| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
1. The writeStarted flag shall be reset to false after bytesWritten() signal
is emitted, but not before it.
2. In case of start of writing from the startAsyncWriteTimer it is necessary
to call the startAsyncWrite() method but not the completeAsyncWrite() method.
It give guarantees to exclude of race condition to multiple call of the
WriteFile twice until the write operation is not completed.
For example, a race condition is watched in case of writing data from the
slots which are connected to the bytesWritten() signal from a device; or to
the timeout() signal from a timer with zero interval. To checking of these
scenarios are added additional autotests.
Tested on Windows 8 with the on-board and USB serial ports, using Qt5.
Task-number: QTBUG-40769
Change-Id: Iedabbf38847debeee795a10fd7a4c54c65d2a338
Reviewed-by: Patrick Noffke <patrick.noffke@gmail.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
|
|
|
|
| |
Change-Id: Ia60bf0d020d98a6fda8f689d631bb8681950d15f
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The method waitForReadyRead() shall call of the write sequence since the
signals from the startAsyncWriteTimer will not be handled inside of
waitForReadyRead() without event-loop.
Tested on Windows 8 with the on-board and USB serial ports, using Qt5.
Task-number: QTBUG-40793
Change-Id: I7806f16a4df30c4ec1643d8f696ad5761decd30a
Reviewed-by: Patrick Noffke <patrick.noffke@gmail.com>
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
|
|
|
|
|
|
|
|
|
|
| |
The enum is already marked as deprecated leading to this error:
qserialport.h:267:46: error: no type named 'DataErrorPolicy' in 'QSerialPort'
void dataErrorPolicyChanged(QSerialPort::DataErrorPolicy policy);
~~~~~~~~~~~~~^
Change-Id: I8d178f0d0921b8f5308183e1f09a4c8388448e26
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Previous commit 76f293bfb0fd7bab391f96ac822eacc40e4c5176 with adding of
this test to the benchmarks were hasty.
After all it makes sense to place it to auto tests to have opportunity to
run with others together.
Besides, this test is renamed since it is not a loopback, it is an I/O
test with the synchronous approach.
Change-Id: I043a6e0075561167bb29a59a384554ef98dbbd4a
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
According to documentation, the flush() method shall be non-blocking.
Tested on Linux 64-bit with auto-tests with the on-board and the USB
serial ports for Qt4 and then Qt5.
Note: On Windows the flush() method still does not work.
Change-Id: Iaee80361e59e0c281206ca24c817a446cdbf6ed1
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Using of the QString::fromWCharArray() we didn't pass length of a string,
hoping that all WCHAR strings are terminated by the null-character.
But it can lead to the wrong results if the initial WCHAR string has no
null character. Therefore it is reasonable to pass length of a string to
exclude an error.
Besides, to do not break of the algorithm of string comparison it is
necessary to return not null-terminated strings. Also now there is no need
to initialize of some allocated arrays by zero values.
Tested on Windows 7/8 with the on-board, the virtual com0com, the USB FTDI
and the PL2303 serial ports, using Qt4 and then Qt5.
Change-Id: I382cf8eacf4ab4d21c54de17fcdd6b9fcfa3d02c
Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
|
|
|
|
|
|
|
|
| |
In case of unsuccessful initialization of a device at the opening, a
valid descriptor has to be closed before return from the open() method.
Task-number: QTBUG-40414
Change-Id: I45568f176e003d9be1fe8c3017da29f39908efb0
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In case of communication event is triggered in the
waitForBytesWritten() method then necessary to process it with
the _q_completeAsyncCommunication() method, but not with the
_q_completeAsyncRead() method.
Otherwise it leads to false filling of the readBuffer with the
data that were read earlier (or with garbage) which remained in
readChunkBuffer.
The synchronous loopback autotest reproduces a bug without of
this patch.
Tested on Windows 7/8 with the on-board and USB serial ports,
using Qt5.
Task-number: QTBUG-40344
Change-Id: I679109b60f4058c4c6f7e5f02c1f70ba6039d8d8
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Good reason to check of data correctness at transferring and receiving.
In this test are used two serial ports, connected in a null-modem mode:
http://en.wikipedia.org/wiki/Null_modem .
The sender port transfers data to the receiver port. After data are
received is carried out check of equivalence of the transferred and
received data.
Before run of testing it is necessary to set two variable environments
QTEST_SERIALPORT_SENDER and QTEST_SERIALPORT_RECEIVER to specify names
of used serial ports.
