This reverts commit d0eba497c1.
A better fix is to use -fPIC instead of just -fPIE fo
-reduce-relocations.
Task-number: QTBUG-45755
Change-Id: I1759291b684fd76d4009e4be9ba1354eb056e659
Reviewed-by: Simon Hausmann <simon.hausmann@theqtcompany.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Suppress QMAKE_LFLAGS_BSYMBOLIC_FUNC (-Bsymbolic-functions for g++)
by default since it causes crashes with gcc 5.X. Since applications
compiled with gcc 5.X might run against Qt compiled with gcc 4.9.X
and CLANG might also be affected, turn it off by default.
Task-number: QTBUG-45755
Change-Id: I5704c3156db6b6f74e1c14576e5d02bcbd3082a4
Reviewed-by: Simon Hausmann <simon.hausmann@theqtcompany.com>
Reviewed-by: BogDan Vatra <bogdan@kde.org>
Allow setting of pg_config path for cross compilation where pg_config is
not in the command search path (do the same as for mysql_config).
This is e.g. used for buildroot (see [1] for details).
[1] http://lists.busybox.net/pipermail/buildroot/2015-February/119714.html
Change-Id: I11d084496ffbb6f8bc350dbcf2971a5be8e3b346
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
While the -fuse-ld=gold flag is related to linking, it is an argument to the
compiler driver to tell it what linker to execute, NOT an option to tell the
linker to behave differently.
So it shouldn't get prefixed with -Wl when passed though the compiler driver.
Change-Id: I2b50cb6d2bd8911aa9b305cd8e755d4dfe923041
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
Reviewed-by: Joerg Bornemann <joerg.bornemann@theqtcompany.com>
Option -separate-debug-info is not enabled by default, even help
claims it to be enabled.
Change-Id: I4a986acbb2b7f2c7bf7feaaff29bcc6bfadcc559
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
Commit d44781730c disabled use of pkg-
config on Mac, which in turn automatically disables all of the tests
that depend on pkg-config too, like D-Bus. If you then use the -dbus
flag, configure would error out telling you that it couldn't find D-Bus,
despite it being there.
Now we print some helpful information on error and extra information for
finding out why it may have got disabled.
Change-Id: Ic4e41c58d62d80eb1a0f3ca1c93529e049aaf231
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
It was not possible to determine whether Qt is being built as
a shared or static library in the config tests, as that information
was not added to QMAKE_CONFIG used in the compileTest function.
For linking to ICU on Windows, there is logic which depends on
it in config.tests/unix/icu/icu.pro, but was broken for the static
target, as the expected library name was not found (icuXX -> sicuXX).
Change-Id: I22c2d543ea14feebbad36961c9a29e99f66323ff
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
Commit a2da88712f was incomplete. Though
it did enable building of QtDBus without the libdbus-1 headers, but it
still kept looking for it. We don't need to do it anymore if CFG_DBUS is
the default (runtime), so skip the actual check.
This commit does not change behavior. All it does is skip the D-Bus
tests if we're about to use dlopen'ed libdbus-1.
Change-Id: I3947443c9ef6cc409cb6b154564f8af6fb5de84e
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
When CFG_DBUS != "no", we enable the build of linuxaccessibility in
src/platformsupport, which requires QT += dbus to work. When no dbus
headers are found, CFG_DBUS is set to runtime, which means we'll
enable a11y but "dbus" will not end up in qconfig.pri and therefore
src/dbus is not built at all (and not before platformsupport). Therefore
the build in platformsupport breaks.
So even if CFG_DBUS is set to "runtime" we need to make sure that dbus appears
in qconfig.pri.
Task-number: QTBUG-43205
Change-Id: Ibbe5e66552226cdcc1ce16fb83556406f3bb0e09
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The compile.test script expects -I... parameters. Passing just the
include path is therefore wrong. Pass cflags instead.
Change-Id: I0f9b155d1e710fe7d77c64a867a89290d4f31db3
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
Most of QtDBus already needs very little from libdus-1, so create an
extra header containing the minimum API we actually need.
One large advantage of this solution is that now QtDBus can always be
enabled, even if the system doesn't have libdbus-1 installed. This is
interesting on OS X, where libdbus-1 is often installed by Homebrew or
MacPorts, which may include extra libraries we don't want in our
packaging.
Change-Id: I1b397121ec12eeca333ef778cf8e1c7b64d6b223
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Conflicts:
dist/changes-5.4.0
7231e1fbe2 went into 5.4 instead of the
5.4.0 branch, thus the conflict.
