Commit Graph

397 Commits

Author SHA1 Message Date
Mark Brand
f4f0c8b99b Merge "Merge branch 'buildsystem' (second round)" into refs/staging/master 2012-06-29 21:26:11 +02:00
Robin Burchell
ddf7233125 Add mkspecs for android cross compilation.
Using these requires setting two environment variables, e.g, for me:

export ANDROID_NDK_ROOT=/Users/burchr/android-ndk-r7c
export ANDROID_NDK_HOST=darwin-x86
./configure -opensource -confirm-license -xplatform unsupported/linux-android-armeabi-v7a-g++ -nomake examples -nomake demos -nomake tests -v

These mkspecs are somewhat based on the work of the Necessitas crew, kudos to
them for their work in getting the NDK integration into qmake.

Change-Id: I591e423ed8dc70616009f681c81890c696110e62
Reviewed-by: Girish Ramakrishnan <girish.1.ramakrishnan@nokia.com>
2012-06-29 21:15:46 +02:00
Oswald Buddenhagen
1491065506 Merge branch 'buildsystem' (second round)
refactoring and cleanup. fixes x-builds between different os families.

Conflicts:
	mkspecs/features/qt.prf

Change-Id: I0205e6f07f77c9b015cf055dd87a471883949a91
2012-06-29 09:06:37 +02:00
Thiago Macieira
96166fa56a Update the macros for shared/DLL and static builds
Up until now, we had a mess of different macros used for building
DLLs, for building shared libraries on Unix systems and for building
static libraries. Some of the macros were contradictory and did not
work. From now on, there shall be only:

 - QT_STATIC: indicates that it's a static Qt build and the export
   macros should expand to empty
 - QT_SHARED: indicates that it's a shared / dynamic Qt build and the
   export macros should expand to Q_DECL_EXPORT or Q_DECL_IMPORT,
   depending on whether the macro corresponds to the current module
   being built (the QT_BUILD_XXXX_LIB macro comes from the module's
   .pro file)

QT_BOOTSTRAPPED implies QT_STATIC since the bootstrapped tools link
statically to some source code.

QT_STATIC is recorded in qconfig.h by configure when Qt is configured
for static builds. Nothing is recorded for a shared / dynamic build,
so QT_SHARED is implied if nothing is defined. This allows for the
existence of a static_and_shared build: with nothing recorded,
defining QT_STATIC before qglobal.h causes the export macros to be
that of the static form. Linking to the static libraries is out of the
scope of this change (something for the buildsystem and linker to
figure out).

From this commit on, the proper way of declaring the export macros for
a module called QtFoo is:

    #ifndef QT_STATIC
    #  ifdef QT_BUILD_FOO_LIB
    #    define Q_FOO_EXPORT Q_DECL_EXPORT
    #  else
    #    define Q_FOO_EXPORT Q_DECL_IMPORT
    #  endif
    #else
    #  define Q_FOO_EXPORT
    #endif

The type of the Qt build is recorded in QT_CONFIG (in qconfig.pri) so
all Qt modules build by default the same type of library. The keywords
are "static" and "shared", used in both QT_CONFIG and CONFIG. The
previous keyword of "staticlib" is deprecated and should not be used.

Discussed-on: http://lists.qt-project.org/pipermail/development/2012-April/003172.html
Change-Id: I127896607794795b681c98d08467efd8af49bcf3
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
2012-06-28 06:21:48 +02:00
Stephen Kelly
c4a05f5e3a Refactor the creation of imported targets into a macro.
This will make it easier to handle cases where the Qt installation
and the build do not match exactly. For example, on Windows, it
would be possible to do:

  ./configure -debug
  make
  make install
  make release
  make release_install

In which case, both debug and release libraries would be installed
even though it was only configured with -debug. On non-Windows, the
debug and release libraries would overwrite each other, so it is
not necessary to support it.

Similarly, we want to handle cases in the future where (on
non-Windows) both static and shared libraries would be
installed (again, not described with a single build configuration).

