Unlocking a mutex can never throw an exception. That doesn't make
sense and our code should make sure it can't happen. Right now,
provided that the system-level functions don't throw, we don't either.
Locking a mutex cannot throw on Linux because we use futexes
directly. A non-recursive mutex is just a futex, whereas a recursive
mutex uses a mutex (a futex) to manage a lock count.
However, on other platforms, due to the freelist, there can be memory
allocation, which means it might throw std::bad_alloc. Not because of
the freelist itself (that uses malloc and will just crash if malloc
fails) but because of Q_GLOBAL_STATIC. In 5.1, the global static will
be noexcept provided the type's constructor is so too (it is, in this
case).
Change-Id: I4c562383f48de1be7827b9afb512d73eaf0792d5
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
There's a comment saying nanosleep's availability is questionable, but
the information of what systems don't have that is now lost in
time. It's quite likely that they were older, Unix systems we no
longer support anyway.
nanosleep comes from POSIX.1b-1993, which is merged into POSIX.1-2001,
so chances are that it's supported almost everywhere where Qt runs
(except for Windows anyway).
Change-Id: I4fd18f8715c43a42429000f3b3d2c3b7343f94b4
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
This requires using the same chain of conditions that QBasicAtomic*
uses in order to provide constructors, so we're using the newly-added
macro QT_BASIC_ATOMIC_HAS_CONSTRUCTORS to check.
Even though QAtomic<> is a template, we can't just use Q_DECL_CONSTEXPR
since the body of the constructors needs to change, too.
Change-Id: I462a80ed175040f7709c30d07d34036c6c5507d8
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This macro is defined when the QBasicAtomic* types have constructors.
It will be used to simplify maintenance of QBasicAtomic subclasses,
as well as classes that aggregate Q(Basic)Atomics and wish to be
literal types if possible.
Change-Id: I7ab181c58a13ec2195449bbae5b3476b9017a8f2
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This allows one to write code that depends on these values at
compile-time.
Change-Id: I7d78524ed9c70d4141360496d1d764dcbfa92e62
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
These methods aren't often used, but currentThreadId() gets used in the
recursive mutex code to store the identity of the thread with multiple
locks.
Change-Id: I031cf3893aa6e7fd7a53733fc1088d295bdb9079
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The code under mutex protection already implicitly handles the case where
queue.empty(), so just removing the pre-lock check suffices.
Change-Id: I01467aff62e61bb4efd16359dd64546ff51c6bad
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
The resulting code is easier to understand, too.
Change-Id: I3cd84d85b3186860dd3ccd67c3771b82695e7f83
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Commit 3ef51efbe7 broke compilation of
qthread_unix.cpp on QNX. This fixes it by passing in the threadId to
setCurrentThreadName().
Change-Id: I24f32d8054baedbd9a65b6a80fb1f6f37e07092d
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
QString::toLocal8Bit() will need to call QTextCodec::codecForLocale(),
which isn't the cheapest of the functions, at least the first time it's
run. So avoid calling it when in most scenarios, the name of the QObject
isn't set, and the information is purely for debugging.
Additionally, avoid allocating memory when setting the thread name to
the class name. The class name coming from the meta object is a static
constant string and we can use it directly.
Change-Id: Ief643bad87a51487b1d41c0a2f323e80bb53e8a7
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Simon Hausmann <simon.hausmann@nokia.com>
Fix some foward references from QtCore in addition. This
will require more work.
Change-Id: Ib1bade18c2cc220a7afe25e9fca6a3f50cb1174b
Reviewed-by: Casper van Donderen <casper.vandonderen@nokia.com>
The qdoc manual currently claims that the command must stand on its own line.
The change follows the consistency with the rest and how the example looks like
inside the qdoc manual for this command.
Change-Id: I6b653dc95cf9d84e4adf32220dace5d313678419
Reviewed-by: Casper van Donderen <casper.vandonderen@nokia.com>
In particular, qEmergencyOut() is now completely exception-free.
Incidentally, this patch shows that Qt isn't consistent in how it
treats empty environment variables used as flags, but that is something
for a separate commit. This patch aims to be behaviour-preserving,
except in exceptional circumstances, of course.
Change-Id: Ie106e7b430e1ab086c40c81cc1e56cd0e5400cb4
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Actually, only the "new" atomics are marked. The old implementation,
based on qoldbasicatomic.h is unchanged, but should still work without a
problem.
