[Bug c++/63323] New: confused by earlier errors, bailing out with no other errors
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63323
Bug ID: 63323
Summary: confused by earlier errors, bailing out with no
other errors
Product: gcc
Version: 4.9.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: tavianator at gmail dot com
Created attachment 33529
-- https://gcc.gnu.org/bugzilla/attachment.cgi?id=33529action=edit
Preprocessed source
The following file gives a confused by earlier errors, bailing out error
message, with no previous errors printed, when compiled with
-fsanitize=undefined:
$ cat confused.cpp
#include memory
#include map
int
main()
{
std::unique_ptrint ptr{new int};
std::mapint, std::unique_ptrint map;
map.insert({0, std::move(ptr)});
return 0;
}
$ g++ -std=c++11 -fsanitize=undefined confused.cpp
‘
/usr/include/c++/4.9.1/ext/new_allocator.h:120: confused by earlier errors,
bailing out
Without the -fsanitize=undefined, it points to the actual error, which is fixed
by using std::make_pair instead of a braced initializer as the argument to
insert().
[Bug tree-optimization/63537] New: Missed optimization: Loop unrolling adds extra copy when returning aggregate
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63537
Bug ID: 63537
Summary: Missed optimization: Loop unrolling adds extra copy
when returning aggregate
Product: gcc
Version: 4.9.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: tavianator at gmail dot com
Created attachment 33715
-- https://gcc.gnu.org/bugzilla/attachment.cgi?id=33715action=edit
Reproducer
At -O2 and above on x86_64, this manually unrolled loop generates much better
code than the automatically unrolled one:
struct vec {
double n[3];
};
struct vec mul_unrolled(struct vec lhs, double rhs) {
struct vec ret;
ret.n[0] = lhs.n[0]*rhs;
ret.n[1] = lhs.n[1]*rhs;
ret.n[2] = lhs.n[2]*rhs;
return ret;
}
This generates the beautiful:
movsd16(%rsp), %xmm2
movq%rdi, %rax
movsd24(%rsp), %xmm1
mulsd%xmm0, %xmm2
mulsd%xmm0, %xmm1
mulsd8(%rsp), %xmm0
movsd%xmm2, 8(%rdi)
movsd%xmm1, 16(%rdi)
movsd%xmm0, (%rdi)
ret
In contrast, at -O2 this:
struct vec mul_loop(struct vec lhs, double rhs) {
struct vec ret;
for (int i = 0; i 3; ++i) {
ret.n[i] = lhs.n[i]*rhs;
}
return ret;
}
generates this:
movsd8(%rsp), %xmm1
movq%rdi, %rax
mulsd%xmm0, %xmm1
movsd%xmm1, -40(%rsp)
movq-40(%rsp), %rdx
movsd16(%rsp), %xmm1
mulsd%xmm0, %xmm1
movq%rdx, (%rdi)
mulsd24(%rsp), %xmm0
movsd%xmm1, -32(%rsp)
movq-32(%rsp), %rdx
movsd%xmm0, -24(%rsp)
movq%rdx, 8(%rdi)
movq-24(%rsp), %rdx
movq%rdx, 16(%rdi)
ret
which puts the result in -40(%rsp) and then copies it to (%rdi). At -O3 it
gets vectorized but the extra copy is still there:
movapd%xmm0, %xmm1
mulsd24(%rsp), %xmm0
movupd8(%rsp), %xmm2
movq%rdi, %rax
unpcklpd%xmm1, %xmm1
mulpd%xmm1, %xmm2
movsd%xmm0, -24(%rsp)
movaps%xmm2, -40(%rsp)
movq-40(%rsp), %rdx
movq%rdx, (%rdi)
movq-32(%rsp), %rdx
movq%rdx, 8(%rdi)
movq-24(%rsp), %rdx
movq%rdx, 16(%rdi)
[Bug tree-optimization/63537] [4.8/4.9/5 Regression] Missed optimization: Loop unrolling adds extra copy when returning aggregate
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63537 --- Comment #2 from Tavian Barnes tavianator at gmail dot com --- Is it possible to make SRA work even if the loop isn't unrolled? If the array size is increased to 4 then -O2 doesn't unroll the loop at all, resulting in: movq%rdi, %rax xorl%edx, %edx .L3: movsd8(%rsp,%rdx), %xmm1 mulsd%xmm0, %xmm1 movsd%xmm1, -40(%rsp,%rdx) addq$8, %rdx cmpq$32, %rdx jne.L3 movq-40(%rsp), %rdx movq%rdx, (%rax) movq-32(%rsp), %rdx movq%rdx, 8(%rax) movq-24(%rsp), %rdx movq%rdx, 16(%rax) movq-16(%rsp), %rdx movq%rdx, 24(%rax) ret which would be a lot prettier as something like: movq%rdi, %rax xorl%edx, %edx .L3: movsd8(%rsp,%rdx), %xmm1 mulsd%xmm0, %xmm1 movsd%xmm1, (%rax,%rdx) addl$8, %edx cmpl$32, %edx jne.L3 ret
[Bug c/61118] New: Spurious -Wclobbered warning generated by gcc 4.9.0 for pthread_cleanup_push
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=61118
Bug ID: 61118
Summary: Spurious -Wclobbered warning generated by gcc 4.9.0
for pthread_cleanup_push
Product: gcc
Version: 4.9.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
Assignee: unassigned at gcc dot gnu.org
Reporter: tavianator at gmail dot com
Created attachment 32762
-- http://gcc.gnu.org/bugzilla/attachment.cgi?id=32762action=edit
Source file
On gcc 4.9.0 (but not on 4.8.2 or earlier), the attached code gives two
spurious warnings:
$ gcc -std=gnu99 -Wclobbered -O2 -c clobber_warning.c -pthread -save-temps
clobber_warning.c: In function ‘dmnsn_future_wait’:
clobber_warning.c:23:54: warning: variable ‘__cancel_routine’ might be
clobbered by ‘longjmp’ or ‘vfork’ [-Wclobbered]
pthread_cleanup_push(cleanup_fn, future-mutex);
^
clobber_warning.c:23:103: warning: variable ‘__cancel_arg’ might be clobbered
by ‘longjmp’ or ‘vfork’ [-Wclobbered]
pthread_cleanup_push(cleanup_fn, future-mutex);
Those variables come from the expansion of the pthread_cleanup_{push,pop}
macros, which look like this (reformatted a little):
do {
__pthread_unwind_buf_t __cancel_buf;
void (*__cancel_routine) (void *) = (cleanup_fn);
void *__cancel_arg = (future-mutex);
int __not_first_call = __sigsetjmp ((struct __jmp_buf_tag *) (void *)
__cancel_buf.__cancel_jmp_buf, 0);
if (__builtin_expect ((__not_first_call), 0)) {
__cancel_routine (__cancel_arg);
__pthread_unwind_next (__cancel_buf);
}
__pthread_register_cancel (__cancel_buf);
do {;
pthread_cond_wait(future-cond, future-mutex);
do { } while (0);
} while (0);
__pthread_unregister_cancel (__cancel_buf);
if (0)
__cancel_routine (__cancel_arg);
} while (0);
The __cancel_routine and __cancel_arg variables are never modified, so I don't
see how they can be clobbered.
