| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
Reviewed-By: Trust Me
|
|
|
|
|
|
|
| |
Functional coverage of QDir, except for trivial shims for deprecated APIs
and the QDebug streaming functions.
Reviewed-by: joao
|
|
|
|
|
|
| |
Missing return statement
Reviewed-by: joao
|
|
|
|
|
|
|
|
|
|
|
|
| |
Adding tests for these functions:
symLinkTarget / readLink (static & non static)
permissions (static)
constructors which take a QObject parent
After this, all function of QFile API are tested except encoding/decoding
functions and some QT3_SUPPORT inlines in the header.
Reviewed-by: joao
|
|
|
|
| |
Reviewed-By: joao
|
|
|
|
| |
Reviewed-By: joao
|
|
|
|
| |
Reviewed-By: joao
|
|
|
|
| |
Reviewed-By: joao
|
|
|
|
| |
Reviewed-By: joao
|
|
|
|
|
|
|
| |
The existing dir test function is appropriate, now it checks the dir
functions as well as the path functions.
Reviewed-By: Prasanth Ullattil
|
|
|
|
|
|
|
|
|
|
|
| |
Creating symbolic links requires admin privilege by default - when UAC
is enabled then even users in the administrators group do not have admin
privileges without an escalation prompt. (which is inappropriate for an
autotest).
When run with insufficient privileges, the test is skipped with a warning
Reviewed-By: Prasanth Ullattil
|
|
|
|
|
|
|
| |
EACCES was not defined at this point, but it is also not needed, since
we know exactly why unmap is failing.
Reviewed-by: Shane Kearns
|
|
|
|
| |
Reviewed-by: Shane Kearns
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Memory mapping of RFile handles is not supported in current symbian OS
versions. However the "open C" libraries provide an emulated mmap()
implementation which was used by QFile::map in Qt 4.6 release.
To avoid breaking applications which rely on this function, QFile::map
will now open the file with open C as well in order to use that handle
to call mmap().
When symbian implements a file mapping API, we can switch the implementation
to use that for RFile handles.
Reviewed-By: joao
|
|
|
|
| |
Reviewed-by: Joao
|
|
|
|
|
|
|
|
|
| |
Since windows doesn't have the concept of user and group IDs, the name
resolution has to be done differently. If the qt_ntfs_permission_lookup
flag is set (default = 0), the user information is retrieved through the
NTFS object security information.
Reviewed-by: Joao
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The filesSelectedSignal() test case was using the Qt source tree as its
test directory - however this is not deployed to the device for symbian
and windows CE.
On symbian, the path evaluates to the root directory, which doesn't meet
the conditions of the test (it contains only directories, while the test
requires both files and directories to be present)
Changed the test to use SRCDIR (on desktops, the source directory of the
test code; on symbian/wince the deployed files location)
This directory is valid for the test as it contains both files and directories.
The hooks() test case crashes on symbian versions prior to symbian^3,
because data exports from dlls are not properly supported by the symbian kernel
So, this test is skipped for old symbian versions (if a test crashes, then
no xml result file is created)
Reviewed-By: Markus Goetz
|
|
|
|
|
|
|
|
|
|
|
| |
Symbian now follows the same convention as windows, where "c:" means the
working directory on the C: drive. As a result, the path "c:/" needs to be
passed to QFileInfoGatherer when getting info for the root directory.
This code already existed but only enabled for windows - now enabled for
symbian as well.
Reviewed-By: Markus Goetz
|
|
|
|
|
|
|
| |
The same workaround used for other autotests, as the symbian build system
doesn't accept \" in DEFINES.
Reviewed-By: Janne Koskinen
|
|
|
|
|
|
|
|
|
|
|
|
| |
The m_filePath should always have / as the separator, while m_nativeFilePath
should always have the native separator.
Almost every place the constructor is used, the path could be one passed
into an API from code outside our control.
So it's easier to do the path conversion in the constructor than to have to
remember to call fromNativeSeparators in every place a QFileSystemEntry is
constructed (especially since unix tests won't reveal any error)
Reviewed-By: joao
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
On Symbian, the Seek() function clamps to the end of file position if a
position beyond the end of file is given. Also the write functions which
take a position silently append at the end of file position if asked to
write to a start position beyond EOF.
To avoid this behaviour, we don't use seek but rather maintain our own
cursor position. When writing, we check if the cursor is beyond EOF and if
so extend the file using SetSize() before writing.
Reviewed-By: joao
|
|
|
|
| |
Reviewed-By: joao
|
|
|
|
|
|
|
|
|
|
|
|
| |
The qprocess autotest launches various sub processes and checks their results,
but this lacked error handling.
Added a QVERIFY when opening files (if fopen returned NULL, the test fails
instead of crashing)
Added a 10s guard when waiting for a subprocess to create a file - previously
this loop did not terminate if the subprocess failed.
Reviewed-By: Markus Goetz
|
|
|
|
|
|
|
|
|
|
|
| |
File engine refactor caused a regression in the TIFF image codec,
as QFile::seek() previously worked like lseek() / fseek() in posix.
But on symbian the native RFile::Seek api clamps to the end of the file
if you attempt to seek beyond there.
This test checks seek behaviour in the appropriate place, the QFile
auto test.
Reviewed-By: joao
|
|
|
|
|
|
|
| |
Because RFs::Entry("") gets the entry for the current directory, we
need to explicitly check for empty strings and treat as a non existant file.
Reviewed-By: joao
|
|
|
|
|
|
|
|
|
| |
File engine refactor caused a regression in QTextDocument, due to a
QFileInfo constructed with an empty string returning true from exists()
Added a test case to QFileInfo to catch this at the correct level.