Change-Id: Ib7916c9664a0f72e40156a03bdfc79a4a6c24350
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>
2012-06-27 14:43:00 +02:00
Oswald Buddenhagen
a2382f7f15 unset QT_MODULE_* after loading the module pris
abuses have been observed in the wild, so make sure these variables are
not available.

Change-Id: I502c3f5db7d341cf6a8bd2ec09e87f129da2fca6
Reviewed-by: Mark Brand <mabrand@mabrand.nl>
2012-06-27 14:35:31 +02:00
Oswald Buddenhagen
b40b91cb39 move initialization of numerous variables to spec_pre.prf
this is cleaner by design and allows removing some hacks.

Change-Id: I3270195b5d62caa476ffde7c1e1ef43cec99c565
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
Reviewed-by: Mark Brand <mabrand@mabrand.nl>
2012-06-27 14:35:25 +02:00
Oswald Buddenhagen
dafc4a884c don't put QMAKESPEC_ORIGINAL into INCLUDEPATH any more
the spec does that itself, given that the real spec is just included
nowadays, instead of copied (which never worked without side effects).

Change-Id: Ibf655b9a943dadb949d3c7a58d8fe50fcd62cef7
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-27 14:35:24 +02:00
Oswald Buddenhagen
caa2263051 use real qmakespec as basis for finding features
doesn't make much of a difference on unix (as the default specs are just
symlinks).
on windows, it makes the gross hack used for finding spec-specific wince
default_post.prfs unnecessary.

Change-Id: Id403dce5be487e1ae22c1f54b8095a6afdd98bc8
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-27 09:39:51 +02:00
Oswald Buddenhagen
4753958db7 revamp handling of qmake target mode
the project evaluator becomes oblivious of the target mode.
the mode is set up in spec_post.prf according to the spec.
$$QMAKE_TARGET contains the feature suffixes to search, and is also
contained in $$CONFIG.
the target_mode variable itself becomes private to the Makefile class.

Change-Id: I3c06d9dab536b753343cec6c5c491d3203e50bd8
Reviewed-by: Mark Brand <mabrand@mabrand.nl>
2012-06-27 09:39:50 +02:00
Holger Hans Peter Freyther
8e5b6647a9 mkspecs: Use the right variable name to find the qdevice.pri
Commit 8b822825c5 introduced the
/get version but used the wrong variable name. Fix it by using
QT_HOST_DATA.

Change-Id: Ia4759b8c6ff2de9726f3aebae2f2f39c6644d4ec
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
2012-06-26 20:02:55 +02:00
Oswald Buddenhagen
34cb86d46b make using testlib automatically "promote" all modules to -private
autotests often need private headers (especially with qpa headers now
being private) and have no compatibility requirements, so it makes sense
to just use the privates of requested modules.

this also suppresses the useless warning about using privates, in case
they are still explicitly specified.

Change-Id: I9e499bedcf6ef25777283ff1432cef7254e9093a
Reviewed-by: Girish Ramakrishnan <girish.1.ramakrishnan@nokia.com>
2012-06-26 11:51:34 +02:00
Stephen Kelly
c42a0a7729 Factorize the cmake test infrastructure to be reusable.
This makes it easy to add cmake module tests for all modules.

Change-Id: I303bf7674ca6ae7a8544488f96e8e02afbaa6ff0
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
2012-06-26 11:49:52 +02:00
Konstantin Ritt
a9ba927406 Avoid macro re-definition after 1838a6c2fb
Change-Id: I2631ff687881278ba42edacd237a2130c103b69f
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
Reviewed-by: Rafael Roquetto <rafael.roquetto@kdab.com>
2012-06-26 11:37:57 +02:00
Stephen Kelly
fa6ca58b8b Don't re-evaluate the lib location.
Rely on the DESTDIR variable being set correctly by qt_module_config.

Change-Id: I1a166124024722ec5a189a7402b38646179aa890
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
2012-06-26 04:18:29 +02:00
Stephen Kelly
70c981f973 Fix the DEFINITIONS for Qt modules.
As the DEFINITIONS to be used for QtAddOns is different to essential
modules, rely on the logic in qt_module_config setting this variable
correctly.