[Bug c/61118] Spurious -Wclobbered warning generated by gcc 4.9.0 for pthread_cleanup_push
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=61118 --- Comment #1 from Tavian Barnes tavianator at gmail dot com --- Created attachment 32763 -- http://gcc.gnu.org/bugzilla/attachment.cgi?id=32763action=edit Preprocessed source file
[Bug c/43904] New: Wrong code with -foptimize-sibling-calls and memcpy on x86_64
On x86_64, gcc 4.5.0, this code generates bad assembly:
--- C code ---
typedef unsigned long size_t;
void *memcpy(void *dest, const void *src, size_t n);
void
buggy_init(void *ptr, size_t size)
{
const char *str = Hello world!;
memcpy(ptr, str, size);
}
--
Compiled with gcc -O -foptimize-sibling-calls, the generated assembly looks
like this:
--
buggy_init:
movq%rsi, %rdx
movq$.LC0, -16(%rsp)
leaq-16(%rsp), %rsi
jmp memcpy
--
which passes rsp-16 as memcpy's second argument. memcpy overwrites this part
of the stack, and copies the wrong value, which causes a crash later.
--
Summary: Wrong code with -foptimize-sibling-calls and memcpy on
x86_64
Product: gcc
Version: 4.5.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: tavianator at gmail dot com
GCC build triplet: x86_64-unknown-linux-gnu
GCC host triplet: x86_64-unknown-linux-gnu
GCC target triplet: x86_64-unknown-linux-gnu
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43904
[Bug c/43904] Wrong code with -foptimize-sibling-calls and memcpy on x86_64
--- Comment #2 from tavianator at gmail dot com 2010-04-26 23:47 --- Created an attachment (id=20497) -- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20497action=view) Full testcase Proper output: Stored: 0x40071c Got:0x40071c Hello world! Output with -O -foptimize-sibling-calls: Stored: 0x40070c Got:0x1 [1]15940 segmentation fault (core dumped) ./a.out -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43904
[Bug c++/63723] New: Narrowing conversion allowed in braced init list in SFINAE context
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63723
Bug ID: 63723
Summary: Narrowing conversion allowed in braced init list in
SFINAE context
Product: gcc
Version: 4.9.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: tavianator at gmail dot com
Created attachment 33877
-- https://gcc.gnu.org/bugzilla/attachment.cgi?id=33877action=edit
Preprocessed source
GCC allows narrowing conversions in braced init lists with only a warning. The
documentation suggests that this extension doesn't apply in SFINAE contexts,
but it appears to be wrong:
$ cat foo.cpp
#include type_traits
#include utility
template typename From, typename To
class is_list_convertible_helper
{
template typename To2
static void requires_conversion(To2 t);
template typename From2, typename To2,
typename = decltype(requires_conversionTo2({std::declvalFrom2()}))
// ^ Braced initializer
static std::true_type helper(int);
template typename From2, typename To2
static std::false_type helper(...);
public:
using type = decltype(helperFrom, To(0));
};
template typename From, typename To
class is_list_convertible
: public is_list_convertible_helperFrom, To::type
{ };
static_assert(!is_list_convertibledouble, int::value,
double - int is narrowing!);
$ g++ -std=c++11 foo.cpp
foo.cpp: In substitution of ‘templateclass From2, class To2, class static
std::true_type is_list_convertible_helperFrom, To::helper(int) [with From2 =
double; To2 = int; template-parameter-1-3 = missing]’:
foo.cpp:18:31: required from ‘class is_list_convertible_helperdouble, int’
foo.cpp:22:7: required from ‘class is_list_convertibledouble, int’
foo.cpp:26:48: required from here
foo.cpp:10:46: warning: narrowing conversion of ‘std::declvaldouble()’ from
‘double’ to ‘int’ inside { } [-Wnarrowing]
typename = decltype(requires_conversionTo2({std::declvalFrom2()}))
^
foo.cpp:26:1: error: static assertion failed: double - int is narrowing!
static_assert(!is_list_convertibledouble, int::value,
^
[Bug c++/63723] Narrowing conversion allowed in braced init list in SFINAE context
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63723
--- Comment #2 from Tavian Barnes tavianator at gmail dot com ---
It's decltype(requires_conversionTo2({std::declvalFrom2()})). Not sure why
it says missing.
[Bug c++/63723] Narrowing conversion allowed in braced init list in SFINAE context
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63723 --- Comment #4 from Tavian Barnes tavianator at gmail dot com --- Still happens with 4.9.2 though. Is a backport of the fix possible?