The following configurations were tested and do work:
- x86 64-bit
- x86 32-bit
- generic GCC
- generic C++11 std::atomic
- bootstrap
- ARMv6 and 7
- MIPS
- MSVC 2010 32-bit
- MSVC 2010 64-bit
The only two configurations untested are IA-64 and ARMv5. Except for
MSVC, all configurations were tested with GCC 4.6 (MIPS and ARM) and 4.7
(x86 and generics).
Change-Id: Iecbfeacd9d20b535453e91335165e9a221e0b47e
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
These functions are inline (on most architectures) but they contain
inline assembly which the compiler could not understand. In any case,
if it decides not to inline them, it would need to generate exception
handlers.
Change-Id: If9d50793d715c51781e76d0a539da03a6d83e255
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Moreover, exporting it causes the horrible side-effect on Windows
(with MSVC) that the compiler will not inline the functions, but
instead will place indirect calls to the functions in QtCore DLL,
even in release mode. For such a critical piece of code, inlining
is necessary.
Change-Id: Ib31c12f6bf8dc8ece1b51824716a480559753c24
Reviewed-by: Marius Storm-Olsen <marius.storm-olsen@nokia.com>
This is so we can insert valgrind (helgrind) annotation macros. They
require the actual address of the variable to work.
Change-Id: I988f6a46385ad58143c53ad34b6cf0f58be2cdb8
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
QMutexLocker does not support being passed already-locked mutexes,
unless they are recursive mutexes. But in that case, it behaves as if
the mutex weren't locked in the first place.
Since that's the case, there's no point in testing the low bit to see if
it's set or not. It's never going to be.
Change-Id: Ie4b81f7e2cca16e6db36f3cb51a5377dbdfc157d
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Change-Id: I19d3b2e9a5180b13deb828b55195404ef20be295
Reviewed-by: Daniel Teske <daniel.teske@nokia.com>
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
See the discussion of this topic on the mailing list:
http://lists.qt-project.org/pipermail/development/2012-May/003943.html
The consensus is to not have these methods have inline code to actually
acquire the lock (i.e. no atomic test-and-set or similar). QBasicMutex
is unchanged, and continues to have inlined lock(), tryLock(), and
unlock(). QMutexLocker has been changed to always call QMutex::lock()
(even though the constructor takes a QBasicMutex parameter).
Change-Id: Ic7d2d9d581e6b254c84fdfdd8ce6c425535a8078
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Based on reviewer comments on
https://codereview.qt-project.org/#change,25992, split
Q_COMPILER_DEFAULT_DELETE_MEMBERS into
Q_COMPILER_DEFAULT_MEMBERS and Q_COMPILER_DELETE_MEMBERS. Clang provides
2 separate feature checks for these, all other compilers always define
both.
Keep the Q_COMPILER_DEFAULT_DELETE_MEMBERS define, though, for any
existing code that may be using it.
Change-Id: I8969e10989c776a822da153152cfecb99dda8cb9
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Android does not support some parts of POSIX. It's also unfortunately not
POSIX-compliant, meaning we need to guard things like pthread_cancel.
Change-Id: I2979eb5579c822f4f5ba27bcc29e118b908e8bdc
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Even though we really shouldn't, we can get away with using __NR_futex instead.
Done-with: Thiago Macieira <thiago.macieira@intel.com>
Change-Id: I0ba449b740acf2c78825f8093d1515a74f0bc9cd
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This ensures interoperability between the Blackbery C and C++ APIs
and makes it easier to expose platform services in C++ that are
exposed in BPS - since events from both APIs can be processed on
the same thread.
Change-Id: I7270adc64c26396f66d9126141500d5e58be51e7
Reviewed-by: Kevin Krammer <kevin.krammer@kdab.com>
Reviewed-by: Giuseppe D'Angelo <dangelog@gmail.com>
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
The timeout given in milliseconds should be converted to a timespec
struct.
To separate the seconds and nanoseconds the milliseconds are first
multiplied to represent the whole value in an int64 variable.
The calculation is done on integers, so we get an overflow if the
milliseconds are bigger then 2148.
If we cast the given value to an int64 we can avoid this problem.
Fix the used cast.
Task-number: QTBUG-24795
Change-Id: I864ae227cf7dda16a6f30aa4db74acc49e20f6eb
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
This change fixes most qdoc errors in QtCore. There are about 900 left.
The main thing this change does is moving documentation from qtcore from
/doc/src to /src/corelib/doc.
