QDateTime takes dd and yyyy. It is confusing when the example format
strings don't work because they use DD YYYY, and it is a silent error
with the characters not getting substituted and can be easy to miss.
Improved Qt::ISODate, TZD isn't a format character so skip writing out
a bad format example.
Task-number: QTBUG-62111
Change-Id: Ia61d561074ae885fc0a99238d93cb34aaa9953bb
Reviewed-by: David Fries <david@fries.net>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Found while working on suppressing the warning about the return value
(which is either 0 or -1) was being ignored.
Task-number: QTBUG-61968
Change-Id: I02d22222fff64d4dbda4fffd14d148b1724547ca
Reviewed-by: Florian Bruhin <qt-project.org@the-compiler.org>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
__has_cpp_attribute(nodiscard) is 1 in all compilation modes, but if you
use it outside of C++1z, you get a warning.
LLVM-bug: https://bugs.llvm.org/show_bug.cgi?id=33518
Task-number: QTBUG-61840
Task-number: QTBUG-62085
Change-Id: I84e45059a888497fb55ffffd14d3683f4808978b
Reviewed-by: Eike Ziller <eike.ziller@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
In the past, we had an undocumented text flag that worked with
one of the QPainter::drawText() overloads. This was never intended
as public API and served a specific cause in Qt WebKit at one point.
But there is a general need for such API, as disabling shaping features
easily gives 25% performance improvement on text rendering even for
fairly short strings.
This patch adds a new style strategy flag to disable shaping and
will just uses the CMAP and HDMX tables to get glyph indices and advances
for the characters. In Qt 6, the TextBypassShaping flag can be removed
completely and be replaced by the style strategy.
[ChangeLog][QtGui][Text] Added QFont::PreferNoShaping style strategy to support
improvements to performance at the expense of some cosmetic font features.
Task-number: QTBUG-56728
Change-Id: I48e025dcc06afe02824bf5b5011702a7e0036f6d
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
This removes a lot of duplicated code that existed in both qglobal.cpp
and qsystemerror.cpp, including the hack to get the correct strerror_r
signature.
This removes the incorrect use of EACCES, EMFILE, ENOENT, and ENOSPC
from qt_error_string on Windows. qt_error_string is supposed to be used
only with Win32 error codes from GetLastError(), despite there being a
lot of uses in cross-platform and even Windows-specific code that pass
errno constants.
It may or may not work: that depends on whether the constants happen to
match. ENOENT matches ERROR_FILE_NOT_FOUND and one could argue that
ENOSPC matching ERROR_OUT_OF_PAPER is acceptable, but EMFILE isn't the
same as ERROR_BAD_LENGTH nor is EACCES, ERROR_INVALID_DATA.
Change-Id: I1eba2b016de74620bfc8fffd14cccb7f77f4b510
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Kai Koehne <kai.koehne@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The classes have recently been cleaned up to be ready for being made
public, and this change adds the final documentation. The classes are
already used in qjson, qtdeclarative and some image format decoders.
Change-Id: Ib6df89f1119162c5bad7a08e9b6f843a2db4040f
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Doesn't affect our current builds since it's just a marker for the
linker on what sections should be merged. Unless you're mixing
namespaced and non-namespaced static builds into one executable.
Change-Id: Ia53158e207a94bf49489fffd14c7bc294fccf8f9
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
First, it can never return, so we can mark it Q_NORETURN and add an
std::termianate at the end. Though if it did, we'd end up in a null-
pointer dereference crash in the caller.
Second, add Q_DECL_NOTHROW to it. It can't throw, but it terminates
execution. This also prevents both puts and fprintf from escaping via
pthread asynchronous cancellation on Linux/glibc.
Third, don't use QMessageLogger, since that allocates memory and
actually uses QString. If we really are in an OOM situation, then
QString's failed allocation would recurse back into qt_check_pointer.
We'd compound the OOM situation with a stack overflow...
Change-Id: Ia53158e207a94bf49489fffd14c81c47971f4e82
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
qt_assert and qt_check_pointer get the function name and line number
from the caller (the functions are called from the Q_ASSERT and
Q_CHECK_PTR macros, respectively), so we don't need to capture the
context from those two functions. Instead, pass the context to
QMessageLogger for proper logging.
I've left the file name and line number in the assertions, for users who
did not add them to their message log pattern, but I've removed from the
almost never used qt_check_pointer function.
Note: how useful is it that we allocate memory in response to failing to
allocate memory?
Change-Id: Ia53158e207a94bf49489fffd14c81b359c5b6537
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Kai Koehne <kai.koehne@qt.io>
The getentropy function, first found in OpenBSD, is present in glibc
since version 2.25 and Bionic since Android 6.0 and NDK r11. It uses the
Linux 3.17 getrandom system call. Unlike glibc's getrandom() wrapper,
the glibc implementation of getentropy() function is not a POSIX thread
cancellation point, so we prefer to use that even though we have to
break the reading into 256-byte blocks.
The big advantage is that these functions work even in the absence of a
/dev/urandom device node, in addition to a few cycles shaved off by not
having to open a file descriptor and close it at exit. What's more, the
glibc implementation blocks until entropy is available on early boot, so
we don't have to worry about a failure mode. The Bionic implementation
will fall back by itself to /dev/urandom and, failing that, gathering
entropy from elsewhere in the system in a way it cannot fail either.
uClibc has a wrapper to getrandom(2) but no getentropy(3). MUSL has
neither.
Change-Id: Ia53158e207a94bf49489fffd14c8cee1b968a619
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
QT_CONFIG and some other macros are unavailable there.
Change-Id: Ia53158e207a94bf49489fffd14c8d306e2dbd9d2
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
According to qcompilerdetection.h, the Dinkumware C++ library does not
have std::initializer_list, even though the compiler supports it.
Add the missing Q_COMPILER_ guards.
Change-Id: I84a7d5054c00dba38bcde15e277ceb0ee05e6cd7
Reviewed-by: Jake Petroules <jake.petroules@qt.io>
Reviewed-by: James McDonnell <jmcdonnell@blackberry.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The values do not need to be 2 x 32-bits. Eight bits suffice for a long time to come.