[Bug libstdc++/63840] New: std::function copy constructor deletes an uninitialized pointer if new fails
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63840
Bug ID: 63840
Summary: std::function copy constructor deletes an
uninitialized pointer if new fails
Product: gcc
Version: 5.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: libstdc++
Assignee: unassigned at gcc dot gnu.org
Reporter: tavianator at gmail dot com
Created attachment 33953
-- https://gcc.gnu.org/bugzilla/attachment.cgi?id=33953action=edit
Reproducer
std::function's copy constructor looks like this:
templatetypename _Res, typename... _ArgTypes
function_Res(_ArgTypes...)::
function(const function __x)
: _Function_base()
{
if (static_castbool(__x))
{
_M_invoker = __x._M_invoker;
_M_manager = __x._M_manager;
__x._M_manager(_M_functor, __x._M_functor, __clone_functor);
}
}
_M_manager(..., __clone_functor) calls _M_clone, which looks like this when the
functor is stored on the heap:
static void
_M_clone(_Any_data __dest, const _Any_data __source, false_type)
{
__dest._M_access_Functor*() =
new _Functor(*__source._M_access_Functor*());
}
If operator new or the copy-constructor throws, __dest._M_pod_data remains
uninitialized. Then the stack unwinds, and ~_Function_base() gets called:
~_Function_base()
{
if (_M_manager)
_M_manager(_M_functor, _M_functor, __destroy_functor);
}
Which ultimately calls
static void
_M_destroy(_Any_data __victim, false_type)
{
delete __victim._M_access_Functor*();
}
Which deletes _M_pod_data. A simple fix could be:
templatetypename _Res, typename... _ArgTypes
function_Res(_ArgTypes...)::
function(const function __x)
: _Function_base()
{
if (static_castbool(__x))
{
+ __x._M_manager(_M_functor, __x._M_functor, __clone_functor);
_M_invoker = __x._M_invoker;
_M_manager = __x._M_manager;
- __x._M_manager(_M_functor, __x._M_functor, __clone_functor);
}
}
I have a test case attached.
[Bug c++/57510] initializer_list memory leak
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57510
--- Comment #4 from Tavian Barnes tavianator at gmail dot com ---
I have a testing tool that automatically inserts operator new failures, to help
test exception safety and check for leaks. This bug causes all kinds of
spurious failures that I have to work around, since anything like this
vectorstring vec = {some, strings};
leaks memory if the second string constructor fails. Usually it can be worked
around like this
string arr[] = {some, strings};
but it's still quite annoying.
If someone can point me in the right direction or give an outline of how to fix
this bug, I'm happy to try and write up a patch myself. Thanks!
[Bug tree-optimization/64308] New: Missed optimization: 64-bit divide used when 32-bit divide would work
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64308
Bug ID: 64308
Summary: Missed optimization: 64-bit divide used when 32-bit
divide would work
Product: gcc
Version: 4.9.2
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: tavianator at gmail dot com
Created attachment 34280
-- https://gcc.gnu.org/bugzilla/attachment.cgi?id=34280action=edit
Test case
The following is a fairly typical implementation of exponentiation modulo m:
$ cat ipowm.c
// Computes (b**e) % m
unsigned int
ipowm(unsigned int b, unsigned int e, unsigned int m)
{
unsigned int ret;
b %= m;
for (ret = m 1; e; e = 1) {
if ((e 1) == 1) {
ret = (unsigned long long)ret * b % m;
}
b = (unsigned long long)b * b % m;
}
return ret;
}
Unfortunately, GCC emits a 64-bit multiply and divide for both ... * b % m
expressions on x86 and x86-64, where a 32-bit multiply and divide would be
equivalent and faster.
$ gcc -std=c11 -O3 -Wall -S -save-temps ipowm.c
$ cat ipowm.s
...
imulq%rdi, %rax
divq%rcx
...
imulq%rdi, %rax
divq%rcx
...
The pattern
mull%edi
divl%ecx
would be much faster. They're equivalent because b is always reduced mod m, so
b m and therefore (for any unsigned int x), x * b / m = x * m / m == x, thus
the quotient will always fit in 32 bits.
[Bug tree-optimization/64308] Missed optimization: 64-bit divide used when 32-bit divide would work
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64308 --- Comment #3 from Tavian Barnes tavianator at gmail dot com --- @Richard Biener: Yes the range for _16 could be [0, 4294967294]. Why can't VRP can't assume division by zero doesn't occur? If it can then it could say anything mod [a, b] fits in [0, b - 1]. That's a reasonable improvement by itself but it's not enough to optimize this PR, because to use divl for (ret * b % m), you need (ret * b / m) to fit in [0, 4294967295] as well. And to know that that, as Marc Glisse suggests, you'd need symbolic ranges. @Marc Glisse: Is there currently no support at all for symbolic ranges? If you can infer that b m is an invariant then that's all you need. Formally it's something like this: If x, y, and z are 32-bit unsigned integers, and x = z || y = z, then (uint64_t)x * (uint64_t)y % z can be computed with mull and divl because x * y / z is always = max(x, y) which fits in 32 bits.
[Bug c++/63723] Narrowing conversion allowed in braced init list in SFINAE context
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63723
--- Comment #8 from Tavian Barnes tavianator at gmail dot com ---
Actually it is a regression: according to http://gcc.godbolt.org/, GCC 4.6.4
compiled it correctly with -std=c++0x and
typedef decltype(helperFrom, To(0)) type;
instead of
using type = decltype(helperFrom, To(0));
while still allowing
int x{1.0};
[Bug c++/64372] New: Spurious warning with throw in ternary operator returning const reference
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64372
Bug ID: 64372
Summary: Spurious warning with throw in ternary operator
returning const reference
Product: gcc
Version: 4.9.2
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: tavianator at gmail dot com
The following program produces a spurious warning:
$ cat ternary-warning.cpp
const char
foo(const char* ptr)
{
return ptr ? *ptr : throw ptr;
}
$ g++ -Wall -c ternary-warning.cpp
ternary-warning.cpp: In function ‘const char foo(const char*)’:
ternary-warning.cpp:4:29: warning: returning reference to temporary
[-Wreturn-local-addr]
return ptr ? *ptr : throw ptr;
^
[Bug c++/64372] Spurious warning with throw in ternary operator returning const reference
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64372 --- Comment #4 from Tavian Barnes tavianator at gmail dot com --- I meant to include -std=c++11 in the OP, it still happens with that flag set. Good suggestion for the replacement though. The actual code was closer to i length ? a[i] : throw ... but I guess that can become (i length ? a : throw ...)[i].