Other issues resolved are mis-use of qdoc commands.
Change-Id: I002d01edfb13575e8bf27ce91596a577a92562d1
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>
Reviewed-by: Jerome Pasion <jerome.pasion@nokia.com>
Replace "contains(QT_CONFIG, coreservices)" with "!ios" in config files.
Replace "QT_NO_CORESERVICES" with "Q_OS_IOS" in source files.
Change-Id: Id3b02316b245a24ce550e0b47596d18a4a409e4f
Reviewed-by: Morten Johan Sørvig <morten.sorvig@nokia.com>
the most-common case is: queue is empty or filled with tasks of
the same priority; so the runnable would be put at the end of queue.
both checks are cheap for us.
also avoid detach()'ing by using const iterators
Change-Id: Iab2255f852211f9accc8d717f778671661210ef3
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
IDEs pick up hints like this, so it makes it more clear to the user
what they need to pass in.
Change-Id: I5de272395fc98391cf43963ce4416a845726bb2c
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Variables that are only used in asserts.
Change-Id: I0984b676ad9b67f5ae72266b8540e8054163918f
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
It's almost exactly the same code in both files, so let's have one file
only. That means we need an #ifdef for the special case of 64-bit types
on i386.
Also take the opportunity to add a comment explaining how this works.
Change-Id: I50d274fa026806ae511b1045aa8a5c25daaa0edc
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
This is the first step in merging the i386 and x86-64
architectures. The next commit will bring i386 support into
qatomic_x86.h.
Change-Id: I24105ea70f3fc29b3fb779a70053f99117440573
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Futexes on Linux can be used across processes, for inter-process
synchronisation. The private flag tells the kernel that this futex is
not used with other processes, so it does not need to check for waiters
outside the current process.
This feature had been proposed in Merge Request 25, but was lost.
Change-Id: Ieafa8b8df0949bd9ae73709b3ec63f7709b0b2a6
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
The qt_global_mutexpool was private API deprecated long time ago.
And there is no reason to call qt_create_tls because it is called in
QThreadData::current that is called from the QObject constructor, even
before QCoreApplication::init can be called.
Change-Id: Idf3576d8591377811b727b12edc43dc898570ba4
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Naming threads is very useful for release builds.
Enabling only on Linux/Mac for now.
The Windows port is using debugger specific API for setting thread
names, so it has to remain debug mode only.
Change-Id: I179521f65f215ff038e8230f958f6aa728ea4cbe
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
QDoc now has support for Doxygen style commands for italics, bold
and list items. This change applies that change in QDoc to the
actual documentation.
Task-number: QTBUG-24578
Change-Id: I519bf9c29b14092e3ab6067612f42bf749eeedf5
Reviewed-by: Shane Kearns <shane.kearns@accenture.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Don't register the thread handle and QThread object pointer to watch list
for adopted thread watcher thread in Windows. Otherwise the watcher thread
will never exit and can't clean up its own QThreadData and QAdoptedThread
object.
Task-number: QTBUG-23961
Change-Id: Ia84326cf3cfd978d2b003ccc1bb6861db950e899
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
Fix potential race condition in data structure of adopted thread watcher
in Windows. Since QVector is not thread safe, it is not safe to read and
append data to qt_adopted_thread_handles or qt_adopted_qthread
simultaneously in qthread_win.cpp. This patch fix this race condition.
Change-Id: I2d0c7a4cdde5390d38d15c98343f0fc6ddd24aba
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
LSB doesn't allow syscalls, so fall back to the normal _unix
implementation
Change-Id: I8aba6147da8b46e3f85b0454cf9aca219811c9fe
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
IBM's POWER and the PowerPC architecture have been merged into a single
ISA, the Power ISA (see http://www.power.org). Use this unified name in
Qt.
Change-Id: Ia41492b0031d890843e43c5f7ecd1e60c65bb75b
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This provides a fallback implementation on UNIX when the Q_PROCESSOR_*
and Q_CC_*/Q_COMPILER_* checks fail to find an implementation.
Note that we always compile qatomic_unix.cpp, but code is only included
when QATOMIC_UNIX_H is defined (meaning the checks above did not find an
implementation).
Change-Id: I8ce047847206003b4fa96eb3fb76b1c2ffbc2dfc
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Make qbasicatomic.h include the OS/compiler/processor dependent
implementation.
For implementations that have not yet been ported to declare a
QAtomicOps, they need to #include <QtCore/qoldbasicatomic.h>, and the
new QBasicAtomicInteger and QBasicAtomicPointer should not be declared.