Change-Id: I70b8597ab52506490dcaf700427183950d42cbd1
There was an override for Android which would disable the configure
error when doing an OpenSSL build without having the headers
available. This has several times lead to packaging errors where
OpenSSL gets disabled but it's not noticed before the package testing,
which delays the process.
I'm not 100% sure of the reasoning behind the override, but I think
it's a left-over from Necessitas where OpenSSL was statically linked
into Qt.
Change-Id: I2bdc33fb60c59cd493987959d4bbbbb4e9735a92
Task-number: QTBUG-42851
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@theqtcompany.com>
Reviewed-by: BogDan Vatra <bogdan@kde.org>
Reviewed-by: Kai Koehne <kai.koehne@theqtcompany.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
The change was made too late in the 5.4.0 release
cycle, and broke the Qt build and deployment in
several areas:
- macdeployqt
- OS X 10.7 builds
- shadow builds
This reverts commit c0a54efc40.
Change-Id: I1c1ad4901228f5516352ccdfa963e8ea2b5013b6
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@theqtcompany.com>
configure script has been silently accepting whatever flags that begin
with "-feature-" even if the feature name does not exist at all.
Since the script validates many other flags, this behavior can make
users believe flags they supply is valid when it isn't.
Besides, this option is currently not protected against typo in any way.
This commit verifies those flags against content of
"qtbase/src/corelib/global/qfeatures.txt" and fails if supplied
flag is not a valid feature name.
Change-Id: Ib19ec66dd5558fb5491e8e080ce07d4807d94c1f
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
Apple will from February 1, 2015, require all applications uploaded to
the App Store to be built for both 32-bit (armv7/s) and 64-bit (arm64).
https://developer.apple.com/news/?id=10202014a
We enable fat Qt binaries by passing both -arch armv7 and -arch arm64
to clang, which takes care of lipoing together the two slices for each
object file. This unfortunately means twice the build time and twice
the binary size for our libraries.
Since precompiled headers are architecture specific, and the -Xarch
option can't be used with -include-pch, we need to disable precompiled
headers globally. This can be improved in the future by switching to
pretokenized headers (http://clang.llvm.org/docs/PTHInternals.html).
Since we're enabling 64-bit ARM builds, we're also switching the
simulator builds from i386 to fat i386 and x86_64 builds, so that
we are able to test 64-bit builds using the simulator, but we're
keeping i386 as the architecture Qt is aware of when it's building
for simulator, as we need the CPU features to match the lowest
common denominator.
Change-Id: I277e60bddae549d24ca3c6301d842405180aded6
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Defaulting to absolute_library_soname on configure -rpath is no longer
necessary as now we support @rpath install name ids on OS X and iOS.
This also sets QMAKE_SONAME_PREFIX to @rpath for Qt modules when built with
rpath configuration.
This makes Qt libraries relocatable on OS X. Qt SDK is not yet relocatable
though, because plugin location (including cocoa plugin) is still resolved
using absolute path (see QTBUG-14150), also there are several absolute paths
hardcoded in qmake mkspecs pri files.
Task-number: QTBUG-31814
Change-Id: Ie9dffefcd2a946c1580293d433621c1adb7e06c4
Reviewed-by: Jake Petroules <jake.petroules@petroules.com>
We set the variables to empty when we start processing, so ${FOO-foo}
will never print "foo". We need to use ${FOO:-foo}.
Change-Id: I00c28edb10d8eaa09df689905a302b576b246806
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
Fixup after d44781730c, which would override whatever the user passed
on the command line.
Change-Id: If4d260801866ff53de3e6dfd6d37016fd8453d8d
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
The code which extracts style assets for the Android style
is licensed under the Apache license, which is not compatible
with LGPLv2.1. It is, however, compatible with LGPLv3. This
means that the Android platform plugin cannot be LGPLv2.1
as long as this code is included.
To minimize licensing confusion, we default to only providing
LGPLv3 for Android. If you want to build a LGPLv2.1-compatible
library, you can add -no-android-style-assets to the
configuration. This will in turn enable the LGPLv2.1 in
the configure output, and it will disable the extraction
code in the platform plugin.
Running the Android style with an LGPLv2.1-compatible platform
plugin will work, but it will look horrible.
[ChangeLog][Android] Default open-source license for
Qt for Android is now LGPLv3. For compatibility with the LGPLv2.1
license, add "-no-android-style-assets" to your configuration.
Change-Id: I6c7b52140f38138520871fa7c69debbb4ee90e6c
Task-number: QTBUG-41365
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Reviewed-by: Christian Stromme <christian.stromme@digia.com>
This license file has to exist, since much of the code is licensed under
LGPLv3.