[Bug middle-end/70054] New: GCC 6 gives a strict-aliasing warning on use of std::aligned_storage
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70054
Bug ID: 70054
Summary: GCC 6 gives a strict-aliasing warning on use of
std::aligned_storage
Product: gcc
Version: 6.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: tavianator at gmail dot com
Target Milestone: ---
GCC 6 warns on this code; GCC 5 didn't.
boost::container::string uses this pattern as part of its SSO implementation,
so I'm getting a lot of new warnings with GCC 6 as a result. As well, I'm
seeing what appears to be a miscompilation as a result, but I haven't reduced
that yet.
$ cat foo.cpp
#include
struct foo {
std::aligned_storage<sizeof(long), alignof(long)>::type raw;
long& cooked() {
return *static_cast<long*>(static_cast<void*>());
}
};
$ g++ -O2 -Wall -c foo.cpp
foo.cpp: In member function ‘long int& foo::cooked()’:
foo.cpp:7:56: warning: dereferencing type-punned pointer will break
strict-aliasing rules [-Wstrict-aliasing]
return *static_cast<long*>(static_cast<void*>());
[Bug middle-end/71177] Spurious -Waggressive-loop-optimizations warning
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71177 --- Comment #4 from Tavian Barnes --- > I remember seeing a similar bugreport. PR57199 is very similar, pretty much an exact dupe actually.
[Bug middle-end/71177] Spurious -Waggressive-loop-optimizations warning
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71177 Tavian Barnes changed: What|Removed |Added Attachment #38516|0 |1 is obsolete|| --- Comment #2 from Tavian Barnes --- Created attachment 38520 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=38520=edit Further reduced testcase > Please don't set regression markers if you don't know what you are doing (but > feel free to suggest that something should have regression markers). > > Happens also in 5.3.0, 4.9.3 and 7.0, so not a regression. Oops, sorry! This only showed up in my actual build on upgrading to GCC 6, but I didn't check the reduced version against older GCCs. > If you could reduce the testcase further and point out at a missing > optimization, the chances of getting this fixed would increase. Otherwise, > someone needs to look at the dump files and figure out why the ranges are not > affected by the check and whether they should. I've attached a further reduced testcase. > The 'else' block is not unreachable, but loop optimizations look at the > possible > ranges within the loop and those may not be taking into account any range > information derived from the caller. The else block is unreachable *from within normalize()*. And since the warning only shows up due to constant propagation from normalize(), it would be nice to avoid it. Specifically, normalize() calls resize(length - 2) which calls (in the unreached else block) append(n - length). (n - length) gets constant-folded to (size_t)-2 which then overflows a ptrdiff_t later.
[Bug middle-end/71177] New: [6 Regression] Spurious -Waggressive-loop-optimizations warning
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71177
Bug ID: 71177
Summary: [6 Regression] Spurious
-Waggressive-loop-optimizations warning
Product: gcc
Version: 6.1.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: tavianator at gmail dot com
Target Milestone: ---
Created attachment 38516
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=38516=edit
Reduced testcase
The following use of boost::container::string gives a warning since GCC 6:
$ cat stringbug.cpp
#include
using boost::container::string;
string normalize(string token)
{
if (token.length() >= 2) {
token.resize(token.length() - 2);
}
return token;
}
$ g++ -Wall -O3 -c stringbug.cpp
In function ‘boost::container::string normalize(boost::container::string)’:
cc1plus: warning: iteration 9223372036854775807 invokes undefined behavior
[-Waggressive-loop-optimizations]
In file included from stringbug.cpp:1:0:
/usr/include/boost/container/string.hpp:2608:10: note: within this loop
for (; first != last; ++dest, ++first, ++constructed){
^~~
cc1plus: warning: iteration 9223372036854775807 invokes undefined behavior
[-Waggressive-loop-optimizations]
/usr/include/boost/container/string.hpp:2626:7: note: within this loop
for ( ; first != last; ++first, ++result)
^~~
What's happening is that resize() is implemented like this:
void resize(size_type n, CharT c)
{
if (n <= this->size())
this->erase(this->begin() + n, this->end());
else
this->append(n - this->size(), c);
}
After inlining/constant propagation, the else block contains undefined
behaviour for resize(length - 2). But the else block is also unreachable due
to the if (length >= 2) check.
Reduced testcase attached.
[Bug middle-end/71002] [6/7 Regression] -fstrict-aliasing breaks Boost's short string optimization implementation
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71002 --- Comment #11 from Tavian Barnes --- Yeah I reported the Boost bug as https://svn.boost.org/trac/boost/ticket/12183.
[Bug middle-end/71002] [6/7 Regression] -fstrict-aliasing breaks Boost's short string optimization implementation
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71002 --- Comment #3 from Tavian Barnes --- Because their long_t is not POD. I don't know why that is though. It could be POD if they removed the default/copy constructors and assignment operator. Actually they're probably worried about custom allocators where the pointer type is not POD. So it couldn't be POD in general, and thus can't appear in a union directly (in C++03).
[Bug middle-end/71002] [6/7 Regression] -fstrict-aliasing breaks Boost's short string optimization implementation
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71002 --- Comment #5 from Tavian Barnes --- > But if it is not POD then assuming it gets copied correctly when > init-constructing a POD union where they placed such object is > an interesting assumption... Hrm? They seem to always copy it manually with long_t's copy constructor: ::new(_repr.long_repr()) long_t(other.m_repr.long_repr());
[Bug middle-end/71002] New: [6 Regression] -fstrict-aliasing breaks Boost's short string optimization implementation
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71002
Bug ID: 71002
Summary: [6 Regression] -fstrict-aliasing breaks Boost's short
string optimization implementation
Product: gcc
Version: 6.1.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: tavianator at gmail dot com
Target Milestone: ---
Created attachment 38438
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=38438=edit
Reduced test case
(This is the miscompilation I reported seeing in PR70054. I have a reduced
test case for it now.)