Change-Id: Ia951834484c9f8dfa75131592e5e716b68ff989b
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This header is not used on Windows anymore, so remove the dead #ifdefs.
Change-Id: I76cfbd13c9fff0eab87cc69e8ca1e0d5ccab9e3a
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This architecture is obsolete and discontinued.
Support for PA-RISC can be re-added if needed, but it would be preferred
to use the GCC intrinsic support from qatomic_gcc.h (on Linux/HPPA, for
example).
Change-Id: I952e521a2c8c68840df0d44843b5487d5c20b135
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Use the new Q_PROCESSOR_* macros to decide which headers to include in
the atomic implementation. This also removes qatomic_arm.h, which isn't
needed anymore, just select the correct qatomic_armv*.h from
qbasicatomic.h
Change-Id: I954848feafb8c420949d066ffcee1dd2b271e13b
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Most of these headers are either forwarding headers, or we explicitly
stop syncqt so that it doesn't generate class includes for the atomic
implementation. Either way, syncqt doesn't see the QT_END_* (and
sometimes not QT_BEGIN_*), which this commit fixes.
Change-Id: Icc8da6f384f38b1ff4eb265c731ce2f2ed92a1a3
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The QT_ARCH_WINDOWS(CE) define(s) aren't used to control which
header to include anymore, so just remove it. We also do not need
the empty src/corelib/arch/windows/arch.pri.
Change-Id: I5400fc852af31907e533d0278540b8cd3da391cb
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
First, we do not need to have the QT_INTERLOCKED_REMOVE_VOLATILE(a)
macro anymore either, since the value stored in QBasicAtomicInteger is
not volatile.
Windows provides Interlocked*Pointer() functions in some configurations,
so we provide a partial specialization of QAtomicOps for pointer types.
For integer types, QAtomicOps selects an implementation based on the
size of the type. At the moment, we only support 32-bit types, but it
will be possible to add 64-bit later.
Note that the 32-bit specialization of QAtomicOpsBySize declares the
Type typedef as long, not int, since the Windows Interlocked*() API
takes parameters as longs and long pointers. Since this typedef differs
from the type given to QBasicAtomicInteger<T> by the QBasicAtomicInt
typedef, we need to templatise the _q_value parameter separately from
the other arguments in QGenericAtomicOps.
This templatisation would be necessary to port other architectures, such
as PA_RISC, where we need to have an int[4] array in the atomic type
while the arguments do not need this array.
Change-Id: Id71fa1ae334da2023553cb402b45e6c285f1d344
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
QBasicMutex is a POD and can be used as a static global object.
in qpicture.cpp factoryLoader is used only once, and under the mutex, so
there is no need for Q_GLOBAL_STATIC for it, it can be a function static
in qhostinfo_unix.cpp the code seemed wrong while compiled with
namespace and QT_NO_GETADDRINFO. I also could get rid of one include
because it was included earlier.
Change-Id: I3c700203c3e067266c20733f4bda8031446dbb86
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
- Always use <qt_windows.h> as the last file to be included.
- Remove it from some headers, use Qt::HANDLE instead of HANDLE.
- Clean up #ifdef, use Q_OS_WIN for Windows/Windows CE.
- Add NOMINMAX to qt_windows.h to avoid problems with the
min/max macros.
- Remove <windows.h> from qplatformdefs.h (VS2005)
Change-Id: Ic44e2cb3eafce38e1ad645c3bf85745439398e50
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
As in the past, to avoid rewriting various autotests that contain
line-number information, an extra blank line has been inserted at the
end of the license text to ensure that this commit does not change the
total number of lines in the license header.
Change-Id: I311e001373776812699d6efc045b5f742890c689
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
The C++11 std::atomic type is very close to our API, to the point one
has to wonder if the committe was inspired by it. It provides all of
the memory semantics that Qt requires and more, plus some
compare-and-swap operations that we don't use.
The idea of returning the actual value in the event of a failed
compare-and-swap is actually quite good, as often we'll retry with
it. We just couldn't come up with a good name (fetchAndTestAndSet?).
The C++11 atomics require that the compiler support constexpr as well,
since std::atomic itself isn't required by the standard to be
trivially-constructible (in fact, it has a constexpr constructor in
the standard). For that reason, we need constexpr so we can add a
constructor to QBasicAtomic too.