Change-Id: I64485ccd6df093216cac4a97fb1cfaac0122a218
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
2012-06-26 04:18:25 +02:00
Oswald Buddenhagen
d921182310 add possibility to set $$MODULE_IMPORT_SUFFIX in module .pri files
quick1 got a suffix, so we need a way to set it

Change-Id: I099b868106abd4d3040047703472faa65f694f31
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
2012-06-24 15:26:34 +02:00
Donald Carr
e6ccfe40eb Set Wayland as default platform on Raspberry-Pi
Update Raspberry-Pi mkspec to indicate that wayland is the default platform
for the Raspberry-Pi.

Change-Id: I10b30ecfb16faed6027137225d9e95409faa7e87
Reviewed-by: Johannes Zellner <johannes.zellner@nokia.com>
Reviewed-by: Girish Ramakrishnan <girish.1.ramakrishnan@nokia.com>
2012-06-24 05:26:10 +02:00
Stephen Kelly
3d2a8d09d1 Add a variable for the private includes for modules.
This is required to use QPA for example in 5.0.

Change-Id: I44bfc6987d778370e55c05c591f63ff84c482d0a
Reviewed-by: David Faure <faure@kde.org>
2012-06-23 14:16:24 +02:00
Girish Ramakrishnan
7042de0894 eglfs: move m_pos into Cursor
Change-Id: I51e92d265e7ea6372ae58b357f75362e2d9a2df9
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>
2012-06-22 15:42:44 +02:00
Girish Ramakrishnan
57ca02b1d2 embedded doesn't exist anymore
Change-Id: Ib56895defb72f51bb790994b9fd72fc81ef25af5
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
2012-06-22 07:41:31 +02:00
Bradley T. Hughes
1618ef236b mkspecs: Add macx-clang-libc++ and macx-clang-libc++-32
Use these mkspecs to switch to the new libc++ C++ runtime library, which
in turn makes it possible to enable C++11 support with clang.

Change-Id: If92908592f8bee4829a1bad747fe396f527d26c7
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
2012-06-20 15:45:07 +02:00
Oswald Buddenhagen
2f1762535b make missing qt modules fatal
there is no point in continuing; it will error out later (in more
cryptic ways) anyway.

Change-Id: I2562eadc2d91bfaa6e85368dc3774da31409a141
2012-06-19 16:46:09 +02:00
Oswald Buddenhagen
0fdf7c3a43 less confusing wording
Change-Id: Ia509fe1a28de803673623d425f5949b35404598c
2012-06-19 16:46:09 +02:00
Oswald Buddenhagen
1ee462604b fix misnomer: qt_module.prf => qt_build_config.prf
qt_module suggests to be congruent to qt_plugin.

Change-Id: I629530bcbe2ba6c0adbdc11a275119c8aff0c953
2012-06-19 16:46:08 +02:00
Oswald Buddenhagen
b3a497d781 fix regexp in qmodule.pri inclusion check
this also allows us removing the redundant flag variable

Change-Id: If26d22280ce7bc2ef3514c7c11ce7cd9baec1293
2012-06-19 16:46:08 +02:00
Oswald Buddenhagen
575a516630 revamp automatic makefile generation for sub-projects
instead of making the "real" targets depend on the makefiles, add
conditional makefile generation to the targets themselves.
this causes makefile generation to follow the recursion order determined
by the project, which is important when dealing with prl and module pri
files.
a side effect of this is that qmake and make calls are interleaved now,
which is entirely different from a 'qmake -r' run.

on the downside, calling make with multiple targets which operate on the
same subprojects without prior makefile generation will make a mess, as
the qmake calls will be racing. this should be no problem, as qmake does
not generate recursive targets where this would be useful - at least by
default.

it is not sufficient to just order the creation of the makefiles
non-recursively (e.g., by using gnu-specific order-only-prerequisites),
as an interrupted and subsequently resumed build would happily skip the
nested makefiles.
workable alternative approaches would be walking the entire tree in a
pre-pass to ensure makefile presence (which is incredibly slow) or
creating additional stamp files only after recursing and having the
makefiles depend on them (which is ugly).