Since GCC 6, the following simple use of boost::container::string is broken:
$ cat foo.cpp
#include
#include
#include
using boost::container::string;
struct foo
{
__attribute__((noinline))
foo(string str)
: m_str{std::move(str)},
m_len{m_str.length()}
{ }
string m_str;
std::size_t m_len;
};
int main() {
foo f{"the quick brown fox jumps over the lazy dog"};
if (f.m_len == 0) {
std::abort();
}
return 0;
}
$ g++ -O2 -Wall foo.cpp -o foo && ./foo
[1]12277 abort (core dumped) ./foo
It works with -fno-strict-aliasing. I reduced the problem to the attached
standalone test case. Boost's code doesn't seem to be 100% compliant, but the
worst thing it does is access a non-active union member (the is_short
bitfield). As far as I know, GCC permits that as an extension.
[Bug c/78584] Bug in GCC argument parser expandargv
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78584 Tavian Barnes changed: What|Removed |Added CC||tavianator at gmail dot com --- Comment #1 from Tavian Barnes --- Can't reproduce with GCC 6.2.1. The `@file` syntax is used to read command line options from `file`. You can probably write `./@.` if you want to compile a file with that name.
[Bug middle-end/61118] Spurious -Wclobbered warning generated by gcc 4.9.0 for pthread_cleanup_push
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61118 --- Comment #7 from Tavian Barnes --- > the warning is issued for variables which are alive after return from longjmp > but not marked as volatile. Such variables will have undefined value > according to C standard > (http://pubs.opengroup.org/onlinepubs/7908799/xsh/longjmp.html). But this condition is not met: > - They are changed between the setjmp() invocation and longjmp() call.
[Bug middle-end/61118] Spurious -Wclobbered warning generated by gcc 4.9.0 for pthread_cleanup_push
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61118
--- Comment #10 from Tavian Barnes ---
> I think it is - __cancel_arg is assigned inside a while loop
Specifically a do { } while(0); loop, which obviously has only one iteration.
[Bug preprocessor/81515] C pre-processor allows invalid words
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81515 Tavian Barnes changed: What|Removed |Added CC||tavianator at gmail dot com --- Comment #2 from Tavian Barnes --- Is this report about BADTEST or #ele?
[Bug c++/85958] Make const qualifier error clear
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85958 Tavian Barnes changed: What|Removed |Added CC||tavianator at gmail dot com --- Comment #4 from Tavian Barnes --- IMHO "discards qualifiers" and even "discards const qualifier" are still confusing. Making it clearly counterfactual, as in "...would discard (const) qualifier(s)...," would be an improvement. But I'd further argue that "discarding qualifiers" is not really how most people think of this kind of error. When a minor tries to get into a bar, they are not told that "entering this bar discards your age," they are told that "minors aren't allowed." So I think "cannot bind 'const int' to non-const reference type 'int&'" would be more intuitive phrasing.
[Bug tree-optimization/86029] gcc -O3 make very slow product
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86029 Tavian Barnes changed: What|Removed |Added CC||tavianator at gmail dot com --- Comment #1 from Tavian Barnes --- Maybe a dupe of https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70291? In the -O3 version, __mulsc3() dominates the profile. │ for(int i=0; i<=decimate_taps_length; i++) decim += samplebuf[i] * decimate_taps[i]; 0.20 │430:┌─→vmovss 0x4(%r13,%rbx,1),%xmm1 3.63 ││ vmovss 0x0(%r13,%rbx,1),%xmm0 12.35 ││ vmovss 0x4(%r12,%rbx,1),%xmm3 0.31 ││ vmovss (%r12,%rbx,1),%xmm2 0.02 ││ add$0x8,%rbx 36.48 ││→ callq __mulsc3 0.01 ││ vmovss -0x78(%rbp),%xmm6 0.00 ││ vmovss -0x80(%rbp),%xmm4 23.70 ││ vmovq %xmm0,-0x68(%rbp) 14.25 ││ vaddss -0x68(%rbp),%xmm6,%xmm5 1.54 ││ vaddss -0x64(%rbp),%xmm4,%xmm7 0.48 ││ vmovss %xmm5,-0x78(%rbp) 5.92 ││ vmovss %xmm7,-0x80(%rbp) │├──cmp$0x2590,%rbx 0.01 │└──jne430 At -Ofast, │ for(int i=0; i<=decimate_taps_length; i++) decim += samplebuf[i] * decimate_taps[i]; 9.36 │5e0: vpermilps $0xf5,(%r12,%rax,1),%ymm0 15.56 │ vpermilps $0xa0,(%r12,%rax,1),%ymm1 11.24 │ vmulps (%rbx,%rax,1),%ymm0,%ymm0 17.55 │ vpermilps $0xb1,(%rbx,%rax,1),%ymm4 3.31 │ add$0x20,%rax 2.11 │ vmovaps %ymm1,%ymm3 6.62 │ vfmadd132ps %ymm4,%ymm0,%ymm3 3.79 │ vfmsub231ps %ymm4,%ymm1,%ymm0 2.91 │ vblendps $0xaa,%ymm0,%ymm3,%ymm0 10.75 │ vaddps %ymm0,%ymm6,%ymm6 │ cmp$0x2580,%rax 5.59 │ ↑ jne5e0 0.01 │ vmovss 0x258c(%rbx),%xmm0 0.01 │ vmovss -0x70(%rbp),%xmm7 0.01 │ vmovss %xmm5,-0xd0(%rbp) 0.05 │ vextractf128 $0x1,%ymm6,%xmm3 0.01 │ vmovss 0x2588(%rbx),%xmm8 0.03 │ vshufps $0xff,%xmm3,%xmm3,%xmm13
[Bug tree-optimization/86029] gcc -O3 make very slow product
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86029
--- Comment #5 from Tavian Barnes ---
(In reply to Zsolt from comment #3)
> What is the difference between gcc's and clang's __mulsc3?