Should save 189 bytes in text size on Android builds.
Change-Id: I78e31e7caa7a698f41c66d7bbac58a766c6e8834
Reviewed-by: Jake Petroules <jake.petroules@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The copy constructor is not disabled by any user-defined move or copy
special member function, and thus does not need to be = default'ed.
Change-Id: I90586d25756885ac77f0946c147079efb5d1b1e0
Reviewed-by: Jake Petroules <jake.petroules@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Makes the qle_bitfield template more generic and moves it to qendian_p.h
It is also hardened to be more reliable.
Change-Id: I53214ec99cceee4f5e8934ae688c99e555a5fb42
Reviewed-by: Ville Voutilainen <ville.voutilainen@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
qdeadlinetimer.cpp and qelapsedtimer_*.cpp are not part of the bootstrap
library. Instead of adding them there, just remove this entropy source
from bootstrapped builds (instead of just qmake).
To compensate, store all the bits instead of trying to combine them into
just one 32-bit word. We've got a few new sources from the stack and
libc, plus two more ELF auxvec values that the Linux kernel supplies
(inspired by OpenBSD's getentropy_linux.c, which is used in Bionic).
Task-number: QTBUG-61492
Change-Id: I1d5a585d4af842f9a66ffffd14c999ae8c44f46c
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
The endian integers have more use than just in QJson, and is already
used separately by QtDeclarative. This patch moves q_littleendian out
of qjson_p.h and matches it with a corresponding q_bigendian, and puts
it to use in simplifying endian handling in the ico image handler.
Change-Id: I975bb701a089256db8ced3cb53b4bd62cdfb02dd
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Marks the bswap methods constexpr and removes the special versions
explicitly using builtins as both clang and gcc automatically
recognizes bswap code and replaces it with builtins.
Change-Id: I07258431ac2588f6046682ffb9c832e650faf66a
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Found this in a few uses of qrand() that assumed the result would be
non-negative.
Change-Id: Ia53158e207a94bf49489fffd14c7c029515cf42c
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
GCC didn't support it until version 5 or 6, so add configure tests for
both <random> and <sys/auxv.h>. Normally I'd say "upgrade", but this is
too low-level and important a feature.
There's a good chance that all our supported compilers have <random>
anyway. As for <sys/auxv.h>, it's present on Glibc, Bionic and MUSL, but
I don't see it in uClibc (AT_RANDOM is a Linux-specific feature).
Change-Id: Ia3e896da908f42939148fffd14c5b2af491f7a77
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
it contains no code.
Change-Id: Ie8a43abb2db3d040f7046206adf2bf555960dd9c
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The Linux kernel gives us 16 bytes of random data and sets a pointer to
it in the ELF AuxV (the same one that allows us to get HWCAP on ARM
systems). So if we end up in the fallback generator, at leat we'll get a
good amount of entropy to seed the Mersenne Twister.
This could happen if the application is run in a chroot(2) or container
without /dev/random or /dev/urandom. That is probably an installation
mistake, so we don't optimize this case for performance.
With this commit, we have now good, high-quality fallbacks for Windows
(rand_s), for BSDs (arc4random) and for Linux. The only missing,
supported OS without a good entropy source is QNX.
Change-Id: Ia3e896da908f42939148fffd14c5b1084051f1a8
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
If we've never read any samples from hardware RNG or from /dev/urandom,
then there are no stored samples for us to seed the Mersenne Twister. In
that case, attempt to obtain an emergency sample that consists of some
random bits from the variable addresses (ASLR should help) and from the
clock (using nanosecond quality, instead of seconds).
There's still the possibility that we'll get poor entropy: very close to
boot, if the kernel entropy pool is empty, ASLR could be poor and the
monotonic clock value could be consistent from boot to boot. There's
nothing we can do about that.
Change-Id: Ia3e896da908f42939148fffd14c5b0c7b608371b
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
This function works even in chroots where /dev is not available. Since
it's part of the OS, it may have better ways of gathering entropy even
when we don't.
Change-Id: Ia3e896da908f42939148fffd14c5afedc8569dbc
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
This commit adds support for the x86 RDRAND instruction for
QRandomGenerator. This is the same that libstdc++-v3 uses for
std::random_device() by default. If it fails because the hardware does
not have enough entropy collected, we fall back to the operating system
generator, which often has more entropy collected from other sources.
Change-Id: Icd0e0d4b27cb4e5eb892fffd14b5167214e1ea3f
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
This class provides a reasonably-secure random number generator that
does not need seeding. That is quite unlike qrand(), which requires a
seed and is low-quality (definitely not secure).
This class is also like std::random_device, but better. It provides an
operator() like std::random_device, but unlike that, it also provides a
way to fill a buffer with random data, not just one 32-bit quantity.
It's also stateless.
Finally, it also implements std::seed_seq-like generate(). It obeys the
standard requirement of the range (32-bit) but not that of the algorithm
(if you wanted that, you'd use std::seed_seq itself). Instead,
generate() fills with pure random data.
Change-Id: Icd0e0d4b27cb4e5eb892fffd14b4e3ba9ea04da8
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Change-Id: Id6533c8a444854f6215f6e47000875ef9751905b
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@qt.io>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Commit d56c6cf7a4 was incorrect. It was a
nice try, but on a 64-bit Mac machine (x86_64 CPU), it returned
hw.cputype = 7, which is CPU_TYPE_X86. CPU_TYPE_X86_64 is only used in
Mach-O slices for fat binaries and does not reflect hw.cputype.
Task-number: QTBUG-61205
Change-Id: Ia3e896da908f42939148fffd14c54b3050b8e64b
Reviewed-by: Jake Petroules <jake.petroules@qt.io>
Clang's definition of the __EXCEPTIONS macro is inconsistent across
platforms. When compiling for Darwin, Clang 3.6 and newer will set the
token when exceptions are enabled in either C++ or ObjC. This change
adds the reliable check described in the Clang 3.6 release notes to
ensure that QT_NO_EXCEPTIONS is defined when required.