Reviewed-By: joao
|
|
|
|
|
|
|
|
|
|
|
|
| |
The pointer is lazily initialised by QWidget, but becomes invalid after
the application is destroyed.
In the QApplication autotest, QApplication is created and destroyed many times
in the same thread (while in normal use it is created at startup and destroyed
on exit)
As a result, the pointer which is stored in global data became stale and
caused a crash due to calling functions on a deleted object.
Reviewed-By: Jason Barron
|
|
|
|
|
|
|
|
| |
Files in /sys/bin/ on any drive can be assumed to be executable.
So the executable flag is set in the metadata for this path, even though
RFs::Entry fails due to permissions (so we don't know if the file exists or not)
Reviewed-By: joao
|
|
|
|
|
|
|
|
| |
CFileMan doesn't support . and .. in relative paths.
Therefore, paths are converted to absolute first.
Also, the error was being ignored which has been fixed.
Reviewed-By: joao
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
First, pgm and pbm formats need the same exception given to ppm format
as they are subtypes of the same format. (as text based image formats they
are vulnerable to line end differences)
Second, a workaround for the readFromDevice test case hanging on the symbian
emulator. The test expects to be able to send a whole image to a socket and
have this fit within the send/receive buffers of the TCP stack.
At least for the emulator's winsock backend, this hangs the autotest inside
a blocking write. On the Nokia 5800, the test passed so it's currently skipped
only for the emulator.
Reviewed-By: Jason Barron
|
|\
| |
| |
| | |
scm.dev.troll.no:qt/qt-file-engines-refactor into earth/file-engine-refactor
|
| | |
|
| |
| |
| |
| |
| |
| |
| | |
In 4.7 and before a QDir::exists() always stats and a refresh() call would
flush metadata, restore this behavior again.
Reviewed-by: Prasanth Ullattil
|
| |
| |
| |
| |
| |
| | |
Also fix the qfileenginemodel test to be less fragile
Reviewed-by: Prasanth Ullattil
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Due to the change from default deployment in symbian from being hardcoded
in qmake to being in default_deployment.prf the .pro files needed to be
changed.
Specifically, "DEPLOYMENT = foo" needs to be "DEPLOYMENT += foo" otherwise
the default deployment lines are not added and the test won't install.
Reviewed-By: Miikka Heikkinen
|
| | |
|
| |
| |
| |
| |
| |
| |
| | |
In 4.7 and before a QDir::exists() always stats and a refresh() call would
flush metadata, restore this behavior again.
Reviewed-by: Prasanth Ullattil
|
| |
| |
| |
| |
| |
| | |
Also fix the qfileenginemodel test to be less fragile
Reviewed-by: Prasanth Ullattil
|
| |
| |
| |
| |
| |
| |
| |
| | |
Qt 4.7 behaviour was to return false when directory already exists for mkdir
but true when directory already exists for mkpath.
This change makes symbian do that in master, even though it's inconsistent.
Reviewed-By: Thomas Zander
|
| |
| |
| |
| |
| |
| |
| |
| | |
These tests did not link when QT_BUILD_INTERNAL was not defined, as one
of the private headers was causing QFileInfoGatherer::~QFileInfoGatherer to
be referenced
Reviewed-By: Thomas Zander
|
| |
| |
| |
| | |
Reviewed-By: Thomas Zander
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Testing done:
win32-msvc2008 and Mac OS X - qfile, qfileinfo, qdir, qdiriterator,
qtemporaryfile autotests
symbian-sbsv2 - qtcore, qtgui, qtxml autotests
win32-g++ - compilation test for qmake
Reviewed-By: Thomas Zander
|
| |
| |
| |
| | |
Reviewed-By: Trust Me
|
|\ \
| | |
| | |
| | |
| | |
| | | |
Conflicts:
src/s60installs/bwins/QtGuiu.def
src/s60installs/eabi/QtGuiu.def
|
| |\ \
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
scm.dev.nokia.troll.no:qt/qt-integration into master-integration
* 'qt-master-from-4.7' of scm.dev.nokia.troll.no:qt/qt-integration:
Fix crash when using Q_GLOBAL_STATIC(QWidget...)
Removed the need for S60main.rsc resource file in Symbian.
|
| | |\ \ |
|
| | | |\ \
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
4.7-integration
* '4.7' of scm.dev.nokia.troll.no:qt/qt-s60-public:
Fix crash when using Q_GLOBAL_STATIC(QWidget...)
Removed the need for S60main.rsc resource file in Symbian.
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
If Q_GLOBAL_STATIC is used with a QWidget (or subclass) then the
destructor of QWidget will be executed after the destructor of
QApplication. Since ~QApplication() destroys the S60 environment and
the trap handler, we need to be sure that if QApplication is destroyed,
we do not attempt to use anything from the S60 environment. This
includes RWsSession and the trap handler.
The fix is to avoid flushing the WSERV buffer if QApplication has been
deleted already.
Reviewed-by: axis
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Qt upgrade was failing since the s60main.rsc is being locked by
S60 application framework. And when installer detects that
old version of Qt has already been installed it first tries to
uninstall the old one and then install the new one. The uninstallion
failed since the file was locked by the running Qt application.
It should be noted that this patch fixes the Qt upgradibility only
for Qt versions where patch is included. I.e. the versions before
4.7.2 need a different mechanism to be upgradable. This different
mechanism is based on partial upgrade SIS packages as described in
QT-4052.
Task-number: QT-3471
Reviewed-by: Axis
|