This test can reveal errors related with the internal data processing
and also errors of synchronous I/O inside of QSerialPort.
This test is placed in separate "benchmarks" category not to mix-up
with "manual" and "auto" categories, because this test is closer to
"benchmarks" by own functionality.
Tested on Windows 7/8 with the virtual com0com serial ports, using
Qt4 and then Qt5.
Change-Id: Ie6f87b50784bce211cf68c16cf75f79d12a8564f
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
-added information about how to run the example by including a file
from qtbase/doc/global
-edited summary sections
-added \brief.
-removed image captions. Our Qt 5 CSS files don't handle them.
Task-number: QTBUG-33597
Change-Id: If3293aa9d98b662f20af2af7030b0ab0fb1e8401
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
Reviewed-by: Leena Miettinen <riitta-leena.miettinen@digia.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The function devicePortName() search in the Registry the names of ports
by names of registry keys. At first looks for value of a key of "PortName",
and then of "PortNumber" key.
Thus, the first found value shall stop search and do not try to continue.
Tested on Windows 7/8 using Qt4 and then Qt5.
Change-Id: I98b00ff043a3b08476fec0a57b0d36ce65fc8d63
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
It is reasonable to wrap implementation of getting properties of port
from availablePorts() to the separate functions. It is simplifies a
code and improves its understanding.
Tested on Windows 7/8 with the on-board, virtual com0com, USB FTDI
serial ports using Qt4 and then Qt5.
Change-Id: I2345fa49b3633960412715b299e068daa8fdcfd0
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
|
|
|
|
|
|
|
|
|
| |
This patch improves workaround implemented in the previous commit
d34fce4a5d12789ded107631e22cb6ef54d35eee
Change-Id: I013b19b47ee343ab00f242d7ee1021b92c07d18f
Reviewed-by: Peter Kümmel <syntheticpp@gmx.net>
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The current implementation of the qserialportinfo_mac.cpp module is a
little confused and complicated.
It is reasonable to re-write it using some ready wrappers of CFTypeRef,
CFStringRef and others objects from the private qcore_mac_p.h file.
Thus, it will allow to bypass of manual releasing of the allocated
resources (using the RAII idiom where it is possible), and also to
simplify a code.
Tested on OS X 10.8.4 with the on-board, USB PL2303, USB FTDI,
USB Android serial ports and with the USB ZTE Modem using Qt4.
Change-Id: Ibbac03a17fdd41bbaf530f863e7c690a15708bd2
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Building in OS X uses some shared code from the serialportinfo_unix.cpp
module with the OS X specific code from the serialportinfo_mac.cpp module.
Thus, all code of the serialportinfo_unix.cpp module, which not related
with the OS X, is shielded by a macro. It adds an excessive garbage for
readability in this module and also some confusion in the *.pri file.
It makes sense to make implementation of the serialportinfo_mac.cpp
module completely independent, that will simplify maintaining of source
code.
Besides are added tests which can reveal declared but not implemented
methods in building, and also to check a correctness of default values
at running.
Tested build on OS X 10.8.4 with Qt4, an then on Android with Qt5.
Change-Id: I67935b64e2b623fb8d4c14d59e1b87f1eac71c3e
Reviewed-by: Bernard Pratz <guyzmo+qt@m0g.net>
Reviewed-by: Peter Kümmel <syntheticpp@gmx.net>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The slots _q_completeXX() returns empty values, therefore the
waitForXX() methods continue to work even if occurs an error
from the I/O completion.
Is reasonable allow to return of boolean values for the I/O
completion slots.
Besides, it allows avoiding indirect detection of an error,
comparing of the sizes of I/O buffers before and after the
operation completion. Though, for the waitForReadyRead(),
for this purpose it is necessary to add additional
modifications in a following patches.
Tested on Windows 7/8 with the virtual com0com ports using
Qt4 and then Qt5. Testing was made using of autotests
and examples.
Change-Id: I95a76461af4595f6658f0cad766a4fff14eb7afc
Reviewed-by: Peter Kümmel <syntheticpp@gmx.net>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Each q_completeXXX() slot does call of GetOverlappedResult()
and differently interprets an error depending on the type of
I/O operation (read, write or communication).
It is more reasonable to make it in the separate method
handleOverlappedResult() which returns the number of the
transferred bytes or -1 in case of error.
Besides, this method accepts the additional integer parameter
which expresses type of overlapped operation using the
existing QSerialPort::Direction enumeration. The type of the
communication operation has not separate value, so is used
zero value for it.