Change-Id: I12c51455ba73350a6f7501aacc2ca7681c4369dd
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
With this implementation, we can have Qt run on any architecture that
GCC supports without having to write specialised code. However, on
some architectures, the code that GCC generates is not optimal: it
uses locking on ARMv5 and it's always fully-ordered. For that reason,
it appears after the Qt native assembly implementations (it's a
fallback, not an override).
Since they all have fully-ordered memory semantics, we define only the
xxxRelaxed functions. The exception is __sync_lock_and_test, which has
acquire semantics, so we need to define the Release and Ordered
versions too.
On some architectures, GCC can support atomics on types different than
32-bit and pointer-sized (like x86, x86-64, ARM and even
MIPS). However, there's no standardised way of telling: GCC seems to
define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8} if those operations
are present, but I couldn't find it on the ARM compiler (it was there
for i386, x86-64, IA-64 and MIPS).
Change-Id: I55ff7a7c0cfc6388b7ad8e2c0dedecffdf2a3e01
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
These defines were there to aid in the commercial
licensing scheme we used long ago, and are no longer needed.
Keep a QT_MODULE(x) define so other modules continue compiling.
Change-Id: I8fd76cd5270df8f14aee746b6cf32ebf7c23fec7
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
The IA-64 architecture supports the actual memory ordering semantics
in many instructions, but not all. We actually implement the functions
for all operations, so we get the best possible output.
It does support proper load-acquire and store-release semantics, but
we don't need instructions for it: the ABI requires that a volatile
load be acquire and a volatile store be release.
The Intel and HP compiler codepaths are rewritten, but untested.
Change-Id: I7aa62a4ec65f63a97d1bbd8418bb2492c2be465f
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The LL/SC instructions are only present on MIPS II and up, so don't
pretend to support MIPS I. The previous implementation emitted the
instructions by telling the assembler to change instruction sets. Now,
the user must pass an -march= option to GCC telling it which
architecture or processor is being targetted.
On MIPS64, the 64-bit implementation allows supporting for long long
too.
Change-Id: I6dae6f8f61e563aba6a663227d91c5ddf554aa6a
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The implementation is divided in two files, as it used to be in the
previous implementation: one for ARMv5, one for ARMv6 and up.
For the ARMv5 implementation:
Drop the non-Linux EABI version of the atomics, as it's not
ABI-compatible with the ARMv6 and ARMv7 implementations. This
means this ARMv5 implementation only works on Linux. If other
systems implement kernel helpers like Linux, they can be added
too.
We use the __kernel_cmpxchg located at 0xffff0fc0 to implement the
operations, except for fetchAndStore, for which we use the SWP
instruction.
Also introduce the use of __kernel_dmb (at 0xffff0fa0) for the
memory barrier. Now this code is SMP-safe even when built with
ARMv5.
The kernel cmpxchg helper was introduced in Linux 2.6.12, whereas
the dmb helper was introduced in 2.6.15. That means 2.6.15 is the
minimum version now.
For ARMv6 and up:
Introduce byte, half-word and doubleword atomics that work on
ARMv6K and up.
For ARMv6 specifically, the memory barrier instruction (DMB) isn't
present, so we need to accomplish the same with the MCR
coprocessor instruction.
Change-Id: Ife7f9b920bcc7d1eef7611761f6c59ea940ec7df
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Both implementations now are very similar to one another, to the point
we could share the code if we wanted to. They are based on assembly
code for the Relaxed functions only, as the i386 and x86-64
architectures only allow for full memory ordering or something that
closely resembles it (see 8.2 "Memory ordering" in the Intel 64 and
IA-32 Architectures Software Developer's Manual Volume 3A). We could
add "lfence/mfence/sfence" in future versions if we wanted to (SSE2+).
Change-Id: I76966d9f8694edfece2c5ebd3387348fac721447
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The code is now much simpler and much more complete. Now the bootstrap
atomics (which aren't atomic at all) contain the full set of
operations.
The only integer supported is int, but all others would work too.
Change-Id: Id99f07818f9da059c4ff02520f9cbf2d1a71a514
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Contact point is the Qt Project, and needs to be included for
all new files added to the repository.