The important difference is that Clang (and GCC trunk) expand the fastpath
inline, and fall back on __mulsc3 for the more complicated cases. So instead
of
complex float a = b * c;
expanding to
complex float a = __mulsc3(crealf(b), cimagf(b), crealf(c), cimagf(c));
it's more like
complex float a = (crealf(b)*crealf(c) - cimagf(b)*cimagf(c) +
I*(crealf(b)*cimagf(c) + cimagf(b)*crealf(c));
if (isunordered(crealf(a), cimagf(a))) {
a = __mulsc3(crealf(b), cimagf(b), crealf(c), cimagf(c));
}
The fastpath and unlikely branch tends to be faster than the function call.
[Bug middle-end/87647] New: ICE on valid code in decode_addr_const, at varasm.c:2958
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87647
Bug ID: 87647
Summary: ICE on valid code in decode_addr_const, at
varasm.c:2958
Product: gcc
Version: 8.2.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: tavianator at gmail dot com
Target Milestone: ---
$ cat test.c
struct a {
};
struct a *const b = &(struct a){};
int main() {
struct {
char *s;
struct a *t;
} a[] = {"", b, "", b, "", b, "", b, "", b, "", b, "", b, "", b, "",
b, "", b, "", b, "", b, "", b, "", b, "", b, "", b, "", b};
}
$ gcc -O1 test.c
test.c: In function ‘main’:
test.c:8:5: internal compiler error: in decode_addr_const, at varasm.c:2958
} a[] = {"", b, "", b, "", b, "", b, "", b, "", b, "", b, "", b, "",
^
Please submit a full bug report,
with preprocessed source if appropriate.
See <https://bugs.archlinux.org/> for instructions.
$ gcc --version
gcc (GCC) 8.2.1 20180831
Copyright (C) 2018 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
Reduced from a reddit post:
https://www.reddit.com/r/C_Programming/comments/9p44be/internal_compiler_error/
[Bug c/87806] Option -Wall should warn about unused structs, typdefs, enums, etc
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87806 Tavian Barnes changed: What|Removed |Added CC||tavianator at gmail dot com --- Comment #4 from Tavian Barnes --- Perhaps this is reasonable for types that are defined in the file itself, not in an included header?
[Bug tree-optimization/90766] strlen(a + i) missing range for arrays of unknown bound with strings of known length and variable i
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90766 Tavian Barnes changed: What|Removed |Added CC||tavianator at gmail dot com --- Comment #2 from Tavian Barnes --- If char b[] = "123\01234"; then it can't be folded to false for i == 4, right? Presumably g() is only folded because you have [i] not [i].
[Bug libstdc++/90246] std::bad_variant_access messages are not useful
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90246 Tavian Barnes changed: What|Removed |Added CC||tavianator at gmail dot com --- Comment #2 from Tavian Barnes --- > It's too late to change this now, but we could still improve the messages Would it be ABI compatible to make a new exception type that derives from std::bad_variant_access, and throw that instead?
[Bug c++/90205] Wformat-signedness detects %d and suggests %d fixit hint
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90205 Tavian Barnes changed: What|Removed |Added CC||tavianator at gmail dot com --- Comment #8 from Tavian Barnes --- Maybe "argument 2 has type 'double' (promoted from 'float')"?
[Bug tree-optimization/93681] Wrong optimization: instability of x87 floating-point results leads to nonsense
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93681 Tavian Barnes changed: What|Removed |Added CC||tavianator at gmail dot com --- Comment #2 from Tavian Barnes --- Similar to (dupe of?) https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85957
[Bug middle-end/93742] New: Optimization request: pattern-match typical addition overflow checks
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93742
Bug ID: 93742
Summary: Optimization request: pattern-match typical addition
overflow checks
Product: gcc
Version: 10.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: tavianator at gmail dot com
Target Milestone: ---
It would be nice if GCC could pattern-match the typical pattern for
overflow-safe addition and optimize it to just branch on the overflow flag.
Right now, this:
_Bool add_overflow(int a, int b, int *res) {
if ((a > 0 && b > INT_MAX - a) || (a < 0 && b < INT_MIN - a)) {
return 1;
} else {
*res = a + b;
return 0;
}
}
compiles to this:
testl %edi, %edi
jle .L2
movl$2147483647, %ecx
movl$1, %eax
subl%edi, %ecx
cmpl%esi, %ecx
jl .L12
.L4:
addl%esi, %edi
xorl%eax, %eax
movl%edi, (%rdx)
ret
.L12:
ret
.L2:
je .L4
movl$-2147483648, %ecx
movl$1, %eax
subl%edi, %ecx
cmpl%esi, %ecx
jle .L4
ret
#48580 is similar but about multiplication, for which the overflow-safe pattern
is a lot more complicated. But the addition one above is a lot simpler and
pretty widespread I think. For example, it's what CERT recommends:
https://wiki.sei.cmu.edu/confluence/display/c/INT32-C.+Ensure+that+operations+on+signed+integers+do+not+result+in+overflow
[Bug tree-optimization/65752] Too strong optimizations int -> pointer casts
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65752 Tavian Barnes changed: What|Removed |Added CC||tavianator at gmail dot com --- Comment #61 from Tavian Barnes --- (In reply to rguent...@suse.de from comment #44) > The other (unfortunate) thing is that in GCC pointer subtraction > is always performed on integers, thus for the C source code > > int idx = ptr1 - ptr2; > > we internally have sth like > > int idx = ((long)ptr1 - (long)ptr2) / 4; > > so you can't really treat pointers as "escaped" here without loss. It seems possible to distinguish between ptr-to-int casts that actually occur in the source, from ptr-to-int casts that are generated for other reasons by the compiler.