The check requires the use of the Clang-specific __has_feature()
syntax for which a new proxy macro QT_HAS_FEATURE(x) is added in
qcompilerdetection.h
Task-number: QTBUG-61034
Change-Id: Ie7b482dfa1a4a5b700a6b97562c26b626be1fc04
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The documentation says that it's equivalent to
qgetenv(varName).toInt()
But the implementation wasn't. QByteArray::toInt() verifies that the
entire string was consumed, so QByteArray("1a").toInt() == 0, but
qstrtoll alone doesn't. That is, qstrtoll("1a", ...) == 1.
The implementation also detected the base, a behavior I kept. Instead, I
updated the documentation.
Change-Id: I0031aa609e714ae983c3fffd14676ea6061a9268
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
The operator double() and operator long double() members of qfloat16
are causing cast ambiguities. This removes them, leaving only
operator float() which seems to be adequate.
Also, additional arithmetic operator tests were added which without
this removal fail to compile.
Change-Id: Id52a101b318fd754969b3de13c1e528d0aac2387
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The definitions of size_t and ptrdiff_t ([support.types.layout] p2 and
p3 respectively) do not specify that they need to be as big as a
pointer. They just need to be big enough to hold the size of the largest
object and the biggest array subscript, respectively, the platform
supports (e.g., 16-bit DOS would have them as 16-bit in all memory
models, except huge).
But we depend on them actually being the size of a pointer in many
places, such as in QArrayData::offset, that stores the linear distance
from the end of the structure to the beginning of the data, wherever it
is in memory.
It's also a good idea to verify that qptrdiff and qssize_t are the same
type.
Change-Id: I9ad33fff8b634979bdbafffd14bbd1223afc58e8
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
We can't depend on QT_HAS_INCLUDE for such an important functionality in
QtQml, so detect at configure time.
alloca() is not a POSIX function (it apparently first appeared in
Version 32V AT&T UNIX), so the actual header that defines it varies from
system to system. Clearly, if alloca.h exists, that's the one, so we try
it first. On most other systems that don't define it, it's in stdlib.h.
The only exception is Windows, where it's actually defined in malloc.h.
Task-number: QTBUG-59700
Started-by: Thiago Macieira <thiago.macieira@intel.com>
Change-Id: Icd0e0d4b27cb4e5eb892fffd14b4b2b389a4684e
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
It was added to support qAsConst(). When <type_traits> became mandatory,
porting qAsConst() to std::add_const was forgotten.
Change-Id: Ifb9b54d12554ce19dca4664642a8644f49aeb6af
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
Tablet events compression has been missing since commit
60cd1c6775 in 5.6.1, and there's
no way to bring the old behavior back. Since tablet events are
not taken into account by AA_CompressHighFrequencyEvents on
purpose, we introduce this new flag.
This new flag is conditional to AA_CompressHighFrequencyEvents
being set in order for it to be effective.
[ChangeLog][QtCore][Tablet support] If the application attribute
AA_CompressTabletEvents is set in addition to AA_CompressHighFrequencyEvents,
even the QTabletEvents will be compressed (only on the X11 platform so far).
AA_CompressHighFrequencyEvents does not enable compression of tablet events
by itself, because paint applications typically need to process
all possible tablet events in order to draw the smoothest curves.
Change-Id: Ie7434ab4f9a4c64f2626c75e661cfd0d6cd22896
Task-number: QTBUG-44964
Reviewed-by: Gatis Paeglis <gatis.paeglis@qt.io>
Reviewed-by: Wayne Arnold <wayne.arnold@autodesk.com>
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
ICC 17 does not yet implement C++17 [[fallthrough]] nor does it support
GCC's __attribute__((fallthrough))
Change-Id: I84e363d735b443cb9beefffd14b8aba60a7e3f81
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
C++ does not specify which kind of floating point implementation is
being used. The C Standard doesn't either, but it includes a normative
reference for implementations adoping it (ISO/IEC 9899:2011 Annex F).
There are a few existing checks in qfloat16.cpp; move them to qglobal.cpp
(next to the other, similar checks), and improve them by actually
checking that the radix used for floating point numbers is 2.
Change-Id: I704a3a8efeb51014b3be23fb236654d647a6f44f
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Since commit bf2160e72c, we can rely on
charNN_t support in all compilers except MSVC 2013, and since that
commit, we use (in 5.10, not 5.9, yet)
!defined(Q_OS_WIN) || defined(Q_COMPILER_UNICODE_STRINGS)
when we only need charNN_t, the type, as opposed to its library
support (u16string, char_traits<char16_t>, ...).
This patch splits the Q_C_UNICODE_STRINGS macro into two, adding
Q_STDLIB_UNICODE_STRINGS for when we need std::uNNstring, leaving
Q_C_UNICODE_STRINGS for when we need just charNN_t support.
In QDebug, when constructing a QChar out of a char16_t, cast to ushort
first, since QChar(char16_t) was only officially introduced in Qt 5.10.
[ChangeLog][Potentially Source-Incompatible Changes] The internal
Q_COMPILER_UNICODE_STRINGS macro is now defined if the compiler
supports charNN_t, even if the standard library does not. To check for
availability of std::uNNstring, use the new Q_STDLIB_UNICODE_STRINGS
macro.
Change-Id: I8f210fd7f1799fe21faf54506475a759b1f76a59
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
That's before the return type or static, inline, constexpr or such
keywords (if any).
Perl Script:
s/^(\s+)(.*) Q_REQUIRED_RESULT(;)?(\s*\/\/.*)?$/\1Q_REQUIRED_RESULT \2\3\4/
Change-Id: I7814054a102a407d876ffffd14b6a16182f159e2
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
_mm_cvtps_ph is usually defined as a macro:
qfloat16.h:122:37: error: use of old-style cast [-Werror=old-style-cast]
Change-Id: Icd0e0d4b27cb4e5eb892fffd14b516ec47826c0c
Reviewed-by: Ville Voutilainen <ville.voutilainen@qt.io>
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
QSysInfo::productType() returned "osx" for all versions of macOS, even
10.12. Change 3e2bde3578 was incorrect.
[ChangeLog][Important Behavior Changes] QSysInfo::productType() and
QFileSelector behavior on macOS was restored to match what Qt used to
return in version 5.7.0 and earlier. The behavior found in Qt 5.6.2,
5.7.1 and 5.8.0 is removed.