Change-Id: I2795a7cc62f6de65a35921e38d2ab5f8f0233f71
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
We don't have an explicit XPLATFORM_OSX, but XPLATFORM_MAC applies
to both OS X and iOS (as now clarified), so we move the default
out of the XPLATFORM_IOS scope and to a XPLATFORM_MAC scope.
Change-Id: I6b9ba9c881c28def08b9ab863d0165fbd9dedc6d
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
qmake variables using $$system() were incorrectly parsed by the custom
qmake parser in the configure script, when using GNU awk 3.1.8 or
earlier. They are parsed correctly with GNU awk 4 or mawk.
This was occurring with such an assignement (from an extra mkspecs file):
QMAKE_CC = $$system($$CMD QMAKE_CC 2>/dev/null)
The custom qmake parser in the configure script first attempts to
expand $$UPPERCASE variables, before running $$system(), using this:
match(value, /\$\$(\{[_A-Z0-9.]+\}|[_A-Z0-9.]+)/)
But when using non-ASCII locales with GNU awk 3.1.8 or earlier,
$$system was expanded (to an empty string) because these earlier awk
versions match lowercase letters for the [A-Z] regexp, which is
traditionally used to match uppercase letters.
This behavior has been changed in GNU awk 4.0.0, which only matches
uppercase letters for [A-Z] by default. A workaround for earlier GNU
awk versions is to run awk with the C locale.
See GNU awk NEWS "Changes from 3.1.8 to 4.0.0":
25. Gawk now treats ranges of the form [d-h] as if they were in the C
locale, no matter what kind of regexp is being used, and even if
--posix. The latest POSIX standard allows this, and the documentation
has been updated. Maybe this will stop all the questions about
[a-z] matching uppercase letters.
THIS CHANGES BEHAVIOR!!!!
See also gawk.info "A.7 Regexp Ranges and Locales: A Long Sad Story"
Change-Id: Ibb3eb28738c3e77d496c634e1f5c9f630957e730
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Fixes missing space before `]` in test condition introduced in 8d57725338.
Change-Id: I741c291677f32056a0a0bec12cb4d9fd293a2021
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
[ChangeLog][Important Behavior Changes] HarfBuzz-NG is now the default
shaper on all platforms. This results in a better shaping results
for various languages, better performance, and lower memory consumption.
Task-number: QTBUG-18980
Change-Id: I4d9454fc37e9050873df3857e52369dfc7f191b2
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@digia.com>
If the user requested the system xkbcommon with -system-xkbcommon, then
don't silently fall back to the bundled version if the functionality
test fails. Instead, print the an error notice.
Also note that the -xkbcommon argument didn't do anything, since it set
the variable to "yes".
Change-Id: I2c9e820bd076995aaaad987ecce76ebddcd79b4a
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
We no longer support the maemo/meego platform, so we can remove the
specific code for that platform.
Change-Id: Ia7f0730eba2d96794b97b7ca4753f63a2d7bc2a8
Reviewed-by: Laszlo Agocs <laszlo.agocs@digia.com>
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
This code hasn't been tested for at least 4 years. It's not maintained
and probably doesn't work.
Change-Id: I4b9a5179e34111b400914f91caa6b741b69771bb
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Building QtDBus on Linux host for QNX target had two issues:
* Configure check failed, because dbus-1 library was not linked in,
if target platform doesn't support pkg-config.
* Host tools were not built, because pkg-config was not used to locate
dbus headers on the host.
Task-number: QTBUG-37324
Change-Id: I71d8309599fd40ef2dd8c9e3b44b93a7482019f1
Reviewed-by: Rolf Eike Beer <eb@emlix.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
We don't actually detect whether the compiler can create Neon code or
provides Neon intrinsics. Most of them do, so that test would be mostly
moot. We removed the detection previously because we couldn't
automatically enable Neon due to leakage of instructions outside the
areas protected at runtime.
Instead, we rely on the mkspec properly passing the necessary flags that
enable Neon support.
This commit does not change that. All it does is verify whether the arch
detection found "neon" as part of the target CPU features. In other
words, it moves the test that was in simd.prf to configure.
It does fix the Neon detection in configure.exe, which was always
failing for trying to run a test that didn't exist
(config.tests/unix/neon).
Change-Id: Id561dfb2db7d3dca7b8c29afef63181693bdc0aa
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
This patch adds the feature use_gold_linker to use the gold linker that
has been part of of GNU binutils since 2008. Gold links C++ libraries
much faster and use less memory.