Change-Id: Id0e7219e1d11a169f1a91439728cbda55ab29eeb
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
QBasicAtomicPointer is forward declared as a class, keep the actual
declaration of QBasicAtomicInteger and QBasicAtomicPointer as class with
all public members (qoldbasicatomic.h does the same).
src/corelib/thread/qbasicatomic.h:158:1: warning:
'QBasicAtomicPointer' defined as a struct template here but
previously declared as a class template [-Wmismatched-tags]
struct QBasicAtomicPointer
^
src/corelib/global/qglobal.h:1861:23: note: did you mean struct here?
template <typename T> class QBasicAtomicPointer;
^~~~~
struct
Change-Id: I38c59c29d7f796dde772e7f403bbf98b04571a08
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Public headers should compile with QT_NO_KEYWORDS defined.
Change-Id: I5620b4b2600f5e39bb402b97d14fdb257dfe9942
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
Reviewed-by: Jonas Gastal <jgastal@profusion.mobi>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
The new implementation is API- and ABI-compatible with the old
implementation, provided that QBasicAtomicInt isn't used as an
argument in a function call or the return value: now, QBasicAtomicInt
is a typedef to QBasicAtomicInteger<int>.
The new design is based on CRTP: the QGenericAtomicOps template class
takes as a template parameter the derived class itself. This way, we
implement a "poor man's virtual" without a virtual table and
everything is inline.
QGenericAtomicOps implements most of the atomics code that is repeated
in many classes all over:
* Acquire semantics are obtained by placing an acquire barrier after
the Relaxed operation
* Release semantics are obtained by placing a release barrier before
the Relaxed operation
* Ordered semantics are obtained by placing an ordered barrier before
the Relaxed operation (either way would be fine)
* fetchAndStoreRelaxed and fetchAndAddRelaxed are implemented on top
of testAndSetRelaxed
* ref and deref are implemented on top of fetchAndAddRelaxed
It also adds load, loadAcquire, store and storeRelease: the default
implementations of loadAcquire and storeRelease operate on a volatile
variable and add barriers. There are no direct operators for accessing
the value.
Each architecture-specific implementation can override any of the
functions or the memory barrier functions. It must implement at least
the testAndSetRelaxed function.
In addition, by specialising one template class, the implementations
can allow QBasicAtomicInteger for additional types (of different
sizes). At the very least, int, unsigned and pointers must be supported.
Change-Id: I6da647e225bb330d3cfc16f84d0e7849dff85ec7
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
After commit 79f675a1e0, the connect and
disconnect callback API effectively becomes useless. Now that these
callbacks cannot/do not, it makes little sense to keep the backdoors
added for Qt Jambi support.
Remove them for now. Should the Qt Jambi team want/need to port to Qt 5,
we can re-add them, possibly designing a better API for doing so as
well.
Change-Id: I6209a1d647d683c979d5294b632b8c12c0f9f91c
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
This is a source incompatible change. There is concern that the
convenience of the implicit cast and assignment operators can lead to
misuse. Several commits have already been done that remove excess use
of the implicit cast, which is a *volatile* read every time it's used.
Users of the QAtomic* API should have to think about when they are
loading the value, and if they do or don't need the acquire memory
barrier on load. The code that people would write using this API is
meant to be multi-threaded, concurrent, and correct. The API should not
allow them to inadvertently, possibly unknowingly, shoot themselves
in the foot.
SC-break-rubber-stamped-by: Lars Knoll
Change-Id: I88fbc26d9db7b5ec80a58ad6271ffa13bbfd191f
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
... and use it in QCoreApplication::notifyInterna() instead of
"reinventing" the wheel there. The constructor and destructor for
QScopedLoopLevelCounter are inline, so the class does not need to be
exported.
Change-Id: I7af5a4ef0987f277bcc50c1057f3c74152f3d79d
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
QAbstractEventDispatcher() does no longer install itself into the current
thread. Instead the new methods QThread::setEventDispatcher() and
QCoreApplication::setEventDispatcher() allow to install a custom event
dispatcher into any QThread as long as there is no default event dispatcher
created yet. That is, before the thread has been started with
QThread::start() or, in case of the main thread, before QCoreApplication has
been instantiated.
Change-Id: I7367e13d8d8aebed5a5651260bb69b8818eb1b90
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
and isRunning() should return false.
This restore the Qt 4.7 behaviour
In Qt 4.7, the finished() was called with the thread's intenal mutex
locked. Which mean that:
- Call to isRunning or isFinished called from a slot connected to
finish within the thread would deadlock. (Hence no compatibility
to keep here)
- Call to isRunning or isFinished from a slot connected with
QueuedConnection in another thread would lock the mutex until
the destructors are finished. and then return as if the thread have
finished.
