivan Maidanski [Fri, 17 Feb 2017 20:48:23 +0000 (23:48 +0300)]
Refine documentation in Makefile.msft how to run all tests (MS VC)
* src/Makefile.msft: Refine comment about all tests execution
("make check" is redundant, it is enough to run only "make" to
auto-generate list_atomic.c and test_atomic_include.h).
Ivan Maidanski [Fri, 17 Feb 2017 20:27:19 +0000 (23:27 +0300)]
Move libraries version info to the beginning of Makefile.am
(code refactoring)
* src/Makefile.am (LIBATOMIC_OPS_VER_INFO, LIBATOMIC_OPS_GPL_VER_INFO):
New variable; add comment about the version update rule.
* src/Makefile.am (libatomic_ops_la_LDFLAGS,
libatomic_ops_gpl_la_LDFLAGS): Use LIBATOMIC_OPS_[GPL_]VER_INFO
instead direct hard-coding of version info.
Ivan Maidanski [Fri, 17 Feb 2017 08:51:30 +0000 (11:51 +0300)]
Travis CI: Also test with Memory and Unexpected Behavior sanitizers
Note: Currently an error found by MSan/UBSan does not cause Travis build
to fail, the log should be examined manually for the absence of
"MemorySanitizer" and "runtime error" character sequences (the error
messages are shown in red).
Ivan Maidanski [Fri, 27 Jan 2017 16:35:20 +0000 (19:35 +0300)]
Use generalized load/store if AO_PREFER_GENERALIZED (gcc/generic)
Note: AO_PREFER_GENERALIZED macro is for testing generalize.h mainly.
* src/atomic_ops/sysdeps/gcc/generic-small.template (AO_XSIZE_load,
AO_XSIZE_load_acquire, AO_XSIZE_store, AO_XSIZE_store_release): Do not
define if AO_GCC_HAVE_XSIZE_SYNC_CAS and AO_PREFER_GENERALIZED.
* src/atomic_ops/sysdeps/gcc/generic-small.h: Regenerate.
* src/atomic_ops/sysdeps/gcc/generic.h [AO_HAVE_DOUBLE_PTR_STORAGE]
(AO_GCC_HAVE_double_SYNC_CAS): Move the definition upper to be before
AO_double_load.
* src/atomic_ops/sysdeps/gcc/generic.h [AO_HAVE_DOUBLE_PTR_STORAGE]
(AO_double_load, AO_double_load_acquire, AO_double_store,
AO_double_store_release): Do not define if AO_GCC_HAVE_double_SYNC_CAS
and AO_PREFER_GENERALIZED.
Ivan Maidanski [Thu, 26 Jan 2017 20:07:57 +0000 (23:07 +0300)]
Enable limited testing in Makefile.msft without Cygwin
New target "check-noautogen" is introduced in Makefile.msft to build and
run only test_malloc and test_stack (which do not need test code
auto-generation by sed).
* src/Makefile.msft: Refine comment about tests execution; add comment
about "check-noautogen" target.
* src/Makefile.msft (check): Add dependency on check-noautogen;
move test_malloc and test_stack to check-noautogen.
* src/Makefile.msft (check-noautogen): New target (executes test_malloc
and test_stack only).
Ivan Maidanski [Thu, 26 Jan 2017 07:38:21 +0000 (10:38 +0300)]
Implement AO_CLEAR using C11 atomic intrinsic (GCC)
* src/atomic_ops/sysdeps/gcc/generic.h [!AO_PREFER_GENERALIZED]
(AO_CLEAR): Implement using GCC built-in __atomic_clear (thus
overriding the default one based on AO_char_store_release).
Hans Boehm [Thu, 26 Jan 2017 07:16:54 +0000 (10:16 +0300)]
Implement compiler_barrier using C11 __atomic_signal_fence (GCC)
Use __atomic_signal_fence(__ATOMIC_SEQ_CST) as AO_compiler_barrier.
That was the intended use of __atomic_signal_fence. The only
standards-relevant context in which you can tell apart a compiler and
real fence is in the case if signal handlers.
* src/atomic_ops/sysdeps/gcc/generic.h (AO_compiler_barrier): Redefine
using GCC atomic intrinsic (__atomic_signal_fence).
Ivan Maidanski [Tue, 24 Jan 2017 19:46:57 +0000 (22:46 +0300)]
Fix test_atomic failure caused unaligned AO_double_t access on x86
The failure is caused by violation of an assertion that checks
AO_double_t variable is 8-byte aligned on x86.
* doc/README.txt (AO_double_t): Add note about required alignment.
* src/atomic_ops/sysdeps/standard_ao_double_t.h (AO_double_t): Add
comment about alignment.
* src/atomic_ops_stack.h (AO_stack_t): Likewise.
* tests/test_atomic_include.template (test_atomicXX): Define old_w, w
local variables as static (as otherwise, e.g., they could have 4-byte
alignment on x86); add comment.
The failure is actually caused by the fact that default alignment of
stack on x86 is 4 bytes.
So, another solution should be proposed to fix the issue for other
compilers (and, potentially, other targets). E.g. do not place
AO_double_t volatile variables on stack.