Tested build on Windows 7/8 using Qt4 and then Qt5.
Change-Id: I8482392d06279bc430656e50b1e4392c513885c6
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
Reviewed-by: Peter Kümmel <syntheticpp@gmx.net>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
|
|
|
|
|
|
|
|
|
| |
When in readNotification() or completeAsyncWrite() an error
is emitted and the loop is not exited, this error signal
will flood the eventloop.
Change-Id: I77b8a4c337041258862b19a07917bef059c11cc1
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The standard USB CDC driver (usbser.sys) does not add information of
serial number to the device instance identifier of the child devices
(serial ports), so our algorithm returns an empty string.
But information about a serial number is stored in the device identifier
of the parental node. Thus, it is necessary to get this parent
identifier, using these recommendations from MSDN:
http://msdn.microsoft.com/en-us/library/windows/hardware/ff549417%28v=vs.85%29.aspx
For this purpose is used the set of new functions from the
"Configuration Manager Functions" family.
Therefore it was necessary:
* Rewrite the deviceInstanceIdentifier() function.
Now this function accepts one DEVINST parameter and returns the
corresponding device instance identifier string.
* Add the new parentDeviceInstanceNumber() function.
This function accepts a DEVINST of the child device and returns a
DEVINST of the parent device.
Thus, the new algorithm tries to extract a serial number from the
current (child) device and in case it is empty, try to extract from
the parent.
Tested on Windows 7/8 with the Segger JLink device (provides the CDC
serial port) using Qt4 (MSVC2010) and then Qt5 (MSVC2012, MinGW).
Also are checked other devices for possible regressions:
* on-board serial port
* com0com virtual serial port
* USB FTDI serial port
no regressions found.
Change-Id: Ic6afc258f45aa374ec77dc0b586d479b5286359f
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
To the ResourceError handling we generally relied on triggering of the
_q_completeAsyncCommunication(), in case of ejection of an USB device
from a host. But it worked with not all USB devices (for example it did
not work with the ZTE modems). Even when this worked (for example with
the PL2303 devices), then we got the non informative "Unknown error" of
the ResourceError string representation.
It is more reasonable not to do any custom checks of an event mask when
triggered of the _q_completeAsyncCommunication() method. We need allow
to detect an errors with means of the
ReadFile/GetOverlappedResult/WaitCommEvent syscalls on an invalid device
handle.
An important note is that returns the different error codes for a
different types of devices.
For example:
* For the USB ZTE and the Samsung Android USB modem returns the
ERROR_ACCESS_DENIED error code.
* For the FTDI and the Samsung Android Diagnostic Serial Port returns
the ERROR_OPERATION_ABORTED error code.
* For the USB PL2303 device returns the ERROR_BAD_COMMAND error code.
Thus, there is no universal solution of this issue. Each vendor of the
driver implements it own way. Therefore for each new type of the device
it is necessary to add processing its error code, if we have no it yet.
Tested on Windows 7/8 with the USB ZTE and the Android modems, with the
USB PL2303 and the FTDI devices and with the Android diagnostic serial
port using Qt4 and then Qt5, no regressions found.
Task-number: QTBUG-32020
Change-Id: I61c02078ee98221ecef18a27cde3d9bae3674d70
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* It is necessary to ignore the NoError error code.
* In case of error from GetOverlappedResult() it is necessary to stop
the I/O operation by a immediate return from the function, because I/O
processing does not make sense. This is consistent behavior similar to
starting I/O operations.
* All errors from the read completion should be interpreted as ReadError.
Exception is only the critical ResourceError error which it is necessary
to leave without changes. This is also consistent behavior similar to
starting I/O operations.
* The write completion error handling should be similar to the read
completion handling.
* For the communication completion we have no specific error codes.
Therefore error handling shall be without modification of their error
code.
Also were made small cosmetic fixes to the related start I/O methods.
Tested on Windows 7/8 with the on-board, the PL2303 and the virtual
com0com serial ports using Qt4 and then Qt5. Testing are made by means
of examples of library and also by means of auto tests, no regressions
found.
Change-Id: I0d08b53627431c42ab5147d4330f1aaf819b4d63
Reviewed-by: Peter Kümmel <syntheticpp@gmx.net>
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
|
|
|
| |
Change-Id: I50ea328c903f3ba0c78f9570b6b6b892bb38f976
|
|
|
|
|
|
|
|
| |
Caused by I1a66164c.