[ChangeLog][Future Compatibility Notice] The identifiers that
QSysInfo::productType() and QFileSelector will use to identify macOS
systems will change in Qt 6.0 to match the Apple naming guidelines which
will be current then.
Task-number: QTBUG-59849
Change-Id: Ib0e40a7a3ebc44329f23fffd14b2b39392210c4f
Reviewed-by: Jake Petroules <jake.petroules@qt.io>
Turns out that different versions of ICC use different warning numbers.
The Linux and Windows compilers emit 1786, but the macOS one emits 1478.
Don't ask me why.
Change-Id: I523b0abacd5148b2bf08fffd14b475a4c4d89ba1
Reviewed-by: Jake Petroules <jake.petroules@qt.io>
Neither the Intel compiler nor Visual C++ have a dedicated switch to
enable F16C support, like GCC and Clang do. So we used the AVX switch
for that in commit 8241d51f70, as it was
the closest, lowest denominator. That was incorrect and insufficient.
The Intel compiler silently miscompiles the intrinsics with -xAVX,
making calls to out-of-line functions like _mm_cvtps_ph, which don't
exist. So we actually have to use AVX2 support to generate correct code.
That might be a problem later, since Ivy Bridge supports F16C but not
AVX2.
Visual C++ is able to generate F16C code with just -arch:AVX.
Either way, since there's no dedicated command-line switch, there's also
no dedicated preprocessor macro. We're using __AVX2__ for both
compilers, as that's a sufficient condition to indicate a processor that
supports F16C.
Change-Id: I27b55fdf514247549455fffd14b205b8d8b86da7
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
MSVC, Apple's Clang and Clang prior to 3.9 do not recognize _cvtss_sh
and _cvtsh_ss. So expand the operation to use directly the packed
intrinsics.
Change-Id: I27b55fdf514247549455fffd14b2046fd638593d
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
qsysinfo.h(235): error #1786: enum "QSysInfo::MacVersion" (declared at line 156) was declared deprecated ("Use QOperatingSystemVersion")
Take this opportunity to merge the two groups.
Change-Id: I27b55fdf514247549455fffd14b1c2a1d8eab869
Reviewed-by: Jake Petroules <jake.petroules@qt.io>
Aligns with EPERM (Operation not permitted) and ENOENT (No such file
or directory), and is what errno is set to on macOS when opening the
/dev/tty device when running inside e.g. Xcode, where isatty() will
return true.
Change-Id: I09b88eaa3ff611d95ab37f0ff4df9aaaca52747d
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
When we call realloc, the alignment of the new block may be different
from the old one. When that happens, we need to memmove the data to the
new position, before we start overwriting things.
Task-number: QTBUG-59804
Change-Id: I27b55fdf514247549455fffd14b07ea78918a3d0
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
GCC 7 defines __has_cpp_attribute even when invoked as "gcc" (possibly,
Clang does the same, according to a comment in the code, did not test
myself).
Hence, define the fallthrough declaration (as C++11 attributes)
only when compiling as C++, otherwise we pick them up even in C mode,
and they cause build failures.
Change-Id: I3f13205e014bb1dea59ee3664b29111521a7eae3
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
Clang implements the _cvtss_sh intrinsic by way of a macro, which
uses a C99 extension and that's not allowed in C++ mode:
float16.h:119:11: error: compound literals are a C99-specific feature [-Werror,-Wc99-extensions]
/usr/bin/../lib64/clang/3.9.1/include/f16cintrin.h:76:55: note: expanded from macro '_cvtss_sh'
Reported at https://bugs.llvm.org/show_bug.cgi?id=32491.
Change-Id: I27b55fdf514247549455fffd14b170df75dd4e1f
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
It replaces Posix' ssize_t in cross-platform code. Most importantly,
it allows QStringView users to refer to QStringView::size_type by
a shorter, less intimidating name.
Change-Id: I2128fefd2ffdd258b0b55cd90802a23c6bc033f6
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Also use canonical contact url.
Change-Id: I43f8c6a2c4949ee0e054045bccc17d82575b072c
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Jani Heikkinen <jani.heikkinen@qt.io>
Fix the condition in QWidgetPrivate::resolveLocale() to decide whether
to propagate locale: make it match setLocale_helper()'s condition when
deciding whether to propagate to descendants. This lead to a
QDateTimeEdit's calendar popup not getting told what locale to use
correctly, unless we setLocale() on it overtly, which then blocked
propagation of locale changes to it unless QDateTimeEdit manually
propagated the changes.
Fix the documentation of WA_WindowPropagation to mention locale as
also being propagated (which it was in several places, only neglecting
this one in resolveLocale).
[ChangeLog][QWidget][Qt::WA_WindowPropagation] Propagate locale
consistently, along with font and palette, within the widget
hierarchy. Previously, locale was propagated on ancestral
setLocale(), but not on creation of the descendant.
Task-number: QTBUG-59106
Change-Id: I92270f7789c8eda66a458274a658c84c7b0df754
Reviewed-by: David Faure <david.faure@kdab.com>
[ChangeLog][QtCore][Logging] If you set QT_FATAL_WARNINGS to a number
greater than 1, Qt will stop the application at that nth warning,
instead of on the first one. For compatibility reasons with previous
versions, if the variable is set to any non-empty and non-numeric value
different from 0, Qt will understand as "stop on first warning".
Change-Id: I0031aa609e714ae983c3fffd14676f1826f34600
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
We need Q_ASSERT in (C++11) constexpr functions, and the only way to
inject them in C++11 is to use the comma operator. E.g. in
QLatin1String:
constexpr QLatin1Char at(int i) const
{ return assert(1 >= 0), assert(i < size()), m_data[i]; }
The main problem with our existing Q_ASSERT is that while it is a
ternary expression in active mode, it was a statement in passive
mode. This is easily fixed by dropping the do-while loop and leaving
just its parenthesized exit condition. Add a cast to void, too,
ensuring that Q_ASSERT has type void in both passive and active modes.