[Bug target/106952] New: Missed optimization: x < y ? x : y not lowered to minss
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106952
Bug ID: 106952
Summary: Missed optimization: x < y ? x : y not lowered to
minss
Product: gcc
Version: 13.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: tavianator at gmail dot com
Target Milestone: ---
Created attachment 53580
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=53580=edit
Assembly from gcc -O3 -S bug.c
The following is an implementation of a ray/axis-aligned box intersection test:
struct ray {
float origin[3];
float dir_inv[3];
};
struct box {
float min[3];
float max[3];
};
static inline float min(float x, float y) {
return x < y ? x : y;
}
static inline float max(float x, float y) {
return x < y ? x : y;
}
_Bool intersection(const struct ray *ray, const struct box *box) {
float tmin = 0.0, tmax = 1.0 / 0.0;
for (int i = 0; i < 3; ++i) {
float t1 = (box->min[i] - ray->origin[i]) * ray->dir_inv[i];
float t2 = (box->max[i] - ray->origin[i]) * ray->dir_inv[i];
tmin = min(max(t1, tmin), max(t2, tmin));
tmax = max(min(t1, tmax), min(t2, tmax));
}
return tmin < tmax;
}
However, gcc -O3 doesn't use minss/maxss for every min()/max(). Instead, some
of them are lowered to conditional jumps which regresses performance
significantly since the branches are unpredictable.
Simpler variants like
tmin = max(tmin, min(t1, t2));
tmax = min(tmax, max(t1, t2));
get the desired codegen, but that behaves differently if t1 or t2 is NaN.
"Bisecting" with godbolt.org, it seems this is an old regression: 4.8.5 was
good, but 4.9.0 was bad.
[Bug target/106952] Missed optimization: x < y ? x : y not lowered to minss
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106952 --- Comment #2 from Tavian Barnes --- (In reply to Alexander Monakov from comment #1) > Note, your 'max' function is the same as 'min' (the issue remains with that > corrected). Whoops, thanks. Also I just noticed that GCC 12.2 does better (but not perfect) with #define min(x, y) ((x) < (y) ? (x) : (y)) #define max(x, y) ((x) > (y) ? (x) : (y)) instead of the inline functions. Doesn't seem to help GCC trunk though.
[Bug sanitizer/113430] New: Trivial program segfaults intermittently with ASAN since Linux 6.7
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113430
Bug ID: 113430
Summary: Trivial program segfaults intermittently with ASAN
since Linux 6.7
Product: gcc
Version: 13.2.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: sanitizer
Assignee: unassigned at gcc dot gnu.org
Reporter: tavianator at gmail dot com
CC: dodji at gcc dot gnu.org, dvyukov at gcc dot gnu.org,
jakub at gcc dot gnu.org, kcc at gcc dot gnu.org
Target Milestone: ---
Since updating to Linux 6.7, I'm getting intermittent segfaults with ASAN and
ASLR enabled.
$ cat foo.c
int main(void) {
return 0;
}
$ gcc -fsanitize=address foo.c -o foo
$ while ./foo; do :; done
AddressSanitizer:DEADLYSIGNAL
=
==337494==ERROR: AddressSanitizer: SEGV on unknown address 0x636c68879e78 (pc
0x7dde493b538f bp 0x sp 0x7ffc78949970 T0)
==337494==The signal is caused by a READ memory access.
AddressSanitizer:DEADLYSIGNAL
AddressSanitizer: nested bug in the same thread, aborting.
tavianator@graphene $ gcc --version
gcc (GCC) 13.2.1 20230801
Copyright (C) 2023 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
$ uname -a
Linux graphene 6.7.0-arch3-1 #1 SMP PREEMPT_DYNAMIC Sat, 13 Jan 2024 14:37:14
+ x86_64 GNU/Linux
Here's the backtrace:
(gdb) set disable-randomization off
(gdb) run
Starting program: /home/tavianator/code/bfs/foo
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/usr/lib/libthread_db.so.1".
Program received signal SIGSEGV, Segmentation fault.
do_lookup_x (undef_name=undef_name@entry=0x761941b3e6d8
"_thread_db_sizeof_pthread", new_hash=new_hash@entry=3872132951,
old_hash=old_hash@entry=0x716f0cc8, ref=0x0,
result=result@entry=0x716f0cd0, scope=, i=0,
version=0x0, flags=3, skip=, type_class=0, undef_map=) at dl-lookup.c:405
405 const ElfW(Sym) *symtab = (const void *) D_PTR (map,
l_info[DT_SYMTAB]);
(gdb) bt
#0 do_lookup_x (undef_name=undef_name@entry=0x761941b3e6d8
"_thread_db_sizeof_pthread", new_hash=new_hash@entry=3872132951,
old_hash=old_hash@entry=0x716f0cc8, ref=0x0,
result=result@entry=0x716f0cd0, scope=, i=0,
version=0x0, flags=3, skip=, type_class=0, undef_map=) at dl-lookup.c:405
#1 0x7619421e20b8 in _dl_lookup_symbol_x (undef_name=0x761941b3e6d8
"_thread_db_sizeof_pthread", undef_map=, ref=0x716f0d58,
symbol_scope=,
version=0x0, type_class=0, flags=3, skip_map=0x0) at dl-lookup.c:793
#2 0x76194197300e in do_sym (handle=, name=0x761941b3e6d8
"_thread_db_sizeof_pthread", who=0x761941afffb3
<__sanitizer::ThreadDescriptorSize()+35>,
vers=vers@entry=0x0, flags=flags@entry=2) at dl-sym.c:146
#3 0x761941973331 in _dl_sym (handle=, name=, who=) at dl-sym.c:195
#4 0x7619418a6ae8 in dlsym_doit (a=a@entry=0x716f0fc0) at dlsym.c:40
#5 0x7619421d94e1 in __GI__dl_catch_exception
(exception=exception@entry=0x716f0f20, operate=0x7619418a6ad0 ,
args=0x716f0fc0) at dl-catch.c:237
#6 0x7619421d9603 in _dl_catch_error (objname=0x716f0f78,
