| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The pkg statements were generated in invalid order if user used
pkg_prerules to redefine languages supported by the pkg file.
Also made it possible to override dependency statements autogenerated
for Qt and QtWebkit, as these statements need to be user defined in
case of multiple languages.
Defining the following in .pro removes all dependencies from pkg rules:
default_deployment.pkg_prerules -= \
pkg_depends_webkit \
pkg_depends_qt \
pkg_platform_dependencies
Task-number: QTBUG-9279
Reviewed-by: Janne Anttila
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Configure parameter -qtlibinfix will now change also plugin names
of plugins built and installed with Qt in Symbian.
Default plugin path is changed to: /resource/qt<libinfix>/plugins.
Other plugins besides the ones installed with Qt are not renamed.
With infixed configuration, Qt can be installed on a phone that already
has a Qt installed on ROM without interfering with the ROM version of
Qt.
Note that since s60main.rsc resource cannot deployed with infixed Qt,
and infixing it is somewhat problematic, currently the phone needs
to have proper Qt installation, too, for infixed Qt to work.
It also means that any changes to Qt that would affect s60main.rsc
cannot be tested on real device using infixed builds. Since this file
is unlikely to need changing, this should not be a big problem.
Task-number: QTBUG-9065
Reviewed-by: Jani Hautakangas
|
|
|
|
|
|
|
| |
RSS_RULES were being overwritten instead of being appended to
in applicaton_icon.prf.
Reviewed-by: Janne Anttila
|
|
|
|
|
|
|
|
| |
Also removed some IBY export paths that target layers below middleware,
as those are not likely to be ever needed by Qt projects, since Qt
itself is part of the middleware layer.
Reviewed-by: Janne Anttila
|
|
|
|
|
|
|
|
|
|
|
| |
The `check' rule ought to depend on the project being built, so one
can simply do `make check' instead of `make && make check'.
Note that there was already an attempt to do this for the autotests,
but it did not work on non-Windows platforms, because
$(DESTDIR_TARGET) is only used in the Windows makefile generators.
Reviewed-By: Lincoln Ramsay
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Usage:
# Set conditional libraries
LIB.MARM = "LIBRARY myarm.lib"
LIB.WINSCW = "LIBRARY mywinscw.lib"
LIB.default = "LIBRARY mydefault.lib"
# Set conditional Epoc Heap Size
EHZ.WINSCW = "EPOCHEAPSIZE 0x2000 0x2000000"
EHZ.default = "EPOCHEAPSIZE 0x40000 0x400000"
# Add the conditional MMP rules
MYCONDITIONS = MARM WINSCW
MYVARIABLES = LIB EHZ
addMMPRules(MYCONDITIONS, MYVARIABLES)
This will generate the following in the mmp file:
#if defined(MARM)
LIBRARY myarm.lib
EPOCHEAPSIZE 0x40000 0x400000
#elif defined(WINSCW)
LIBRARY mywinscw.lib
EPOCHEAPSIZE 0x2000 0x2000000
#else
LIBRARY mydefault.lib
EPOCHEAPSIZE 0x40000 0x400000
#endif
Task-number: QT-2909
Reviewed-by: axis
|
|
|
|
| |
Reviewed-by: Marius Storm-Olsen
|
|
|
|
|
|
|
| |
Builds the target executable if needed before running the test
LR on BKing's machine.
Reviewed-by: Lincoln Ramsay
|
|
|
|
|
|
|
|
|
| |
Libstdcpp.dll version autodetection was broken for clean builds,
so made it default for new version when neither new or old can be
detected.
Task-number: QT-1171
Reviewed-by: Janne Anttila
|
|
|
|
|
|
|
|
| |
Having dependency to moc.exe makes no sense since we cannot build it
anyway when building with Symbian toolchain.
Task-number: QTBUG-7836
Reviewed-by: Janne Koskinen
|
|
|
|
|
|
|
|
|
| |
A separate "check" rule added for running the tests. Use
"make check" to run the tests.
Task-number: QTPROD-442
Reviewed-by: Harald Fernengel
Reviewed-by: Rohan McGovern
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Fix bug where all target types, including apps, plugins and static
libraries were getting DEFFILE statements - now it's just libraries that
get it.