Task-number: QTBUG-23376
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
Change-Id: I88d3e7610215677d362026de316513d3bea04b06
2012-06-19 16:46:08 +02:00
Oswald Buddenhagen
e6d97f3f77 support for private depends in modules
if a module's private headers add additional dependencies, QT_PRIVATE is
now the place to declare them. note however, that this may not contain
other private headers in turn - that would be much harder to implement,
and we want the explicitness anyway.

Change-Id: Ic516fcf1a003c95798df4fbe216f92016afaf47e
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
2012-06-19 16:46:08 +02:00
Oswald Buddenhagen
5290f6df89 introduce $$QMAKE_CHK_EXISTS_GLUE
instead of hard-coding platform differences, use a variable.

Change-Id: I20e98811ad5f07429148c6f88aedbabc3ba58fff
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-19 16:46:07 +02:00
Oswald Buddenhagen
45a201b6a3 centralize definitions of shell commands
there are only two types. everything else is duplication.

Change-Id: I87f2bdd3d56b94bb2ecdb60e8861afeb9af3666f
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-19 16:46:07 +02:00
Oswald Buddenhagen
3c144b54cf normalize $$QMAKE_QMAKE
everything in the projects should be normalized. only the makefile
generators need to adjust it to the native form.

Change-Id: I06a4e997f32134d13949ec4a9dd1b44367aab7cb
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
2012-06-19 16:46:07 +02:00
Oswald Buddenhagen
2cfde5d9c3 correctly find idc even for makefiles
this is a qt tool, so it needs appropriate treatment

Change-Id: I0cb30ba07e03c72ee275cd916ca0a39a99fc3705
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
2012-06-19 16:46:07 +02:00
Oswald Buddenhagen
98470fd9be scan module path backwards
the last read file wins, so reading in inverse order ensures that we
respect the list's sorting by decreasing priority.

Change-Id: I2e6539a52d4195ed6af4c0143b035c39577b8310
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-19 16:46:04 +02:00
Oswald Buddenhagen
a56ef02b2d let default_pre add modules to the qmake path
that way qmake is made aware of the forwarding pris which are generated
for this module even when a top-level .qmake.cache prevents the module's
root from being found automatically.
the path is also added to the cache, so that subsequent partial
qmake-ing of the tree will still find the module.

this also makes the -cache-module-fwd parameter of syncqt useless, so
remove it.

Change-Id: I2afbc52a465c0b3260e9bcaf032c43a82ae8061f
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-19 16:46:03 +02:00
Oswald Buddenhagen
a943f40483 recognize QMAKEMODULES env var, property and cache variable
the latter allows sub-projects to dynamically extend the qmake search path
specifically for modules. the others are just for congruency with
QMAKEPATH and QMAKEFEATURES.

Change-Id: I0c099035f8dc8ee8645566dbc635644a15ed9da5
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-19 16:46:03 +02:00
Oswald Buddenhagen
32042413f2 move generation of module master headers to qt_module_headers
this has the advantage that the %mastercontent assignments in sync.profile
are not necessary any more. as it happens, most modules got them wrong
anyway.

Change-Id: Ibdf689be408f18e1d90c44ef4ecacd7c24b1f1c9
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-19 16:46:01 +02:00
Oswald Buddenhagen
eae5135808 remove detection-based usage of modules which are frameworks
each qt module comes properly declared and located, so there is simply
no point in performing a search.

Change-Id: I86fad21bb8e128b85f1000cc116cc44a23642eb4
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-19 16:46:01 +02:00
Oswald Buddenhagen
a0374fd490 fix direct includes from modules which are frameworks
Change-Id: I1f2c1e5ebdb2fa258500a08b32dcafb4150ee424
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-19 16:46:01 +02:00
Oswald Buddenhagen
5cea63d2f2 let modules which are frameworks remember that fact
this will allow us to not rely on the modules matching the global
qt_framework setting.

