Issue with read_input_registers in version 2.1.0 #353
Comments
|
@mpf82 Thanks for the report, I tried the latest (v2.1.0) from pypi , as well as on dev and master branches. I was not able to reproduce this issue (against simulator). I was able to read the maximum permissible register count (125) with out any issues. From the logs it seems that the response received is partial. If you could add the below log format and paste the logs again (both with 2.1.0 and 2.0.1) we can compare the timings and see if something is going wrong there. |
|
Thanks for the fast reply. 2.0.12.1.0And for what it's worth, this is how I am switching between the pymodbus versions:
|
dhoomakethu
added a commit
that referenced
this issue
Oct 16, 2018
|
@mpf82 Thanks for the quick response, Could you please execute your test against this branch ? I have just added the debug logs nothing else, just to see whats happening. You could also edit pymodbus/client/sync.py and add |
|
I changed Thus I changed def _recv(self, size):
if not self.socket:
raise ConnectionException(self.__str__())
if size is None:
size = self._wait_for_data()
print("size", size) # added print
result = self.socket.read(size)
return resultand this seems to be executed twice: So if I do the math, 39+2 is 41 - and that equals the expected 41 bytes, right?!? But the module complains that only 15 bytes have been received. |
|
My bad, thats correct it's not called for register reads, sorry for that. The socket read is for some reason not reading the full data , If you print Are you running these tests in separate virtualenv and is the |
|
Yes, the tests are in a separate venv.
which would match the output of pip:
(Yes, it really print "pymodbus-2.1.0" when installing 2.0.1, but I can tell by version.py and the behaviour that the version really changed.)
And yes, you are right, even though 39 bytes should be read, result has fewer bytes: def _recv(self, size):
if not self.socket:
raise ConnectionException(self.__str__())
if size is None:
size = self._wait_for_data()
result = self.socket.read(size)
print("size", size, "result", result)
return result
|
|
As a quick fix, this works for me: def _recv(self, size):
""" Reads data from the underlying descriptor
:param size: The number of bytes to read
:return: The bytes read
"""
if not self.socket:
raise ConnectionException(self.__str__())
if size is None:
size = self._wait_for_data()
result = self.socket.read(size)
while len(result)<size: # keep reading
remaining_size = size-len(result)
result += self.socket.read(remaining_size)
return resultBut it does not explain why it's working in 2.0.1 and is broken (at least for me) in 2.1.0 |
|
Could you please share the logs with timing info with this patch? Ideally , the response should be ready within in the read timeout. If possible and time permits, could you please also run the tests with out this fix but with bigger timeout (say 6 sec) ?. Thanks. |
|
Increasing the timeout to e.g. 10 or even more seems to have no effect at all. I also tested with a timeout of 1000, but it really makes no difference and I always get the afore mentioned error when I run the sample program w/o my fix. The response of the Maybe it's worth noting that I'm connected to the device using an USB adapter, similar to this one. But still, everything works out-of-the-box with version 2.0.1 I have added debugging information to my quick fix: def _recv(self, size):
if not self.socket:
raise ConnectionException(self.__str__())
if size is None:
size = self._wait_for_data()
_logger.debug("Before first read. Expected size: {}".format( size ))
result = self.socket.read(size)
_logger.debug("After first read. Actual result size: {}".format( len(result) ))
while len(result)<size:
remaining_size = size-len(result)
_logger.debug("Keep reading: {}".format(remaining_size))
result += self.socket.read(remaining_size)
_logger.debug("New result size: {}".format( len(result) ))
return resultHere's the log: I've run the sample like 20 times, and the first time the function attempts to read the 39 bytes, the actual size is something between 12 and 15 bytes, most of the time 13. After entering the while loop, the other bytes are read in one go. So far I didn't encounter a case where the loop was executed more then once - at least not when reading 18 registers. If I attempt to read all 125 register at once, the log looks like this (timeout set to 1000): |
|
Thanks for the logs, interesting to know that socket.read() is not waiting for the complete data till read timeout. I have tried with almost all the modbus devices we have here (on linux) and the issue is not reproducible. The only major diff between 2.0.1 and 2.1.0 is introduction of |
|
That line seems to be the culprit. If I comment it out, the whole data gets read in one go: Edit: Here's the value |
|
So my guess was correct, I went through the serial implementation for windows and found this line in The windows documentation for COMTIMEOUT has this description for
So it looks like the data is being received slower than expected (which basically does not adhere with modbus specs, refer image below ). |
|
Like I said, I'm connected via an USB adapter, maybe that's slowing down the communication? And a timeout of |
