that way the users don't have to duplicate the connect everywhere.
as a side effect, MaemoPackageContents will have error reporting now.

externalize the error message and evaluation notification callbacks into
an own handler interface.

more types, and callback even for top-level project file

feature files are already generally excluded. add the explicitly
injected files (qmake.conf and .qmake.cache) to that. this makes
path-based exclusion superfluous.

this fixes the recent appearance of qt_webkit_version.pri in all qt 4.7
based projects.

in fact, the patch does somewhat more: it makes the evaluator provide
the information to the higher layers which is necessary for a truly
hierarchical display of includes. TBD later.

Reviewed-by: dt
Task-number: QTCREATORBUG-1590

This is a big change touching almost all of our .pro file parsing.
With this patch we only evaluate once exact for all needs and once
greedy for the filelist. That is the qt runconfigurations don't have own
evaluaters but reuse the project wide exact evaluation.

We reevaluate if the user changes the build directory, the qmake
buildconfiguration or the qmake arguments. That is if you open src.pro
(or projects.pro) of qt with a shadow build you still don't get all the
files, but after correcting the build directory, we reevaluate the .pro
files and find all files. So for a suitable definition of fixed, that
bug is now fixed.

We now get the exact defines of all .pro files instead of all defines for all
buildconfigurations. We still don't distinguish in which
.pro file a DEFINE is set. So the code model now knows about all the
defines set for the given configuration but not for which files it is
actually set. Also that includes all DEFINES set in .qmake.cache or the
mkspecs. This means all defines from .pro files should now work.

The intial loading is still synchronous. I haven't looked into it to
deeply, but it seems possible to make it also async.There are probably a
few issues which need to be solved fist.

Also due to the asynchronous nature of the code, the executable is
updated a few seconds after actually changing the build configuration

no point in re-parsing them over and over during the scan of a bigger
project ...

shaves off another 20% of the loading time of Qt 4.6.

there is no point in throwing away and re-recreating invariant data over
and over ...

shaves off ~15% of the load time of Qt 4.6.

Reviewed-by: dt <qtc-committer@nokia.com>

they were added to facilitate .prf exclusion from the project tree, but
that should be handled differently already.

that way they are auto-inherited by nested parsers, and it's the right
thing to do anyway.

almost too simple to be right, given the TODO ...

- parse .qmake.cache, remove workaround for not doing it
- look for the mkspecs in all the places qmake would do
- use the passed in properties instead of QLibraryInfo to find the qt
  data dir
- fix loading of generic features from within os-specific ones with the
  same name
- add multiple inclusion guard for features (including what seems to be
  a bug in qmake ...)
- misc cleanups and beautifications

to be able to share the option object among evaluators, it is passed by
pointer.

only field_sep stays static and becomes a string - that's effectively
just a cache to avoid repeated construction.

There were both variants, with and without extra semicolon.

this also removes the bogus special casing of various
filename-containing variables inside the pro parser.

Reviewed-by: Trust Me

Conflicts:

	src/libs/cplusplus/CppBindings.cpp
	src/libs/cplusplus/CppBindings.h
	src/libs/cplusplus/FastPreprocessor.cpp
	src/libs/cplusplus/FastPreprocessor.h
	src/plugins/cpptools/cppcurrentdocumentfilter.cpp
	src/plugins/cpptools/cppcurrentdocumentfilter.h
	src/plugins/cpptools/cppsemanticsearch.cpp
	src/plugins/cpptools/cppsemanticsearch.h
	src/plugins/duieditor/parser/qmljs.g
	src/plugins/duieditor/parser/qmljsast.cpp
	src/plugins/duieditor/parser/qmljsast_p.h
	src/plugins/duieditor/parser/qmljsastfwd_p.h
	src/plugins/duieditor/parser/qmljsastvisitor.cpp
	src/plugins/duieditor/parser/qmljsastvisitor_p.h
	src/plugins/duieditor/parser/qmljsengine_p.cpp
	src/plugins/duieditor/parser/qmljsengine_p.h
	src/plugins/duieditor/parser/qmljsgrammar.cpp
	src/plugins/duieditor/parser/qmljsgrammar_p.h
	src/plugins/duieditor/parser/qmljslexer.cpp
	src/plugins/duieditor/parser/qmljslexer_p.h
	src/plugins/duieditor/parser/qmljsmemorypool_p.h
	src/plugins/duieditor/parser/qmljsnodepool_p.h
	src/plugins/duieditor/parser/qmljsparser.cpp
	src/plugins/duieditor/parser/qmljsparser_p.h
	src/plugins/duieditor/parser/qmljsprettypretty.cpp
	src/plugins/duieditor/parser/qmljsprettypretty_p.h
	src/plugins/duieditor/rewriter/rewriter.cpp
	src/plugins/duieditor/rewriter/rewriter_p.h
	src/plugins/duieditor/rewriter/textwriter.cpp
	src/plugins/duieditor/rewriter/textwriter_p.h
	src/shared/cplusplus/ASTClone.cpp
	tests/manual/cplusplus/main.cpp

Reviewed-by: Trust Me

After all I don't need all that stuff in the cmakeplugin, all i needed
is now bundled in debugginghelper.h/cpp

The Debugger, cmake and Qt4 plugin all need it. More refactoring to it
coming.

Checking paths against each other is a little more complex.

Reported on irc, and probably also have a task for that. Simply
forgotten when i first implemented it.

Details:  Free the whole ProFile*, ProItem*, ProValue* and etc
structure after parsing, that frees around 200Mb for projects.pro.

Task:     Reported by eskil on irc.

Details:  ProFileCache be gone, as is ManagedProFile, not to be missed.
Drastically simplified how ownership works.