But even in C++14 constexpr functions, which accept several
statements, Q_ASSERT needs to have a path through its conditionals
that is constexpr, but neither qt_assert(_x) nor qt_noop() are
constexpr. Nor can they be in C++11 (no void returns in C++11
constexpr functions). I fixed this by replacing qt_noop() with
static_cast<void>(0). The void cast is required so both 2nd and 3rd
arguments to the ternary are void (mixing void and non-void branches
in the ternary is only allowed if the void leg is a
throw-expression[1]).
As a drive-by, adjust to style guide, remove overparenthesization and
reverse the conditional in the ternary.
Apply it to QLatin1String where we had the problem that constexpr
functions had a narrow constract.
[1] should probably be extended to any [[noreturn]] void function,
e.g. std::terminate().
[ChangeLog][QtCore][QtGlobal] Q_ASSERT() and Q_ASSERT_X() now always
expand to expressions of type void that are usable in constexpr
contexts. This makes them usable in both C++11 and C++14 constexpr
functions.
Change-Id: I09c396bc0034ac344cfaadc6f8cbeb1b7b0cbabc
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Shortcuts are universally not shown on macOS or mobile platforms, making
applications look very obviously out of place.
Windows and GNOME desktop environments almost never use them.
Only KDE appears to do so commonly; default accordingly.
Task-number: QTBUG-49435
Change-Id: Ieac4cee57b15a02be5258f3d07749af6316af62b
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@qt.io>
For the windows file system engine, we add an extra macro to use
library loading if configured to do so, but avoid it on WinRT, as
none of the symbols would be found.
We also QT_REQUIRE_CONFIG(library) in the library headers and
exclude the sources from the build if library loading is disabled.
This, in turn, makes it necessary to clean up some header inclusions.
Change-Id: I2b152cb5b47a2658996b6f4702b038536a5704ec
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
The FP16 extension in IEEE mode is mandatory for Aarch64, so there is
no aarch64 configuration where the tables will be needed for conversion.
Change-Id: I9804e55c193cc9b5adcaedb720d8b980624139cc
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Reviewed-by: Glen Mabey <Glen.Mabey@swri.org>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Use of these macros implies defined(Q_OS_DARWIN) so don't complicate the
usage point (since the preprocessor cannot handle, for example
defined(Q_OS_DARWIN) && QT_DARWIN_PLATFORM_SDK_EQUAL_OR_ABOVE when the
latter macro is not defined).
Change-Id: I47995351f0e46d8a1d07708117f8eed63d87ba0f
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Replace all QT_NO_PROCESS with QT_CONFIG(process), define it in
qconfig-bootstrapped.h, add QT_REQUIRE_CONFIG(process) to the qprocess
headers, exclude the sources from compilation when switched off, guard
header inclusions in places where compilation without QProcess seems
supported, drop some unused includes, and fix some tests that were
apparently designed to work with QT_NO_PROCESS but failed to.
Change-Id: Ieceea2504dea6fdf43b81c7c6b65c547b01b9714
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
This adds the tabletTracking property in the same way that mouseTracking already
existed: there is a WA_TabletTracking attribute, and a TabletTrackingChange event
to notify when it changes. So for widget applications it's an opt-in feature.
QtQuick applications don't yet make use of tablet events, but when they do
in the future, we don't yet have a mechanism to turn the move events off;
it remains to be seen whether that will be necessary.
[ChangeLog][QtWidget] QWidget now has a tabletTracking property, analogous
to mouseTracking, which will enable TabletMove events while the stylus is
hovering, even if no button is pressed. This allows applications to show
feedback based on the other tablet event properties such as rotation and tilt.
Task-number: QTBUG-26116
Change-Id: Ie96e8acad882b167e967796cdd17f1ad747a2771
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
This is one of the pillars of my static container checking toolbox,
one of the main checks being that every type put into a Qt container
has been marked up with Q_DECLARE_TYPEINFO.
Obviously, we cannot upstream such a checker and inflict it upon the
world, but we can put some foundations in. This is the most central
one.
Change-Id: I9185facb2f37ba9fcc12c9aae5675eed454d755c
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
I don't know the historical reasons for that, but C++11 mandates
long double as a type. On Darwin this means that long double
was a complex type, so the best for now is to mark it relocatable.
Change-Id: Ic933947a282ad963d5d0168c2768cc98fdd456bc
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
We fully specialize QTypeInfo for most C++ built-in types,
but enums and extended integral types (like GCC's int128_t)
were not covered.
Now that we depend on <type_traits>, we can stop pessimizing
enums and extended integral types in QVector and QVLA by
defaulting QTypeInfo::isComplex to true for such types.
Fix a test that checked that enums were complex types. This should
have been a XFAIL test. Enums are not complex types.
Change-Id: Ibb0fb38cc83e980a428b5573d1db5666593418ae
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
also actually deletes qfeatures.txt, which was already claimed by
a668c6a6, but not actually done.
Task-number: QTBUG-58411
Change-Id: I686760632fee7c10b01bd2e83f2481b01bc2b774
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Use F16C or ARM FP16 if available at compile time.
Configure check added because older clang compilers have F16C defines
and flags but not all the intrinsics.
Change-Id: I71f358b8fd003e70ab8fcf35097414591e485112
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
GHS compiler is not fully compliant with iec559.
Therefore we need to replace is_iec559 assertion
with separate checks to build quint16.
Change-Id: I88c57e394b8d4e7899ee7d4a13cbfbac9436b2fc
Reviewed-by: Rolland Dudemaine <rolland@ghs.com>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
We have observed that MSVC 2013 gets lost with the overloads that this
header adds, causing compilation bugs. This is believed to be a compiler
bug, but it's not something we can work around.
Task-number: QTBUG-58555
Change-Id: I536c32a88bff44dab37afffd14a1bad1d31dc16d
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
Reviewed-by: Glen Mabey <Glen.Mabey@swri.org>
We're already using it, for example in qpaintengine_pic.cpp:502:
d->s << p << ti.text() << fnt << ti.renderFlags() << [...]