Change-Id: I78973a509ff91a1302f435b1a23ecce432e56dd4
Reviewed-by: Laszlo Papp <lpapp@kde.org>
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The exception notifier does not do any useful operation while processing
I/O but only complicates error handling. It has been thoughtlessly
introduced at beginning in hope that in the future it can it is useful,
but the practice showed that it not so.
Problem is that in case of ejection of an open USB device from a host
there is the infinite triggering of the exception notifier that leads to
program hangup. We could stop an exception notifier in this case, but it
not rationally because after the exception notifier is triggered the read
notifier which also enters the infinite loop.
Thus the read notifier anyway duplicates the exception notifier, as for
functionality and as for an error code. Therefore more rationally to
completely remove the exception notifier. So in case of the ResourceError
we need just to stop of the read notifier, because it is a critical
error. Then an user will be notified on the ResourceError appearance and
must to close the device himself.
Though work was checked only on Linux, but we can do not worry, and to
refuse from this notifier in MacOSX also. Because anyway in MacOSX this
type of notifier is not supported, at least from the Qt 5.1 and above:
https://qt.gitorious.org/qt/qtbase/source/982fdd4ef556f63ba77384a0dbe97928e610091b:src/platformsupport/cfsocketnotifier/qcfsocketnotifier.cpp#L210
Besides in documentation on QSocketNotifier is recommended do not
use the Exception mode:
http://qt-project.org/doc/qt-5/qsocketnotifier.html#Type-enum
Tested on ArchLinux 64 bit with the USB PL2303 device, ZTE modem,
Android serial port using Qt4 and then Qt5.
Task-number: QTBUG-36727
Task-number: QTBUG-35829
Change-Id: I4ea6db2a1c7f10c096d57817c00edefc4f0595ff
Reviewed-by: Rafael Roquetto <rafael.roquetto@kdab.com>
Reviewed-by: Peter Kümmel <syntheticpp@gmx.net>
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
|
|\
| |
| |
| |
| |
| |
| |
| | |
Manually adjusted:
src/serialport/qserialport_unix.cpp
to remove isCustomBaudRateSupported
Change-Id: I22480b57d8c1b91ea7ea4a1da4ba3060e1a5f93f
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This regression was caused by 7c737e0edcb93585856c65890ef34e5c5a28ee6b.
The initialization of the (current) serial port info was removed in that commit.
Tested on Linux x86_64 with Qt 5.1.1 and 4.8.6.
Task-number: QTBUG-38961
Change-Id: Idd3c5dc7eb37093848f33387563ff14567ac54fe
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Reviewed-by: Rafael Roquetto <rafael.roquetto@kdab.com>
|
|\ \
| | |
| | |
| | | |
Change-Id: I65f7cd2fdeced526c8bd7dc5bc4081e7687e7310
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Implementation of the availablePortsByUdev() function is a little
complicated by excess "if/else" conditions and also too long lines,
that worsens readability.
It is reasonable to make the following:
* To get rid of the big "if {...}" blocks in favor to immediate return
from function in case of an error. It will allow to reduce a quantity
of lines and will shift code alignment to the left.
* To split declaration of some long variables and functions into separate
lines, with length at least up to 80~100 characters.
* To drop of the 'struct' keywords and the '::' global namespace
operator for variables.
Tested on ArchLinux 64 bit with the on-board and PL2303 serial ports
using Qt4 and then Qt5.
Tested build on Android x86 using Qt5.
Change-Id: Iddc2a9511230e56e4a9d01a4c22af7b2eaeae60c
Reviewed-by: Peter Kümmel <syntheticpp@gmx.net>
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
It is a good idea to get rid from the manually releasing of the
allocated udev resources.
Tested on ArchLinux 64 bit with on-board and PL2303 serial ports using
Qt4 and then Qt5.
Change-Id: Ib25e100bea37ad3cc0bee015b7de6d4de762f2cc
Reviewed-by: Peter Kümmel <syntheticpp@gmx.net>
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
There is no sense to use udev_device_get_subsystem() for comparing with
the predetermined names of subsystems (like "pnp", "usb" and so on).
This idea was preliminary and required further revising after
accumulation of some statistic.
The main ideas in favor of refactoring are given below:
* It seems that reasonable to use the "ID_MODEL", "ID_VENDOR" and others
identifiers to query of properties for all types of serial ports. We
lose nothing, because in the worst case the getUdevPropertyValue()
function will simply return an empty string, as before.