The feature is autodetected when building Qt on Linux, but can be disabled
in configure. On MingW builds it is default off but can be enabled for
cross builds.
Change-Id: Icdd6ba2e706b2c791bcf44b6e718c2b7a5eb2218
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Contrary to expectations, various <foo>-config tools sometimes spit out
denormalized paths, which breaks the text-based filtering, as it relies
on exact matches with normalized paths.
Change-Id: I0613ed24953a3bde19939d28d09572c88b43a361
Task-number: QTBUG-39216
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
the windows version is already sane in this regard
Change-Id: Ib9c6fa35ede0e5e98b38b7b97086b9245c79d48f
Reviewed-by: Joerg Bornemann <joerg.bornemann@digia.com>
[ChangeLog][configure] The -process/-fully-process/-dont-process options
have been removed due to being unnecessary and counterproductive.
-fully-process has always been broken to a degree under unix (and since
5.0 under windows) - rcc isn't built before running qmake -r, so the
dependencies are unreliable (and there are many warning messages from
qmake).
also, it is a lot slower nowadays, as qmake -r is not parallelized.
-dont-process doesn't make any sense any more - even if you don't need
the Makefile for some obscure reason, the time spent on creating it is
not relevant without the recursion.
this leaves -process as the only option.
Task-number: QTBUG-36955
Change-Id: Ifd3949d9ff773780646c6f65db1629e1c19e53d2
Reviewed-by: Joerg Bornemann <joerg.bornemann@digia.com>
If -qreal float is passed, fullCpuArchitecture() will now include
"-qreal_float". If something else other than "float" is passed to
-qreal, we'll try to encode it (e.g., -qreal "fixed<int, 7>").
Change-Id: Ie33fd1a643f4376e6f01a7966e01c7c34e6fcffd
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
GCC 4.9 now allows us to #include any and all intrinsics headers, not
just the one for which we're compiling code, a behavior that ICC and
MSVC have had for some time. With that, we're able to have the functions
for different targets in the same source file. See the GCC manual:
http://gcc.gnu.org/onlinedocs/gcc/Function-Multiversioning.html
This functionality is notified by the QT_COMPILER_SUPPORTS_HERE(XXX)
macro, which indicates that all the intrinsics from
QT_COMPILER_SUPPORTS_xxx are available and enabled. To complement, a
QT_COMPILER_SUPPORTS(XXX) macro is also added.
Unlike ICC and MSVC, GCC requires a special function attribute, which
will also cause code optimization. That's the QT_FUNCTION_TARGET macro.
Note: because of the absence of the target attribute, ICC and MSVC will
not generate instructions with the VEX prefix unless they only exist
with the VEX prefix or if -mavx / -arch:AVX are enabled.
Change-Id: I0c1880c20324bd8e0fc68a863e36d1fa7755dff0
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@digia.com>
This matches the -ffunction-sections from bootstrap.pro, which tells the
compiler to create a section for each function. The -gc-sections option
tells the linker to drop what wasn't used (normally, it only drops
entire files).
Before (on Linux, built with -O3, no LTO):
text data bss dec hex filename
1746385 7920 3750 1758055 1ad367 bin/moc
1444101 6664 1894 1452659 162a73 bin/rcc
4407725 1568 4896 4414189 435aed bin/qmake
After:
text data bss dec hex filename
1131655 6520 3494 1141669 116ba5 bin/moc
1027043 5480 1766 1034289 fc831 bin/rcc
3578489 1656 5313 3585458 36b5b2 bin/qmake
Gain: 35% on moc, 28% on rcc, 19% on qmake
Before (on OS X):
__TEXT __DATA __OBJC others dec hex
1495040 12288 0 4294993008 4296500336 100176470 bin/moc
1265664 8192 0 4294983904 4296257760 10013b0e0 bin/rcc
5279744 81920 0 4297912320 4303273984 1007ec000 bin/qmake
After:
__TEXT __DATA __OBJC others dec hex
806912 8192 0 4294988132 4295803236 1000cc164 bin/moc
720896 8192 0 4294979764 4295708852 1000b50b4 bin/rcc
4841472 77824 0 4295580688 4300499984 100546c10 bin/qmake
Gain: 46% on moc, 43% on rcc, 8% on qmake.
Change-Id: Icc7cdc9fd6f5db15537b4adabaac7e7a27e539d4
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
This might lead to a smaller binary if we use --gc-sections too.
Change-Id: I7e17b956a85ecefc3e187054848393d2855152b6
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