Fix bug where duplicate DEFFILE blocks were being added to projects that
manually specified their own DEFFILE - this now tests for the qmake
variable "defBlock" being set, and doesn't add additional DEFFILE
statements if it is.
NOTE: This means that adding DEFFILE statements to MMP_RULES
manually should be done by creating a variable called defBlock, and
adding that to the MMP_RULES (ie. MMP_RULES += defBlock)
Fix bug in configure.exe, where using -nokia-developer for Windows builds
would warn about Symbian DEF file usage (or lack thereof)
Reviewed-by: Janne Koskinen
Reviewed-by: Jason Barron
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Provide configure flag to enable/disable the use of DEF files on Symbian.
A useful side-effect was that it cleaned up how we control DEF files from
qbase.pri and in WebKit.
-nokia-developer still disables DEF files, as it triggers the autotest
exports, which are not frozen into the DEF files.
Disabling DEF files means that there is no BC with previously released
versions of Qt, so this should only be used for development purposes.
.pro files can specify custom locations for DEF files by setting
defFilePath.
Task-number: QTBUG-6556
Reviewed-by: Jason Barron
|
|
|
|
|
|
|
|
|
|
|
| |
changed /epoc32/include layout, including a subdirectory called "platform".
Since that particular directory is used as test for detecting SDKs with "New SF structure", the test returns with false result in case of pf_5250 (which has old SF structure).
Change was made to test for a more obscure directory "mw" that appears the be present in new structure only.
Merge-request: 2281
Signed-off-by: axis <qt-info@nokia.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Restore original Qt 4.4 behaviour. The capital P headers are not
correct because of frameworks on Mac.
More info, see thread:
http://lists.kde.org/?l=kde-multimedia&m=126045273702498&w=2
Task-number: QTBUG-4685
Reviewed-by: Trust Me
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In preparation for future removal of Webkit from Qt, Qt.sis no longer
contains QtWebkit.dll. All projects that have webkit dependency now
automatically also add dependency qtwebkit.sis into their .pkg files.
To create qtwebkit.sis, run "make sis" in src\3rdparty\webkit\WebCore
directory.
Task-number: QTBUG-6841
Reviewed-by: Simon Hausmann
Reviewed-by: Janne Koskinen
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
To support building Qt as part of Symbian SDK where epoc32 is not yet
populated, changed the autodetection logic in determining stlport
version. Now, we assume that we want the new version, unless only the
old version exists on SDK already.
Sqlite binaries export autodetection is now skipped if CONFIG value
symbian_no_export_sqlite exists, allowing clean builds to explicitly
suppress exporting.
Task-number: QTBUG-6971
Reviewed-by: axis
|
|
|
|
|
|
|
|
|
|
| |
This patch was written by Harald Fernengel for Maemo5 port. This effectively
just adds -g to QMAKE_CFLAGS & QMAKE_CXXFLAGS and is mainly for packagers who
want to build Qt in release mode and still have debug symbols, but will want
to strip those debug symbols out themselves (rather than let Qt do it).
Reviewed-By: Harald Fernengel
Reviewed-By: Thiago Macieira
|
|
|
|
|
|
|
|
|
|
|
|
| |
Some S60 SDK has a bug which make mifconv to crash with absolute paths.
Switched to use QMAKE_FILE_IN variable in mifconv.commands instead of
using join of ICON keywords. Joining is not needed anymore since ICON
keyword supports only one filename. Using input variable
(QMAKE_FILE_IN) also esnures that absolute path is converted to
relative before invoking mifconv.
Task-number: QTBUG-4745
Reviewed-by: Miikka Heikkinen
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
With SBSv2 toolchain only first file from ICON keyword is used and
it is actually the right behaviour. Originally support for multiple
files in ICON keyword was introduced to support localized application
icons. However, in that case the different localized icons should not
go to same MIF file since it is not possible to define in .rss file
which icon number should be used for which language.
To support a localized application icons separate MIF for each locale
should be created and installation of ICON should depend on selected
language. Implementing proper support for localized icons is left
for another task.
Task-number: QTBUG-4745
Reviewed-by: Miikka Heikkinen
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The linker warning that indicates symbol visibility changes is not useful
as it is an expected change to symbol visibility and produces lots of
clutter in the build logs.