Change-Id: I963eccae8f7634aff90cc4bbab6ca886a78e35eb
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
QAtomic* has a copy constructor, so it may make sense to allow
assignment of one atomic variable to another.
Change-Id: Ic754d13765080e2fcd13dc583940e354ad4404cd
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
... since the latter is now deprecated and will be removed.
Change-Id: I456c1bf93ebf119c028bc4a63f1f8a31f069b83b
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
QMutexPrivate takes more memory than necessary, and also initialize
platform specific ressources.
Change-Id: I70be1b89b1c21499645785ae47693a6b2514e28b
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Because we use d as a local variable.
We used this->d to refer it, but this can be confusing to have twice the same
name
Change-Id: I570aa5f444ada358eb456d6b3d9b8bfa60b10bbf
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Those are already outdated and do not compile anymore
(QMutex has changed too much)
Better to remove that dead code so it do not show up in grep anymore
Change-Id: I096e7a73e23cbb77050843c2e1c10929086fdb8f
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Add Q_DECL_DEPRECATED to all methods that are pending removal once
http://codereview.qt-project.org/#change,6243 is merged.
This is necessary to give people time to react to the source
incompatible change.
Change-Id: Ia72f184d630d593f96dd1d95ad6cd0d3bd5d811c
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Q_WS_QPA is the only active code path after merging
refactor, other Q_WS-macros are no longer used.
Enable compilation without -qpa.
- Remove Q_OS_MSDOS, Q_OS_OS2
- Remove Q_WS_QWS
- Remove/replace definitions/conditionals of Q_WS_XX
- Remove qpa branches from profiles
- Replace Q_WS_MAC by Q_OS_MAC
- Replace Q_WS_MAC && !Q_WS_QPA by
Q_OS_MAC && !QT_NO_CORESERVICES
- Similarly in profiles: mac:contains(QT_CONFIG, coreservices)
- Replace Q_FS_FAT by Q_OS_WIN
Change-Id: Icce5a6c55b052c8f72b3b979ddf31a4f388ea9c9
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
When using methods from a template base class, the lookup needs to be
qualified. See http://clang.llvm.org/compatibility.html#dep_lookup_bases
Change-Id: I5b7cd71e0d45414ac0eff97fe9ba5d3ccd5bd9e6
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
Reviewed-by: Thiago Macieira (Intel) <thiago.macieira@intel.com>
Reviewed-by: Olivier Goffart <ogoffart@kde.org>
Now, users of QBasicAtomicInt and QBasicAtomicPointer must be sure to
use .load() and .store() to access the values.
Change-Id: I6b48ed175618baf387dd38d821bd50e6e93c082e
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Change Ie3271abd1728af599f9ab17c6f4868e475f17bb6 didn't build on Mac in
debug mode. This commit fixes the build.
Change-Id: I925af83ea577228b60679af804aa7875a779ec3f
Reviewed-on: http://codereview.qt-project.org/6331
Sanity-Review: Qt Sanity Bot <qt_sanity_bot@ovi.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Most of these changes are search-and-replace of d->ref ==, d->ref !=
and d->ref =.
The QBasicAtomicPointer in QObjectPrivate::Connection didn't need to
be basic, so I made it QAtomicPointer.
Change-Id: Ie3271abd1728af599f9ab17c6f4868e475f17bb6
Reviewed-on: http://codereview.qt-project.org/5030
Reviewed-by: Qt Sanity Bot <qt_sanity_bot@ovi.com>
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
This should be replaced later with proper loadAcquire and storeRelease
functions in the architectures that require them. The code right now
is enough for x86, x86-64 and IA-64. It's probably not good for ARM
and MIPS.
Change-Id: Iaa28a18cad64772539662439c2c14690ec26940f
Reviewed-on: http://codereview.qt-project.org/5029
Reviewed-by: Qt Sanity Bot <qt_sanity_bot@ovi.com>
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
And added a POD QBasicMutex. (QBasicMutex* can safely be
static_cast'ed to QMutex*)
The d pointer is not anymore always a QMutexPrivate.
If d == 0x0: the mutex is unlocked
If d == 0x1: the mutex is locked, uncontended
On linux:
if d == 0x3: the mutex is locked contended, waiting on a futex
If d is a pointer, it is a recursive mutex.
On non-linux platforms:
When a thread tries to lock a mutex for which d == 0x1, it will try to
assing it a QMutexPrivated (allocated from a freelist) in order to wait
for it.