^
[ChangeLog][QtCore][QFlags] Added operator<< and operator>> for
streaming QFlags into and out of QDataStreams.
Change-Id: I33dc971f005a4848bb8ffffd1478e79d6102d1bc
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
If these lists weren't created in the first place, then they are empty.
We don't need to create it in order to conclude that. Unlike most
Q_GLOBAL_STATICS, these are almost never used and yet they were
always created due to where they were checked.
Since we're calling exists() before, there are two consequences: first,
since the list already exists, we're not allocating memory so it cannot
throw std::bad_alloc when being accessed. Second, since we've just
checked it exists, we can use QGlobalStatic's operator*(), which is
slightly faster than operator()(). The weird &(*list) syntax is only to
avoid changing the rest of the code that used a pointer
Change-Id: Ifaee7464122d402991b6fffd14a0e44f533dc3d9
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
MSVC < 1800 (older than 2013) is no longer supported, so we can simplify
the code. And the implementation for C never worked -- "if it's not C++,
let's use a namespace!"
Change-Id: Iaeecaffe26af4535b416fffd148c2c6788c43881
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Erik Verbruggen <erik.verbruggen@qt.io>
This constitutes a fairly complete submission of an entirely new
floating point type which conforms to IEEE 754 as a 16-bit storage
class. Conversion between qfloat16 and float is currently performed
through a sequence of lookup tables. Global-level functions
qRound(), qRound64(), qFuzzyCompare(), qFuzzyIsNull(), and
qIsNull() each with a qfloat16 parameter have been included
for completeness.
[ChangeLog][QtCore] Added new qfloat16 class.
Change-Id: Ia52eb27846965c14f8140c00faf5ba33c9443976
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Although it's permitted to specify the project name together with
a version number for \since, it's unnecessary for Qt classes and
functions.
This change also normalizes the version formatting: '<major>.<minor>'
Change-Id: Ie5a43662077d13c31e241bcde8a7a2849d27d330
Reviewed-by: Leena Miettinen <riitta-leena.miettinen@qt.io>
It's included by qglobal.h, so we get it for free in other headers.
Change-Id: I90072156e313271a5354a39cbf78a83a6885c431
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Instead of using the actual compiler version to determine which C++11
features to enable, let's just use the Mirosoft compiler version that
the actual compiler is claiming compatibility with. That is because the
limiting factor is often the standard library, not the compiler itself.
This will cause some features that do not depend on the library to also
be disabled, but oh well. Better upgrade your Visual Studio version
instead.
Task-number: QTBUG-57696
Change-Id: I3e4e5051937c40319d6efffd14912cd4fdab25fb
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Sérgio Martins <sergio.martins@kdab.com>
This patch replaces QRegExp by QString search and replace.
Change-Id: I11165afa45f8f9a856e6fb9b64929e4bdacb913d
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Gcc defines neither _MIPS_ARCH_MIPS32 nor __mips32 on MIPS32
architectures, instead __mips is defined to 32.
This fix exposed bit-rot in qdrawhelper where qt_memfill32 was set as
a function pointer despite not being one since Qt4.
Change-Id: I87461823e54fa3166223ebf97175fd05d2f2fd16
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
There should be no blank lines in \value content, otherwise qdoc
ends the list and starts a new one.
Change-Id: Idddc7992317894487445aea36397136df40b9691
Reviewed-by: Martin Smith <martin.smith@qt.io>
QHostAddress is one of the few classes in Qt which is pimpl'd but not
implictly shared, making it suprisingly expensive to copy around, return
by value and so on. Being pimpl'd it is also still lacking a move
constructor, like most of such types in Qt.
Remove a bit of the surprise factor and make it implictly shared. In
practice this means making it eagerly parse host addresses from strings.
Since it was entirely implemented out of line, replacing it with a
implictly shared implementation is binary compatible.
[ChangeLog][QtNetwork][QHostAddress] QHostAddress is now implicitly
shared.
Change-Id: Ia7ff94efcb74e7321b7607cd690c5c162f685605
Reviewed-by: Richard J. Moore <rich@kde.org>
[ChangeLog][QtBase][General] Removed support for WinRT/Windows Phone 8.1.
Task-number: QTBUG-57288
Change-Id: Ifd6d6780cbbdb710d99556ba3d2fb2e514d4f789
Reviewed-by: Oliver Wolff <oliver.wolff@qt.io>
operator== and operator!= have been disabled, as they are likely to be
misused and are not particularly useful in practice. The same goes for
the QVersionNumber conversion convenience functions.
The constructor normalizes version component values so that invalid
versions like [5, -1, 3] cannot be constructed and made to wreak havoc
on assumed logic.
Change-Id: Iabb6876bd5dc11522032837f78cf825b921a49b2
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
clang/g++ still warn when encountering the implementation of a
deprecated function.
Follows up 21a247adb4
Change-Id: I6ab1695acb520ef7ce7cb1896545d02607c3ce29
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@qt.io>
So that QFlags can use an (un)signed int matching the
underlying type as identified by the compiler and not by us.
Requires fixing a few warnings about sign conversion due to
QFlags misusages in qtbase that were either plain wrong, or
were relying on the enum being backed by an (un)signed int
when it wasn't.
Keep qtypetraits.h in the source tree in order to prevent
source breaks if some downstream #includes it (note however
that it did not contain any public API).
Change-Id: Ib3a92b98db7031e793a088fb2a3b306eff4d7a3c
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
It's supposed to be a full alignment (like Top | Right), not
just one flag.
Change-Id: I656adda83742d7e4f31955322e937e979b32747c
Reviewed-by: David Faure <david.faure@kdab.com>
Initial support for INTEGRITY to build QtBase
Change-Id: I18f36b4dea9107f01e1c281e4b62880590c777a1
Reviewed-by: Tuukka Turunen <tuukka.turunen@theqtcompany.com>
Reviewed-by: Nikola Velinov <nvelinov@ghs.com>
Reviewed-by: Risto Avila <risto.avila@qt.io>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Honor the rule of five. Copy assignment was already disabled.
Disable also copy construction, but re-enable the move special
member functions.
