| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
The test machine is configured to pop up dialogs in case of CRT
assertion failures. The CRT has a debug assert whenever an invalid
file handle is used, while the test is expecting to get the EBADF
return code.
Due to some behaviour change in windows 7 (maybe SP1), we need to
call _CrtSetReportMode in addition to setting an error handler.
Reviewed-By: mread
|
|
|
|
|
|
|
|
| |
This reverts commit 2e7e8a2b805b868eade36cc44fa14bf8ca2f8c2f.
This broke the QNetworkconfigurationManager auto tests.
Reviewed-by: Peter Hartmann
|
| |
|
|
|
|
|
|
|
|
|
|
| |
do not use a null configuration, but a default configuration in
QNetworkRequest by default. In addition, setting an SSL configuration
explicitly will cause the on-demand loading of root certs to be
disabled (because it could be that the user has set the CA certificates
explicitly).
Reviewed-by: Markus Goetz
|
|
|
|
|
|
| |
This also fixes the autotest failure on mac with macDefaultLocale test.
Reviewed-by: trustme
|
|
|
|
|
|
|
|
|
|
| |
On some Windows machines the wifi detection in
QNativeWifiEngine::available does not work as WlanOpenHandle still
succeed altough there is no WLAN card. So instead we can detect this
using the WlanEnumInterfaces in requestUpdate() instead.
Task-number: QTBUG-17219
Reviewed-by: Xizhi Zhu
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This change takes the QTBUG-17986 fix, which deletes QThreadData for
adopted threads that have created QEventLoops, and applies it to
qthread_symbian.cpp, which didn't exist at the time of the original
fix.
One complication is that Symbian uses a separate thread to monitor
adopted thread lifetime, as there is no API to intercept thread exit
to have cleanup code run within the context of the thread. However the
cleanup for the thread involes deleting active objects that were
created in the adopted thread, not the monitor thread. If these active
objects are completed but not run, their cancellation could deadlock.
In particular the wake up active object in the event dispatcher is
typically in this state. We deal with it by detecting the situation
and re-completing/cancelling the active object in the adopted thread
monitor thread, which prevents deadlock and allows correct operation of
the monitor thread. It is possible for this problem to affect other
active objects owned by the event dispatcher. They symptom would be
that finished signals from adopted threads are not sent, or they arrive
much later than they should.
Task-number: QTBUG-18622
Reviewed-by: Shane Kearns
|
|
|
|
|
|
|
|
|
|
|
| |
This patch uses icu to do string collation via QString::localeAwareCompare
function and for QString::toLower/toUpper - which is important e.g. for turkish
locales where uppercased 'i' is not a latin 'I'.
Based on the patch by Harald Fernengel
Reviewed-by: Harald Fernengel
Reviewed-by: Zeno Albisser
|
|\
| |
| |
| |
| |
| |
| |
| | |
Conflicts:
configure.exe
src/s60installs/bwins/QtCoreu.def
src/s60installs/bwins/QtGuiu.def
src/s60installs/eabi/QtCoreu.def
|
| |\
| | |
| | |
| | | |
scm.dev.troll.no:qt/qt-symbian-network into symbian-socket-engine
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This reverts commit e5d27e7aeac984e46f3aa8de20160cc00fc63155.
Conflicts:
tests/auto/qabstractnetworkcache/tst_qabstractnetworkcache.cpp
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Change df9491b302f6404ad2ccc6dc2eb3377176d994c6 optimised auto connections
by comparing thread ID rather than comparing TLS addresses.
However it was implemented on a branch that didn't have the native
symbian threads. So merging the two branches caused a regression without
merge conflicts.
Reviewed-by: mread
|
| |\ \
| | |/
| |/|
| | |
| | | |
Conflicts:
tests/auto/qabstractnetworkcache/tst_qabstractnetworkcache.cpp
|
| |\ \
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
symbian-socket-engine
Conflicts:
src/s60installs/bwins/QtCoreu.def
src/s60installs/bwins/QtGuiu.def
src/s60installs/bwins/QtNetworku.def
src/s60installs/eabi/QtCoreu.def
src/s60installs/eabi/QtGuiu.def
src/s60installs/eabi/QtNetworku.def
src/s60installs/eabi/QtOpenVGu.def
tests/auto/qabstractnetworkcache/tst_qabstractnetworkcache.cpp
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
If there is no event loop when a socket notification is enabled,
then invoke the method via a queued connection so that it is run
again when the event loop is started.
This covers sockets created and having an asynchronous API called
before calling QCoreApplication::exec().
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
1. Start the default bearer, when Qt is built with bearer management
2. Allow server sockets to have IPv6 any addresses
3. Broadcasting test - broadcast on each valid broadcast address
4. Enable IPv6 tests on symbian
5. writeDatagram test - workaround symbian having a smaller recieve
buffer for UDP packets than the maximum packet size it can send.
6. Add QEXPECT_FAIL for some multicast tests on symbian, the OS support
for multicast is incomplete.
7. Choke the echo test, as the test server shuts down the echo service
for 10 seconds when receiving packets too fast.
8. For manual testing, added the FORCE_SESSION macro to test behaviour
of UDP sockets when they have an explicit network session associated
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | | |
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | | |
Reviewed-by: Shane Kearns
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
1. start default network at start of test (this makes the test faster
and more reliable, as the network can be stopped automatically if
we don't use a network session)
2. remove unnecessary ifdef from the ipv4 data
3. add new IPv6 test hosts to the ipv6 data
4. extend ipv6 timeout to match the ipv4 timeout
5. check error in LookupThread::run() to avoid test crashing on failure
6. extend timeout for multipleDifferentLookups (uncached DNS is SLOW)
7. enable IPv6 lookups on symbian statically (don't use the broken POSIX
functions)
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
For reverse lookups, don't report errors, instead just return the IP
address as a string for the host name (for behavioural compability
with the windows and unix backends)
Don't sort IP4 addresses before IP6 addresses for the same host. The
symbian host resolver internally sorts the list so the usable addresses
are returned first. (usable means has a valid route)
Task-number: QTBUG-18135
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Test calling the abort from the same thread or from a different thread.