(It is not desirable to follow the ELF spec when it comes to symbol
visibility in this case, which is why the linker warns and we ignore it)
This is likely to be a Raptor-only issue - I believe abld suppresses the
warning by default (at least in ABIv1 mode)
Also update gui.pro to use LFLAGS rather than MMP_RULES to alter the
arguments to ARMCC toolchain. qmake ought to detect this (incorrect)
usage of MMP_RULES and abort MMP file generation - raised QTBU-5961 to
look at this.
Reviewed-by: Shane Kearns
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This change makes it possible to define ICON keyword in pro file
without UID3.
It is safe to use target as an icon name since Symbian devices cannot
have two different binaries with the same name as all binaries are
located in \sys\bin. This means there cannot be also two mif files
with same target name. Also all native S60 apps use target name as an
MIF basename. Target name is also much easier to associate with app
than UID.
Task-number: QTBUG-4677
Reviewed-by: Miikka Heikkinen
|
|
|
|
|
| |
Task-number: QTBUG-5148
Reviewed-by: Tom Cooksey
|
|
|
|
|
|
|
|
|
| |
Moved duplicated and broken code for setting up the include and library
paths into the egl.prf feature file, which egl.pri and opengl.pro now
includes using the qmake feature system.
Task-number: QTBUG-5148
Reviewed-by: Tom Cooksey
|
|
|
|
| |
Does nothing if no declarative directories installed.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
introduce QMAKE_LIBS_OPENGL_ES1, QMAKE_LIBS_OPENGL_ES1CL and
QMAKE_LIBS_OPENGL_ES2 so we do not have to force users to copy/paste
around entire mkspecs just to change the OpenGL backend.
This should make the "-opengl es2" (and friends) configure option
work out of the box with any mkspec on Linux+WinCE.
Also removes a WinCE specific hack that is not required anymore.
Reviewed-by: Tom Cooksey
Reviewed-by: Trond Kjernåsen
Approved-by: Lars Knoll
|
|
|
|
|
|
| |
This is the same workaround as Janne did for QtWebkit.
Reviewed-by: Janne Koskinen
|
|
|
|
| |
RevBy: mread
|
|
|
|
|
|
|
|
|
| |
Using just -lQtUiTools instead of -lQtUiTools.lib will make qmake
attempt to autodetect whether or not the lib is static or dynamic,
which will not work since QtUiTools is not necessarily yet built.
Task-number: QT-1018
Reviewed-by: Janne Koskinen
|
|\ |
|
| |
| |
| |
| |
| |
| | |
This is required by the Helix plugin. As discussed with Dallas team/Fu Liz.
Reviewed-by: Shane Kearns
|
|/
|
|
|
|
|
|
| |
Fix for applications not compiling containing #include<QtWebkit>
due to relative inclusion problem in Symbian build environment.
Task-number: QTBUG-4846
Reviewed-by: Iain
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The VCPROJ/SLN generator generates dependencies on the glue project,
so the glue must use the correct library for the dependencies to be
correct. The qtAddLibrary() would add the project 'default' to the
glue, which could end up adding a debug lib to the glue, so the
dependency checker wouldn't find the proper project.
We therefore force qtAddLibrary to only add d/_debug if we're not
using debug_and_release, or not in the glue part
Reviewed-by: Rohan McGovern
|
|
|
|
|
|
|
|
|
| |
It was decided on weekly telco that Symbian pkg and sis files can use
plain Qt name, since it is already clear that user is installing
"Qt for Symbian" version of Qt.
Task-number: QT-772
Reviewed-by: Miikka Heikkinen
|
|\ |
|
| |\ |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Prepending adds epoc32\include too early into generated
systeminclude path search order causing Webcore build
to break due to wrong animation.h header getting included.
Reviewed-by: Jason Barron
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The new public classes in the 's60framework' sub-directory of QtGui
introduce a header file dependency to the middleware layer of the SDK
by including 'aknapp.h'.
Since the build system doesn't know if you have included QtGui or not,
the closest approximation is to append the middleware layer include
path in cases where the QT variable contains 'gui'
Reviewed-by: axis
|
| |/
| |
| |
| |
| |
| |
| |
| | |
Phonon.