This requires making the container non-const; to avoid detaches
and keep compatibility with containers that only have begin/end
(but not cbegin/cend), use qAsConst. This requires moving qAsConst
definition a bit up in the file.
Change-Id: I19cd74437cf69baceada9483920a21e96d7b1727
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Don't use Dinkum choices when the C++ library is libC++ (QNX 7.0).
Change-Id: I18c3f716ccfb0c02dbfdc01eac4b707d3ae9aab6
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
turns out that just appending builtin-qt.conf isn't a good idea:
executable-editing tools (objcopy, prelink, etc.) will happily drop the
"attachment".
a safe method would be adding a proper section to the executable, but
there doesn't appear to be an objcopy equivalent in msvc, and using
entirely different methods of embedding the file with different
toolchains seems like a rather bad idea.
so instead go back to the old method of building qmake with a generated
qconfig.cpp. of course, as said file is now created by qmake itself, we
have to compile qlibraryinfo.cpp a second time, and link a second qmake
executable.
Task-number: QTBUG-57803
Change-Id: I9e232693550aa870cec154e49cc06add13017cc2
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
We have to enable qt_safe_ftok with either sharedmemory or
systemsemaphore. In order to make the resulting QT_CONFIG work with the
bootstrap library we switch the features off for bootstrapping. Some
tests and examples have to be excluded when sharedmemory is not
available.
Change-Id: I3fc3926d160202b378be2293fba40201a4bf50c5
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This allows the Dptr parameter of the Q_DECLARE_PRIVATE_D macro to be
a smart pointer, not just a raw pointer.
Change-Id: Iaf27352e327e9aedea149461d47f2f11460a42dc
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
ISO 14443-3 is for nfc communication and uses 2 different checksums.
The existing one is from ISO 3309 and the other one is from ITU-V.41.
Both are needed to implement an own transport layer defined in ISO
14443-4 to allow nfc commands with a length above 250 byte independent
from the smartphone.
This change will avoid code duplication in QNearFieldTarget.
The private function qNfcChecksum is a copy of qChecksum.
Change-Id: I790ffec8e2ea46f88b2db6f48b64fdcb140e7b70
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
initially, the idea was that QLibraryInfo would receive a
pre-sysrootified ExtPrefix from the builtin qt.conf. matching this
against the sysroot would then tell us whether to sysrootify the Prefix
from a "regular" qt.conf as well.
however, this would have lead to some major ugliness and inconsistency
between the code paths, so i changed my mind.
unfortunately, i failed to adjust the remaining code, leading to
169a40d51 entirely breaking sysrootification ...
the proper (and nicely consistent) solution is to introduce a
SysrootifyPrefix key to qt.conf. this is user-accessible as well, so as
a bonus it is now possible to adjust the setting at qmake installation
time. incidentally, this omission was the last thing that prevented
using the same qmake host build for any imaginable configuration of the
same qt version ... i think.
Change-Id: Ic0eebf21f93651f6374628c0ad8b206d696a4a7e
Reviewed-by: Andy Nichols <andy.nichols@qt.io>
Clang pre-3.4 didn't like this and it's used in Xcode 5.1 (which we need
to support for 5.8).
error: 'this' cannot be implicitly captured in this context
typename T::const_iterator i = c.begin(), e = c.end();
^
Task-number: QTBUG-57488
Change-Id: I63e21df51c7448bc8b5ffffd148e688d7c9b89d6
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
g++ still warns when encountering the implementation of a deprecated
function.
Change-Id: I6a25fc8c814590e5337069f9bced0cdec97653bf
Reviewed-by: Jake Petroules <jake.petroules@qt.io>
this moves us another step towards the "outer" configure doing just
minimal bootstrapping of qmake.
a challenge here was that so far, qmake itself needed qconfig.cpp. this
was replaced by usage of a qt.conf file instead of compiled-in values.
however, to make the executable still self-contained, that qt.conf is
embedded into it (by simple appending of a fixed signature and the text
file).
the qmake with the embedded qt.conf is not used for the qt build itself,
which instead relies on the qt.conf in bin/ as before. however, due to
the missing built-in values, this file now needs to contain more
information than before. but except for a minimal version that is needed
to start up qmake/configure at all, that file is now also generated with
qmake. as some of the newly set up properties are subsequently used by
configure itself, qmake gains a (deliberately undocumented) function to
reload the qt.conf after it's fully populated.
unlike the old implementations, this one doesn't emit redundant qt.conf
entries which match the hard-coded fallbacks. omitting them leads to
leaner files which are more comprehensible.
Started-by: Paolo Angelelli <paolo.angelelli@qt.io>
Change-Id: I4526ef64b3c89d9851e10f83965fe479ed7f39f6
Reviewed-by: Jake Petroules <jake.petroules@qt.io>
Commit c35fef9d3b wasn't sufficient. The
problem is that there's a complex combination of libc headers (math.h),
C++ headers (cmath), which may be provided by three different sources on
Linux (glibc, gcc and ICC). On some combinations, the isnan macro leaks
from math.h or cmath and that's what the the commit above tried to fix.
On some other combinations, there's no macro but there's an ::isnan
function defined. When we do "using namespace std; return isnan(x);",
that causes a compilation error. This commit solves that by detecting
whether there is a macro defined.
error: more than one instance of overloaded function "isnan" matches the argument list
function "isnan(double)"
function "std::isnan(double)"
argument types are: (double)
Change-Id: Iaeecaffe26af4535b416fffd148bf71826541bdd
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Namely: decltype(). Ideally we'd want C++17's template constructor
argument deduction, but instead use the C++11 solution: a factory
function. This enables using things such as lambdas in the container
argument.
Change-Id: Idba64d8069d15bbafe54cfdebe24b1fba1eb8d0a
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Provide appropriate alternative documentation where applicable.
Change-Id: I73d810938bb961a74d06d8cedb05c38675363ef0
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This adds documentation for QOperatingSystemVersion::OSType and cleans
up some extraneous quotes.
Change-Id: Idaeb163caded9a51ce0fbcc812eb622b4227844e
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
That's what happens when you don't test and just rely on an the warning
listing. ICC has two warning numbers for deprecated warnings: one that
matches Q_DECL_DEPRECATED and one for Q_DECL_DEPRECATED_X.