errstring=0x716f0f80, mallocedp=0x716f0f77, operate=,
args=) at dl-catch.c:256
#7 0x7619418a64f7 in _dlerror_run (operate=operate@entry=0x7619418a6ad0
, args=args@entry=0x716f0fc0) at dlerror.c:138
#8 0x7619418a6b75 in dlsym_implementation (dl_caller=,
name=, handle=) at dlsym.c:54
#9 ___dlsym (handle=, name=) at dlsym.c:68
#10 0x761941afffb3 in __sanitizer::ThreadDescriptorSize () at
/usr/src/debug/gcc/gcc/libsanitizer/sanitizer_common/sanitizer_linux_libcdep.cpp:298
#11 0x761941b017ae in __sanitizer::ThreadDescriptorSize () at
/usr/src/debug/gcc/gcc/libsanitizer/sanitizer_common/sanitizer_linux_libcdep.cpp:294
#12 __sanitizer::GetTls (size=0x716f1098, addr=0x7619421b0040) at
/usr/src/debug/gcc/gcc/libsanitizer/sanitizer_common/sanitizer_linux_libcdep.cpp:498
#13 __sanitizer::GetThreadStackAndTls (main=true,
stk_addr=stk_addr@entry=0x7619421b0020, stk_size=stk_size@entry=0x716f10a0,
tls_addr=tls_addr@entry=0x7619421b0040,
tls_size=tls_size@entry=0x716f1098) at
/usr/src/debug/gcc/gcc/libsanitizer/sanitizer_common/sanitizer_linux_libcdep.cpp:595
#14 0x761941af0ff4 in __asan::AsanThread::SetThreadStackAndTls
(this=this@entry=0x7619421b, options=) at
/usr/src/debug/gcc/gcc/libsanitizer/asan/asan_thread.h:77
#15 0x761941af14ee in __asan::AsanThread::Init
(this=this@entry=0x7619421b, options=options@entry=0x0) at
/usr/src/debug/gcc/gcc/libsanitizer/asan/asan_thread.cpp:234
#16 0x761941af
[Bug sanitizer/113430] Trivial program segfaults intermittently with ASAN with large CONFIG_ARCH_MMAP_RND_BITS in kernel configuration
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113430 --- Comment #5 from Tavian Barnes --- (In reply to Xi Ruoyao from comment #3) > Updated the title to make it more precise. > > Note that even with Linux 6.7 the default value of CONFIG_ARCH_MMAP_RND_BITS > is still 28 (32 is set by some distro maintainer who apparently does not > know this will hit the sanitizer runtime), so "since Linux 6.7" is just > misleading. Yep agreed. I didn't expect such a patch from Arch, so I assumed it was a change in the default kernel config. For completeness, here's the Arch bug: https://gitlab.archlinux.org/archlinux/packaging/packages/linux/-/issues/20
[Bug c/112748] New: memmove(ptr, ptr, n) call optimized out even at -O0 with -fsanitize=undefined
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112748
Bug ID: 112748
Summary: memmove(ptr, ptr, n) call optimized out even at -O0
with -fsanitize=undefined
Product: gcc
Version: 13.2.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
Assignee: unassigned at gcc dot gnu.org
Reporter: tavianator at gmail dot com
Target Milestone: ---
This is counter-productive, as I wrote the memmove() specifically to get the
sanitizers to check that ptr really points to a big enough allocation.
$ cat foo.c
typedef __SIZE_TYPE__ size_t;
void *memmove(void *dest, void *src, size_t n);
void foo(void *ptr, size_t n) {
memmove(ptr, ptr, n);
}
$ gcc -O0 -fsanitize=undefined -S foo.c
$ cat foo.s
.file "foo.c"
.text
.globl foo
.type foo, @function
foo:
.LFB0:
.cfi_startproc
pushq %rbp
.cfi_def_cfa_offset 16
.cfi_offset 6, -16
movq%rsp, %rbp
.cfi_def_cfa_register 6
movq%rdi, -8(%rbp)
movq%rsi, -16(%rbp)
nop
popq%rbp
.cfi_def_cfa 7, 8
ret
.cfi_endproc
.LFE0:
.size foo, .-foo
.ident "GCC: (GNU) 13.2.1 20230801"
.section.note.GNU-stack,"",@progbits
[Bug middle-end/112748] memmove(ptr, ptr, n) call optimized out even at -O0 with -fsanitize=undefined
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112748 --- Comment #2 from Tavian Barnes --- (In reply to Andrew Pinski from comment #1) > Does -fsanitize=address remove it? Yes, it's still removed with -fsanitize=address. While ASAN is necessary to check that the memory is really allocated, UBSAN should at least check that ptr is not NULL. So it shouldn't be removed in either case.
[Bug middle-end/94787] Failure to detect single bit popcount pattern
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94787 Tavian Barnes changed: What|Removed |Added CC||tavianator at gmail dot com --- Comment #4 from Tavian Barnes --- (In reply to Wilco from comment #3) > I actually posted a patch for this and popcount(x) > 1 given the reverse > transformation is faster on all targets - even if they have popcount > instruction (since they are typically more expensive). This is true on x86 > as well, (x-1) u (x ^ -x)
[Bug c/71613] Useful warnings silenced when macros from system headers are used
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71613
Tavian Barnes changed:
What|Removed |Added
CC||tavianator at gmail dot com
--- Comment #11 from Tavian Barnes ---
Mostly commenting to make this searchable in case someone else runs into it:
-Wtsan is also affected by this:
$ cat foo.c
#include
int main(void) {
atomic_thread_fence(memory_order_relaxed);
__atomic_thread_fence(__ATOMIC_RELAXED);
return 0;
}
$ gcc -c -fsanitize=thread foo.c
foo.c: In function ‘main’:
foo.c:5:9: warning: ‘atomic_thread_fence’ is not supported with
‘-fsanitize=thread’ [-Wtsan]
5 | __atomic_thread_fence(__ATOMIC_RELAXED);
| ^~~