* Earlier we excluded from the list all ports which have a subsystem
with the "platform" name. Practice showed that for these devices is used
the serial8250 driver. Therefore it is reasonable to exclude this type
of devices by the name of drivers instead of by the name of subsystems.
Also are made related changes:
* Is added the resolving of chars for the new udev_device_get_driver()
function
* The functions of getting properties are wrapped for reduction of
length up to 100 characters.
Tested on ArchLinux 64-bit with the on-board, PL2303 serial devices
using Qt4 and then Qt5.
Tested build on Android x86 using Qt5.
Change-Id: I1a66164ca1893180e1ed97524cff98b4a933a63b
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
|/ /
| |
| |
| |
| |
| | |
Change-Id: Ia5bcc1230af988f06923447f8cb1de5a8d585518
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The commit 2360c401ae2012ed1b5a2b470a088cbbdb0d7f27 introduced an
regression into waitForReadyRead(). Before this commit each read
transaction was started again until the zero number of bytes will
returns from FIFO. When FIFO is empty, all data has been read, and
only then the waitForReadyRead() return true. I.e. condition
"qint64(readBuffer.size()) != currentReadBufferSize" does not mean
that reading is finished.
But after that commit this condition is incorrect, because now each
read operation started once and can not return zero bytes (in case
the number of read bytes less than ReadChunkSize). Thus it led to
returning of TimeoutError error.
Now, this issue has been fixed. Also is added the set of auto-tests
to testing of the waitForReadyRead() method.
Tested on Windows 7/8 using Qt5 and then Qt4.
Task-number: QTBUG-39314
Change-Id: I8abbf986c2a1cc77af634ddbc1747fb46f416a39
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
* The ASYNC_SPD_CUST flag shall be cleared only if it was earlier
set; otherwise it is not necessary to touch of the serial struct.
* Seems, that is more reasonable to return an error code when
writing back of the serial struct is failed. In this case it is
an error, because we know that serial struct is supported, since
we got it successfully earlier.
Tested build on Archlinux 64 bit using Qt4 and then Qt5.
Tested build on Android x86 using Qt5.
Change-Id: I54ca4772dc770dabac5e3555e6cb42ecc23255a4
Reviewed-by: Rafael Roquetto <rafael.roquetto@kdab.com>
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
At timeouts from the waitForX() methods are set a wrong
error string. Thing in that the WAIT_TIMEOUT constant is
not a system error, therefore call of GetLastError()
returns a code of a system error which was set by someone
earlier.
Thus, it is necessary to extract text representation of
an error directly from the WAIT_TIMEOUT constant without
relying on GetLastError().
Tested on Windows 7/8 using Qt4 and then Qt5.
Change-Id: Id7fdc71c1b64c39eed658148bb1b66ac8806f119
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
Reviewed-by: Dyami Caliri <dyami@dragonframe.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
QSerialPort::~QSerialPort() calls close() which emits an error if the
port is already closed. This error signal may then call slots of
already (partially) deleted objects.
Task-number: QTBUG-39369
Change-Id: I28bf94564ae120619054c6920779da974dccdad6
Reviewed-by: Laszlo Papp <lpapp@kde.org>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Handling of the ResourceError is moved directly to
_q_completeAsyncCommunication() method.
Reasons:
* The processIoErrors() should handle only the
EV_ERR event.
* The handling of ResourceError in the processIoErrors()
it is the old heritage when the event processing logic
was into the notifiers.
Also the method processIoErrors() is renamed in favor to
handleLineStatusErrors() for readability, according to
documentation on EV_ERR mask:
http://msdn.microsoft.com/en-us/library/windows/desktop/aa363479%28v=vs.85%29.aspx
Tested build on Windows 7/8 using Qt4 and then Qt5.
Change-Id: I6dc331933155f1e182b21f40885e47d574c4ed9f
Reviewed-by: Dyami Caliri <dyami@dragonframe.com>
Reviewed-by: Rafael Roquetto <rafael.roquetto@kdab.com>
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| | |
Return QSerialPort::SerialPortError instead of bool
Change-Id: I79d731c1a0178ef0238495d7e9bb3e1b866e82c0
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| | |
There are two source files for this module only. Adding a header
precompilation is more expensive than the benefit.