Change-Id: Ic1dce757ff63d06af54a2428e23a1bbcf1c81ba1
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-19 16:46:00 +02:00
Oswald Buddenhagen
88a07df5e5 remove workarounds for broken syncqt
SYNCQT.HEADER_FILES contains qconfig.h just fine nowadays.

Change-Id: I52b35d2ac41d772acaef06f3931df75742c867cb
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-19 16:46:00 +02:00
Oswald Buddenhagen
c595988a58 automatically add QT_BUILD_FOO_LIB to DEFINES
Change-Id: I35d9861e48469eb5cc8824e361450684047e6559
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-19 16:46:00 +02:00
Oswald Buddenhagen
0f548b1744 remove backwards compatibility hacks
there won't be terribly many projects relying on it. now's the time to
find out for sure ...

this reverts commit 3279b07302fde0eb14f9b197c9ad2e14d512817e

Change-Id: Id36687ab3bfc7dd5ce35b584621a8f5b3ee00fc9
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-19 16:46:00 +02:00
Oswald Buddenhagen
131ba187f2 auto-generate QMAKE_PKGCONFIG_REQUIRES and QMAKE_PKGCONFIG_DESCRIPTION
less boilerplate, more accuracy

Change-Id: I6cc2abd50eafb4901d987c122f10a62ec9ea9da3
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-19 16:45:59 +02:00
Oswald Buddenhagen
f3cca88ac4 move QMAKE_PKGCONFIG_VARIABLES stuff out of qt_module_config.prf
this only needs to be set in one module each - the one which provides
the relevant tool.

this is moderately source-incompatible, in that a package which queries
a given variable from the wrong library won't get the path it looks for
any more. as it's likely that everyone was using QtCore as a reference
anyway, this will only affect uic - which is in the new QtWidgets
library, to which people need to adjust anyway.

Change-Id: If05d3c33fda6cd12466e261391b825c59651d3e4
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-19 16:45:59 +02:00
Oswald Buddenhagen
19009cde45 remove ACTIVEQT_IDC assignment from QMAKE_IDC
the specs have been cleaned of it

Change-Id: Ica0ab4d5f9d36f1f9566759326f3b3144e34cd67
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-19 16:45:59 +02:00
Oswald Buddenhagen
65c1ed6281 remove support for CONFIG+=qt_compat
this is truly prehistoric stuff

Change-Id: Ia781fc03c081bff895f4d2b5325d3624441bdf5f
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-19 16:45:58 +02:00
Oswald Buddenhagen
8b75fc7bac don't explicitly inject qt header/library paths
the modules already do that on demand.

Change-Id: I889a5cd7b62631e2551ada4b096f20b3b79b2c75
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-19 16:45:58 +02:00
Oswald Buddenhagen
4816ca6c56 always add library path from module
that way we can theoretically support modules outside $QTDIR.
also, it's just cleaner.

Change-Id: I6139ebc7328b64ace8552b3e54f9a8c69248ceec
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-19 16:45:58 +02:00
Oswald Buddenhagen
5d1b4c79d2 make module includepath addition less convoluted
don't prepend the paths, as this will only mess up building of
subsequent modules (e.g., building qtdeclarative against an installed
qtbase would pick up the headers from the qtdeclarative previously
installed into the same directory as qtbase).
for frameworks this was a rather pointless exercise in the first place,
as their headers are properly isolated anyway.

however, make sure that we don't add system locations to the search
paths, as this is a) unnecessary and b) messes up subsequent libraries
in non-standard locations which want to shadow versions in standard
locations (pkg-config .pc files which add standard paths are considered
broken as well).

Change-Id: Ie1dc65d4767e98e1df6e49012505141935a6c704
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-19 16:45:58 +02:00
Oswald Buddenhagen
46ae5598e5 add qtCompileTest()
this fully replaces qtmodule-configtests.
it is way shorter and it actually integrates reasonably with qmake.

Change-Id: I819cc6807ad3661c419b54fa253894936dd88a64
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
2012-06-19 16:45:57 +02:00