Only revert stuff that is linked to fast string concat.

This reverts commit e70530c5. It
did break builds on older gcc versions (gcc 4.1.2 64 bit Linux,
gcc 4.0.1 on Mac OS X):

parser/qmljsgrammar_p.h: In static member function ???static int
QmlJSGrammar::nt_action(int, int)???:
parser/qmljsgrammar_p.h:188: error: ???QmlJSGrammar::<anonymous enum>???
is/uses anonymous type
parser/qmljsgrammar_p.h:188: error:   trying to instantiate
???template<class T> struct QConcatenable???
...

Reviewed-by: Oswald Buddenhagen

this makes the i/o part of the evaluator thread-safe.
for safety, assert absolute paths in IoUtils::exists().

on the way, i "stole" some code i wrote for KDE. there have been no
copyright-worthy contributions from others to it, so this is legal.

somewhat faster again

now items have no vtable any more

it's overengineered for our purpose and would just get in the way of
planned optimizations.

use a hash of magic variable names to ints, which allows fast lookup
followed by a jump table dispatch

doing it in 1.3 as well to avoid possible later conflicts

potentially reduces the number of file::exists() calls

partly by removing unnecessary calls, partly by providing minimalistic
reimplementations.
this gives a quite incredible performance boost ...

this gives a rather impressive speed boost.
if qmake wasn't trying to be so clever, we could get even more
out of this.

that way we save the creation of a temporary stringlist

now that we know the size of the entire file, we can use a single fixed
buffer instead of growing one on-demand - in the case where we just read
the file, we can even re-use the file buffer itself.
this should improve performance for files with very many line
continuations, but the practical impact drowns in the noise.
anyway, the code is nicer, as we don't abuse qstring that obviously any
more.

again several percent faster ...

in release mode they expand to nothing ... not so in debug mode.

- nest conditionals properly to avoid unnecessary tests
- avoid usage of QString::append(QChar) to save allocations

the latter will do it anyway.
this eradicates the performance penalty from the previous commit at its
root.

creating tons of items is a tad allocation-intensive and thus slow.

this is probably slower for often-included files, as now the splitting
is done in every evaluation pass.

now that the tree will not be written out any more, we can save some
cpu and memory by not recording the information.

that way the file formatting is better preserved.

they were remnants of the pro editor, which was removed long time ago.
they were still used for automatic editing of pro files, but were
totally overengineered for that purpose (removed 1300 LOC, added 40).

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.

makes forward delcs in headers possible

Move the code that compares the Qt version's debug|release property
with the build configuration's debug|release property
to the new build configuration class.

Reviewed-by: dt

These .pri files which add to INCLUDEPATH and have FORMS should also add
to DEPENDPATH.

Adding to INCLUDEPATH does not add to DEPENDPATH by default with qmake.
That means qmake may not find headers in those paths and won't
consider them dependencies when compiling the source files which include
them.

This doesn't matter when all the headers in the new INCLUDEPATH are
always present, but when some of them are generated (e.g. from FORMS),
the dependencies are important and omitting them means a race condition
in parallel builds.

Fixes errors of the form:
  ..\..\shared\help\topicchooser.h(33) : fatal error C1083: Cannot open
    include file: 'ui_topicchooser.h': No such file or directory
(cherry picked from commit 9a7e0294)

These .pri files which add to INCLUDEPATH and have FORMS should also add
to DEPENDPATH.

Adding to INCLUDEPATH does not add to DEPENDPATH by default with qmake.
That means qmake may not find headers in those paths and won't
consider them dependencies when compiling the source files which include
them.

This doesn't matter when all the headers in the new INCLUDEPATH are
always present, but when some of them are generated (e.g. from FORMS),
the dependencies are important and omitting them means a race condition
in parallel builds.

Fixes errors of the form:
  ..\..\shared\help\topicchooser.h(33) : fatal error C1083: Cannot open
    include file: 'ui_topicchooser.h': No such file or directory

evaluating a file (which might not exist) is a bit different than just
merging a (possibly empty) map ...