Change-Id: I27f0756693ec655f56a781f94d14840ce48f71b4
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
|
| |
| |
| |
| |
| |
| | |
Change-Id: I4a6767196d205cf1f0715edfb9ea117a2d7fa5c9
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
1. The QCOMPARE macro can not make type cast between "int" and
"qint64", though on qt5 there are no problems.
2. The QSKIP macro should accept two input parameters.
Tested build on Lixux 64bit and Windows with Qt4 and then Qt5.
Change-Id: I1f8497201e8d704098faea789bc61829f0d2f9e4
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
| |
| |
| |
| |
| |
| |
| | |
Tested on Linux x86_64 with Qt 5.1.1
Change-Id: Ibea1204ece921c2c18ee37992380fbc33897be8d
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
QSP::open should read values of DTR/RTS after a successful open,
not try to write them. The user should not expect to set DTR/RTS
before opening the port (unlike the other settings).
Tested on Mac OS X 10.8.5.
Task-number: QTBUG-38640
Change-Id: Iffea464f412c6972aa26a408ba01e304fa813c76
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
QSPI was relying on QLockFile.isLocked() to determine if the port
was in use. This was due to a misunderstanding of the QLockFile
class design.
Ideally the code in QSPI.isBusy() should be moved into a new method,
QLockFile.available(), since it contains implementation details of
that class.
Note that isBusy() is for information purposes only, since checking
for a lock without obtaining it is always racy.
Change-Id: I675bd4033e4228e6bf864bf2e7a00a2520ccc50e
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
Reviewed-by: David Faure <david.faure@kdab.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
In qt4 the macro QCOMPARE can not make type cast between "int" and
"qint64", though on qt5 there are no problems.
Tested build on Lixux 64bit with Qt4 and then Qt5.
Change-Id: I258f8ee4d26b65cc81720d66714ef69df23e7266
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
WinCE implementation was mixed a common code with the Win32
implementation that caused some problems with maintenance.
More correct decision is moving a common code which was used
in *win.cpp into *wince.cpp module without modifications.
Now the Win32 and the WinCE implementation are independent
from each other.
Tested build with Qt5 on "win32-msvc2012" and then with Qt4
on "wincewm50pocket-msvc2008" configurations.
Change-Id: I63f687468beffa9a75b534a4fbe536b854b12210
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The slot _q_completeAsyncWrite(), which is used inside of
waitForBytesWritten(), returns void. To define the status
of write completion method was used the last error code,
where any value are not equal to NoError was interpreted
as an error.
But this behavior was incorrect because the error code
number could be set by any other previous I/O operation
and remained not cleared. Therefore waitForBytesWritten()
always returns false.
A good solution it is to compare size of the writeBuffer
after completion of each write operation with zero, because
the buffer should be empty.
Tested on Windows 8 with the on-board and the USB serial
ports and the cwritersync example using Qt5.
Task-number: QTBUG-36758
Change-Id: I2cf6f9e08056d2e237211b19cff59990aac53bc1
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The Win32 API has not possibility to do same behavior as on *nix to defer
a write operation to next event-loop. In *nix implementation enabling of
the write SocketNotifier guarantees that write operation will begin on the
following cycle of event-loop in case the TxFIFO of the driver is empty.
It give an feature of deferred data transfer where is possible to use chain
of the QSerialPort::write() methods which just accumulate data in
writeBuffer. Also it gives the chance to simple implement of the
waitForBytesWrite() method.
But in Windows implementation the first call of the QSerialPort::write()
leads to startup of data transfer. And the subsequent calls of this
method simply accumulate data while the previous operation won't completed.
It leads to a payload data are transferred by chunks and to increase of
loading of the write notifier. Also it complicates implementation of the
waitForBytesWritten() method which works not in compliance with
documentation.
The decision is use of additional QTimer with a zero interval which
launches data transfer on the following event-loop cycle.
Tested on Windows 8 using the virtual com0com serial ports and set
of the cwriter(a)sync examples with use Qt5.
Change-Id: Iaea404198e440f6cb191f168561c977d18e9f4ca
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Tested on Windows 8 using the on-board and the USB serial
ports and using Qt4 and Qt5.
Change-Id: Idd25b81c9b6b2cd0b94e83fe30e6e7dad4dddfd6
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
Reviewed-by: Denis Shienkov <denis.shienkov@gmail.com>
|
| |
| |
| |
| |
| | |
Change-Id: Iec4124886514c5bdeae1cf00e3b2e9432a2c75ee
Reviewed-by: Sergey Belyashov <Sergey.Belyashov@gmail.com>
|