The different thread is to check thread safety of the API.
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Updating comment to indicate that the feature isn't supported on Symbian.
Reviewed-by: Shane Kearns
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Each thread needs at least one current request if it has any queued
requests, this is to stop the queue stalling.
When starting a queued request, start it in the same thread it belongs to
When aborting a request from the wrong thread, just detach it (it will
complete normally but the slot isn't connected, and then delete itself)
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Manger -> Manager
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
The existing threadSafety test only tested calling the blocking API
from multiple threads. As the asynchronous API may be implemented in
a different way, it should be tested separately.
Create 10 threads, each of which queues 10 async name lookups.
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Add additional check when starting a queued request, as the cache may
have been populated while the request was queued.
Put completed requests in the cache (note cache code internally discards
errored results and only caches successful results)
Reviewed-by: Aaron Tunney
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
When an IP6 address contains a valid IP4 address, report the address
in IP4 format (this is needed for SOCKS5 socket engine)
When binding Any or AnyIPv6, create a dual mode socket (KAfUnspec),
but report the local address as being the same as what the user
requested. (by default, symbian returns ::0 for the dual mode socket,
which causes problems for SOCKS and UDP code)
The intent is that most applications written for IP4 can work
transparently in an IP6 environment.
QTcpServer or QUdpSocket can accept either IP4 or IP6 from the same
socket.
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Although symbian's multicast support is incomplete, this is enough to
pass the QUdpSocket test case. (no IGMP/MLD support in the OS)
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Enough debug logging allows time for the proxy server to close the http
connection after sending a 407 response. This errors the connection,
but then the queued _q_startNextRequest is immediately run, causing an
assertion failure.
Changed this to a soft failure which allows the error to propagate
rather than crashing.
Test case is tst_qnetworkreply::ioPostToHttpFromSocket
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
An IPv6 socket can send to IPv4 addresses when in a dual mode stack.
On symbian, autobinded sockets have the ::0 local address rather than
0.0.0.0 - this check was causing the socket to be destroyed and recreated
with every call to writeDatagram.
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
The mutex added needs to be recursive, as stop calls close internally,
which also needs to lock the mutex.
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Allows testing of plugins such as bearer without needing to reinstall
all of Qt. Equivalent to the rules in qbase.pri for the normal DLLs
Reviewed-by: Miikka Heikkinen
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
De-duplicate error handling into a helper function.
RHostResolver can return a few different errors at the end of the list,
so treat these benign ones the same.
When Next() gives an error, return any results we already obtained.
Reviewed-by: Aaron Tunney
|
| | | |
| | | |
| | | |
| | | | |
Reviewed-By: Shane Kearns
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
ReadUserData needed to avoid platsec warnings accessing cert store.
Removed incorrect deployment from the two new on demand tests.
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Skipped because of open C bug, but these tests are passing in N8 with
symbian socket & file engines
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
This avoids the test starting and stopping the WLAN for each test case
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Loosen requirements for generated error pages (these vary with exact
version and locale of the web server). Don't care what the length is,
but if a content-length header is sent then it should match the size
of the error page data. (apache2 does send content-length for errors)
Don't fail the gzip test if the server doesn't send a content-length
header. That's a server bug rather than Qt. (apache2 doesn't send)
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Due to threaded http, the shared QNetworkSession can have its last
reference removed from a http delegate thread. To avoid this deadlocking
use a deleteLater custom deleter so that the QNS is deleted from the
thread it has affinity for.
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
The frox ftp proxy closes connections using TCP RST instead of a normal
TCP FIN. On symbian, this causes the received but unacknowledged
preceding packet to be discarded (the "221 Goodbye" ftp response).
As QNAM and QFTP don't care about this, the network self test won't
either. (only for symbian, as windows and unix TCP/IP stacks retain
the unacknowledged packet in the receive queue)
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Start default bearer when running network self test
Fix typo in qnetworkreply test
Reviewed-by: Markus Goetz
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
A timer of 0ms completes on the next kernel tick (up to 0.999ms)
There is a poll GetOpt we can use instead that doesn't require using
a timer to cancel the select ioctl.
Reviewed-by: Markus Goetz
|
| | | | |
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
The new test added three reference data files required by the test,
but these were not being included in the sis file.
Reviewed-by: mread
|
| |\ \ \
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Conflicts:
src/network/access/qhttpnetworkconnectionchannel.cpp
src/network/socket/qlocalsocket.cpp
src/s60installs/bwins/QtCoreu.def
src/s60installs/bwins/QtGuiu.def
src/s60installs/bwins/QtTestu.def
src/s60installs/eabi/QtCoreu.def
src/s60installs/eabi/QtGuiu.def
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Depending on the OS scheduling, the TCP data sent and flushed may not
have been looped back to trigger the socket's ready to read notification
before the test called processEvents.
This caused a failure on CI system windows 7.
Added a short delay before processEvents to allow the OS to catch up.
Reviewed-By: Markus Goetz
|
| | | | |
| | | | |
| | | | |
| | | | | |
Reviewed-by: Aaron Tunney
|
| | | | |
| | | | |
| | | | |
| | | | | |
Reviewed-by: Peter Hartmann
|