Both epoc32/include and $QTDIR/include/Phonon contain a file called videoplayer.h. Both of these directories are listed as SYSTEMINCLUDE paths in the generated MMP file, with the Phonon path coming first. This means that '#include <videoplayer.h>' picks up the Phonon header rather than (as intended) the Symbian one.
A new qmake variable, PREPEND_INCLUDEPATH, is defined, allowing the .pro file to specify that /epoc32/include should be the first SYSTEMINCLUDE.
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Some changes was done on Mac to support Qt building statically
on Cocoa. Unfortunatly, the change lead to many visibility warnings
because, with the change, src/tools/bootstrap was made with symbols
hidden by default (-fvisibility=hidden). Then, when building the
tools (moc, rcc etc), visibility was public by default (for internal
classes used by Qt). This caused warnings. The change is now temporarily
reverted until Morten get a second chance to look at it.
Rev-By: prasanth
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This introduces an undocumented "phonon_compat" subdir and also adds
it to the INCLUDEPATH when QT += phonon is specified. With this, these
styles of #includes should be supported:
#include <Phonon>
#include <Phonon/Phonon>
#include <phonon>
#include <phonon/phonon>
#include <Phonon/MediaObject>
#include <phonon/mediaobject.h>
Still need to check if the headers get installed during "make
install". I couldn't find where in our code that is done.
BT: yes
Reviewed-By: Trust Me
|
| |
| |
| |
| | |
Reviewed-By: TrustMe
|
| |
| |
| |
| |
| |
| |
| | |
Several script and webkit tests require increased maximum heap
in emulator, so just increase the maximum heap for all autotests.
Reviewed-by: Janne Koskinen
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This is a workaround for the fact that removing --export_all_vtbl from
the RVCT command line seems to cause problems when linking the static
s60main lib with apps built with GCCE (at least 4.3, haven't tested
earlier versions)
The root cause still needs to be found and fixed, after which this
workaround can hopefully be removed.
Problem observed on Symbian^3 release, RVCT 2.2 build 686,
GCC-E 2009q1-162 release from CodeSourcery.
Reviewed-by: axis
|
| |
| |
| |
| |
| |
| |
| | |
SBSv1 has also been updated to define the appropriate macros and add the
required library statements, so all that is needed is the STDCPP keyword
Reviewed-by: mread
|
| |
| |
| |
| |
| | |
Add -all_load to make Objective-C categories work.
Enable hide_symbols to remove visibility issues.
|
|/
|
|
|
|
|
| |
Skipped addExclusiveBuilds function in debug_and_release.prf for all
Symbian platforms, as it doesn't work with Symbian toolchain.
Reviewed-by: Janne Anttila
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This way we avoid having a lot of code in a static (and
unmaintainable) library. The s60main static library now currently has
only one task: to call main().
To move the initialization into QtGui also meant a change in how the
S60 framework is created, because we can no longer use the trick
where we create and start the the S60 event loop and then have the
framework call us back to start main(). The initialization now
follows the creation and destruction of QApplication, which is a lot
more in line with how other platforms do it.
Since S60 doesn't support creating the environment, and *then*
starting it (both are executed by the same call), we had to open up
the S60 framework construction classes and just mirror what they do.
This means that after QApplication construction is done, the S60
framework is initialized, but nothing will run yet and control will
return to main(), where the user can start the event loop himself.
One of the quirks of this approach is that the construction of the
S60 framework makes a new cleanup stack. This means that any active
traps will not be active anymore, and leaving without setting a new
trap will most likely panic. This shouldn't be a problem for us,
since Qt is never supposed to leave, but it means that if anyone uses
the cleanup stack without setting a new trap, they will receive a
panic.
It was considered to add a trap mark in QApplication construction and
then removing it on destruction, but it was dropped because leaving
from main() is still undefined (even if the old cleanup stack would
be restored in the destructor, we wouldn't be able to stop the
exception from unwinding the stack, and the cleanup stack would then
be unbalanced).
RevBy: Jason Barron
RevBy: Janne Anttila
AutoTest: QWidget passed with same failure count
|