Change-Id: I73fa1e59a4844c43a109fffd148ca7a05eda8f13
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Turn them into proper private features, and remove setting of
defines in the pri file.
Change-Id: Iafc11e93d4a9349bf15971dc1adac9a828ea03f6
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
This is the only warning we disable in a lot of places in Qt 5.8 source
code. If other warnings become common, we can add macros for them too.
Change-Id: Iaeecaffe26af4535b416fffd1489d1968e29c52a
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
The user can call qSetMessagePattern after program start, so we need to
be sure that the parsed argument data is properly cleared.
Task-number: QTBUG-57144
Change-Id: I1978c6b95bd84639a8c4fffd1487429b04725522
Reviewed-by: Kai Koehne <kai.koehne@qt.io>
Introduce a glProgramBinary-based disk cache in QOpenGLShaderProgram.
By switching the typical
program->addShaderFromSourceCode(QOpenGLShader::Vertex, ...)
program->addShaderFromSourceCode(QOpenGLShader::Fragment, ...)
invocations to
program->addCacheableShaderFromSourceCode(QOpenGLShader::Vertex, ...)
program->addCacheableShaderFromSourceCode(QOpenGLShader::Fragment, ...)
the compilation may be skipped via gl(Get)ProgramBinary and a disk
cache, when supported. Such QOpenGLShaderProgram instances will have
no QOpenGLShader instances attached. Instead, the entire program
binary (which is driver-specific) is loaded as-is.
Support means OpenGL ES 3.0 or the presence of
GL_ARB_get_program_binary, in combination with >= 1 supported binary
formats. Note that some drivers claim program binary support but
expose no formats. This amounts to no support in practice.
When support is not present, calling the new functions is equivalent
to the non-cacheable variants. If the OpenGL driver changes (vendor,
renderer, version strings), recompilation and storage of the new,
potentially incompatible binary program will happen transparently.
The cache can always be disabled by setting
QT_DISABLE_SHADER_DISK_CACHE=1 or the new application attribute
Qt::AA_DisableShaderDiskCache.
Location-wise the primary choice is the shared cache
(GenericCacheLocation). If this is not available or is not writable,
the per-process one (CacheLocation) is used instead.
In addition to the new public APIs in QOpenGLShaderProgram, the main
shader users in QtGui are migrated as well. (OpenGL paint engine,
glyph cache, blitter, eglfs mouse cursor). This means that any
application using QPainter on OpenGL or widgets with eglfs will
benefit from the improved startup times. Qt Quick will follow suit as
well.
[ChangeLog][QtGui][OpenGL] QOpenGLShaderProgram offers a built-in
program binary disk cache for systems with OpenGL ES 3.x or
GL_ARB_get_program_binary. This can lead to significant increases in
performance when it comes to application startup times for
example. Usage is opt-in for direct C++ users of the class, however
Qt's own main users of shaders, including Qt Quick and QPainter's
OpenGL engine, are migrated to use the new, cache-enabled APIs.
Opting out on application level is always possible via
Qt::AA_DisableShaderDiskCache.
Task-number: QTBUG-55496
Change-Id: I556f053d258bfa6887b1d5238c9f6396914c5421
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Andy Nichols <andy.nichols@qt.io>
They should be enabled/disabled through the configuration system.
Remove some unused defines, and move one define from qglobal.h to
a proper feature definition in Qt Gui.
Change-Id: Ie8d5bff9712ba745af60b42ceca3f0440bed2706
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Turn iconv off if ICU is being used (in line with codecs.pri)
and get rid of the DEFINES += GNU_LIBICONV in the pri file.
Change-Id: I6fbca975498adbb3e67f913ae9b1dd5cc53ee8da
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
swap the branches for (not) reading from qt.conf, and use a state
variable instead of an 'else' for mutual exclusion. this is somewhat
more self-documenting, and allows for a saner handling of the mkspec
fallbacks (which really should have been in a separate [QMake] section
along with Host* and Sysroot, but changing that now is way too much
hassle downstream).
Change-Id: I80a73294022fd1e8d84fe501b737c4fc7758662f
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Any other use than for enums should use std::is_[un]signed. Make this
explicit by renaming the type traits.
Change-Id: I494158563c95c710e710d0d337f4e547006df171
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Since the macro is now just a wrapper for std::is_enum,
its use is also deprecated.
[ChangeLog][QtCore][Global] Q_IS_ENUM is deprecated.
Use std::is_enum<>::value instead.
Change-Id: I09b9f4559c02c81f338cace927873318f2acafde
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Remove most type traits from qtypetraits.h, but keep the custom
implementation of is_signed/is_unsigned. This gets rid of
BSD-3 licensed code from Google in a public header (hugh!).
The custom implementations for is_signed/is_unsigned are kept
because the implementations in gcc's standard headers do not
work as we expect for enums - both is_signed and is_unsigned
always returns false there - see also
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=59027
[ChangeLog][QtCore][General] Qt now relies on type traits from
the C++ standard library.
Change-Id: I3f2188b46949f04ca4482a6ac9afd3482103f0e1
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Define NTDDI_VERSION, just like _WIN32_WINNT, to be Windows Vista.
Usually NTDDI_VERSION is automatically set by MinGW headers to the value
that matches _WIN32_WINNT. However, for precompiled headers the inclusion
order is that _WIN32_WINNT is set _after_ the relevant MinGW header is
parsed, so this can fail.
The alternative would be to set _WIN32_WINNT via a compiler flag, e.g.
in the mkspecs.
Change-Id: Id59e7083f0d3e00491b54e87647c6c9fabb99795
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
By some unfortunate oversight, this enum was never registered.
Change-Id: I2227ccf294d2cf717187a3dcaaf4cbfacc4ac65d
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
I'm guess I introduced the regression in the commit
18ed6f20ad, which I wasn't sure about.
Change-Id: Ic46ff326a6ba46bc877cfffd14839f84fdf796e7
Reviewed-by: Jake Petroules <jake.petroules@qt.io>