|
Hmm, that 859.375 micro seconds is still ok as per modbus spec per-say but yes, using an external USB-RS485 can impose its own limitations. How about providing an option to disable this timeout during client creation ? By default it is enabled but for cases like yours , one could choose to turnoff it . |
|
I tested around, and the "sweet spot" in my case seems to be where the communication would sometimes fail and sometimes succeed. As in But yeah, an option to disable it completely would be nice. Or maybe an option to override the default? Edit: Ok, so as a work-around, I could create the client like so: and maybe you could add That way, one could disable it by setting it to if self.method == "rtu":
if self.baudrate > 19200:
self.silent_interval = 1.75 / 1000 # ms
else:
self._t0 = float((1 + 8 + 2)) / self.baudrate
self.inter_char_timeout = kwargs.get('inter_char_timeout', 1.5 * self._t0)
self.silent_interval = 3.5 * self._t0
self.silent_interval = round(self.silent_interval, 6) |
|
I am planning to introduce another kwarg |
dhoomakethu
added a commit
that referenced
this issue
Oct 18, 2018
dhoomakethu
added a commit
that referenced
this issue
Oct 18, 2018
|
Fix Now part of branch pymodbus-2.2.0, please give it a try and feedback |
|
Sorry for the late reply. Yes, I can confirm that with version |
dhoomakethu
added a commit
that referenced
this issue
Apr 18, 2019
* Rebase to dev3.7 * Adding 3.7 to travis configuration * Updated documentation to resolve warnings introduced with the longer names Updated requirements-docs.txt to include missing modules * Fixed reference to deprecated asynchronous * Adding gmp disable to fix pypy build issues * Adding gmp disable to fix pypy build issues * Removing travis python 3.7 configuration Commenting out python3.7 from Travis while waiting for support. You can run teh 3.7 tests with tox without issues * Adding asserts for Payload Endianness * Fixing example of Payload. Same Endianness for builder and decoder. * Fix Sql db slave context validate and get methods - #139 * #353 - debugging, Add debug logs to check size of avaialble data in read buffer * #353 Provide an option to disable inter char timeouts * #353 Bump version, update changelog * check self.socket (#354) * check self.socket self.socket might be None at this point * Update pymodbus/client/sync.py Co-Authored-By: mpf82 <[email protected]> * Fix typo (#378) * Pymodbus 2.2.0 (#375) * #357 Support registration of custom requests * #368 Fixes write to broadcast address When writing to broadcast address (unit_id=0) there should be no response according to the Modbus spec. This fix changes expected_response_length to 0 when writing to unit_id=0. This will break any existing code that is improperly using unit_id 0 for a slave address. * Bump version to 2.2.0 Fix #366 Update failures in sql context Update Changelog Fix major minor version in example codes * Fix #371 pymodbus repl on python3 * 1. Fix tornado async serial client `TypeError` while processing incoming packet. 2. Fix asyncio examples. 3. Minor update in factory.py, now server logs prints received request instead of only function cod * [fix v3] poprawa sprawdzania timeout * Release candidate for pymodbus 2.2.0 * Fix #377 when invalid port is supplied and minor updates in logging * #368 adds broadcast support for sync client and server Adds broadcast_enable parameter to client and server, default value is False. When true it will treat unit_id 0 as broadcast and execute requests on all server slave contexts and not send a response and on the client side will send the request and not try to receive a response. * #368 Fixes minor bug in broadcast support code * Fixed erronous CRC handling If the CRC recieved is not correct in my case my slave got caught in a deadlock, not taking any new requests. This addition fixed that. * Update Changelog * Fix test coverage * Fix #387 Transactions failing on 2.2.0rc2. * Task Cancellation and CRC Errors Alternate solution for #356 and #360. Changes the RTU to make the transaction ID as the unit ID instead of an ever incrementing number. Previously this transaction ID was always 0 on the receiving end but was the unique transaction ID on sending. As such the FIFO buffer made the most sense. By tying it to the unit ID, we can recover from failure modes such as: - - Asyncio task cancellations (eg. timeouts) #360 - Skipped responses from slaves. (hangs on master #360) - CRC Errors #356 - Busy response * Cherry pick commit from PR #367 , Update changelog , bump version to 2.2.0rc4 * #389 Support passing all serial port parameters to asynchronous server * Fix BinaryPayloadDecoder and Builder wrt to coils * Misc updates, bump version to 2.2.0 * ReportSlaveIdResponse now tries to get slave id based on server identity for pymodbus servers * Update missing bcrypt requirement for testing
dhoomakethu
added a commit
that referenced
this issue
Apr 19, 2019