Change-Id: Ie1431cd9402a576fdd9a693cfd747166eebf5622
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-on: http://codereview.qt.nokia.com/2116
Reviewed-by: Qt Sanity Bot <qt_sanity_bot@ovi.com>
Reviewed-by: Olivier Goffart <olivier.goffart@nokia.com>
The commit 412ef92162 introduced a fix, but the
fix was incomplete. Fix it for good.
Change-Id: I3e7fbdb294f8e960fbbf2e830790750240ed813a
Merge-request: 30
Reviewed-by: Olivier Goffart <olivier.goffart@nokia.com>
Reviewed-on: http://codereview.qt.nokia.com/1991
Reviewed-by: Qt Sanity Bot <qt_sanity_bot@ovi.com>
Recalculate how much time is remaining.
This commit needs to be backported to 4.8.
Change-Id: Ib587335bb90306e65969bb26256fb388f8f6bd24
Merge-request: 20
Reviewed-by: Olivier Goffart <olivier.goffart@nokia.com>
Reviewed-on: http://codereview.qt.nokia.com/1666
Reviewed-by: Qt Sanity Bot <qt_sanity_bot@ovi.com>
Some functions (such as QObject::moveToThread) did not keep
the event ordered by priority.
And because qUpperBound is used to add events, that mean new
events would not be inserted in order.
Task-number: QTBUG19637
Change-Id: I38eb9addb1cdd45b8566e000361ac6e5f1f2c2b8
Reviewed-on: http://codereview.qt.nokia.com/733
Reviewed-by: Qt Sanity Bot <qt_sanity_bot@ovi.com>
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Olivier Goffart <olivier.goffart@nokia.com>
After the mutex optimizations on Mac, we did not handle the case where
semaphore_wait() could return KERN_ABORTED. Under heavy contention, this
happens, and when running in release mode, the assert in qmutex.cpp is not
executed. The code silently allows multiple threads to continue as if it
had acquired the mutex exclusively.
Fix this by checking for KERN_ABORTED from semaphore_wait(), and retry the
wait. We do not handle KERN_ABORTED for timed waits, simply return false and
let the code doing the tryLock() handle it how it deems best.
Reviewed-by: joao
(cherry picked from commit b54af0a9d6406356616889826e31925d2fa05718)
The thread callback doesn't align the stack on 16-bytes on WinXP.
That causes a crash when we call SSE code. So now we tell the
compiler to force that alignment of the stack.
Task: QTBUG-18631
Reviewed-By: Olivier
(cherry picked from commit 364ce5b7f5379499562b4f4f5a68da7ba068fe1e)
The native thread implementation in Qt 4.8 did not call
CTrapCleanup::New() which resulted in E32USER-CBASE 69 panics
in applications when they tried to use the cleanup stack in
a thread's run() function. In 4.7 this was working because
OpenC's pthread implementation created a CTrapCleanup automatically.
Now we do it also in the native Symbian thread implementation.
Trask-number: QTBUG-18822
Reviewed-by: Murray Read
(cherry picked from commit 41aa023ef6019ac9745b780c953f48b8bbc42a42)
This is the beginning of revision history for this module. If you
want to look at revision history older than this, please refer to the
Qt Git wiki for how to use Git history grafting. At the time of
writing, this wiki is located here:
http://qt.gitorious.org/qt/pages/GitIntroductionWithQt
If you have already performed the grafting and you don't see any
history beyond this commit, try running "git log" with the "--follow"
argument.
Branched from the monolithic repo, Qt master branch, at commit
896db169ea224deb96c59ce8af800d019de63f12
Thiago Macieira
Marc Mutz
Sean Harmer
Lars Knoll
Laszlo Papp
Jeremy Katz
Sergio Ahumada
Bradley T. Hughes
Robin Burchell
Jeff Kehres
Markus Rathgeb
Sean Harmer
Casper van Donderen
Ian Dean
Konstantin Ritt
Oswald Buddenhagen
Stephen Kelly
Friedemann Kleint
Harald Fernengel
Olivier Goffart
Robin Burchell
Leonard Lee
jian liang
Jason McDonald
Thiago Macieira
Denis Dzyubenko
Gunnar Sletta
Robin Burchell
Sven Anderson
Olivier Goffart
David Faure
Samuel Rødal
Olivier Goffart
Paul Olav Tvete
Jyri Tahtela
Olli Werwolff
David Boddie
Thierry Bastian
Laszlo Agocs