* Rebase to dev3.7 * Adding 3.7 to travis configuration * Updated documentation to resolve warnings introduced with the longer names Updated requirements-docs.txt to include missing modules * Fixed reference to deprecated asynchronous * Adding gmp disable to fix pypy build issues * Adding gmp disable to fix pypy build issues * Removing travis python 3.7 configuration Commenting out python3.7 from Travis while waiting for support. You can run teh 3.7 tests with tox without issues * Adding asserts for Payload Endianness * Fixing example of Payload. Same Endianness for builder and decoder. * Fix Sql db slave context validate and get methods - #139 * #353 - debugging, Add debug logs to check size of avaialble data in read buffer * #353 Provide an option to disable inter char timeouts * #353 Bump version, update changelog * check self.socket (#354) * check self.socket self.socket might be None at this point * Update pymodbus/client/sync.py Co-Authored-By: mpf82 <[email protected]> * Fix typo (#378) * Pymodbus 2.2.0 (#375) * #357 Support registration of custom requests * #368 Fixes write to broadcast address When writing to broadcast address (unit_id=0) there should be no response according to the Modbus spec. This fix changes expected_response_length to 0 when writing to unit_id=0. This will break any existing code that is improperly using unit_id 0 for a slave address. * Bump version to 2.2.0 Fix #366 Update failures in sql context Update Changelog Fix major minor version in example codes * Fix #371 pymodbus repl on python3 * 1. Fix tornado async serial client `TypeError` while processing incoming packet. 2. Fix asyncio examples. 3. Minor update in factory.py, now server logs prints received request instead of only function cod * [fix v3] poprawa sprawdzania timeout * Release candidate for pymodbus 2.2.0 * Fix #377 when invalid port is supplied and minor updates in logging * #368 adds broadcast support for sync client and server Adds broadcast_enable parameter to client and server, default value is False. When true it will treat unit_id 0 as broadcast and execute requests on all server slave contexts and not send a response and on the client side will send the request and not try to receive a response. * #368 Fixes minor bug in broadcast support code * Fixed erronous CRC handling If the CRC recieved is not correct in my case my slave got caught in a deadlock, not taking any new requests. This addition fixed that. * Update Changelog * Fix test coverage * Fix #387 Transactions failing on 2.2.0rc2. * Task Cancellation and CRC Errors Alternate solution for #356 and #360. Changes the RTU to make the transaction ID as the unit ID instead of an ever incrementing number. Previously this transaction ID was always 0 on the receiving end but was the unique transaction ID on sending. As such the FIFO buffer made the most sense. By tying it to the unit ID, we can recover from failure modes such as: - - Asyncio task cancellations (eg. timeouts) #360 - Skipped responses from slaves. (hangs on master #360) - CRC Errors #356 - Busy response * Cherry pick commit from PR #367 , Update changelog , bump version to 2.2.0rc4 * #389 Support passing all serial port parameters to asynchronous server * Fix BinaryPayloadDecoder and Builder wrt to coils * Misc updates, bump version to 2.2.0 * ReportSlaveIdResponse now tries to get slave id based on server identity for pymodbus servers * Update missing bcrypt requirement for testing * Fix docs (#407) * Fix document generation * Formatting updates in Changelog
dhoomakethu
added a commit
that referenced
this issue
Oct 29, 2019
* Rebase to dev3.7 * Adding 3.7 to travis configuration * Updated documentation to resolve warnings introduced with the longer names Updated requirements-docs.txt to include missing modules * Fixed reference to deprecated asynchronous * Adding gmp disable to fix pypy build issues * Adding gmp disable to fix pypy build issues * Removing travis python 3.7 configuration Commenting out python3.7 from Travis while waiting for support. You can run teh 3.7 tests with tox without issues * Adding asserts for Payload Endianness * Fixing example of Payload. Same Endianness for builder and decoder. * Fix Sql db slave context validate and get methods - #139 * #353 - debugging, Add debug logs to check size of avaialble data in read buffer * #353 Provide an option to disable inter char timeouts * #353 Bump version, update changelog * check self.socket (#354) * check self.socket self.socket might be None at this point * Update pymodbus/client/sync.py Co-Authored-By: mpf82 <[email protected]> * Fix typo (#378) * Pymodbus 2.2.0 (#375) * #357 Support registration of custom requests * #368 Fixes write to broadcast address When writing to broadcast address (unit_id=0) there should be no response according to the Modbus spec. This fix changes expected_response_length to 0 when writing to unit_id=0. This will break any existing code that is improperly using unit_id 0 for a slave address. * Bump version to 2.2.0 Fix #366 Update failures in sql context Update Changelog Fix major minor version in example codes * Fix #371 pymodbus repl on python3 * 1. Fix tornado async serial client `TypeError` while processing incoming packet. 2. Fix asyncio examples. 3. Minor update in factory.py, now server logs prints received request instead of only function cod * [fix v3] poprawa sprawdzania timeout * Release candidate for pymodbus 2.2.0 * Fix #377 when invalid port is supplied and minor updates in logging * #368 adds broadcast support for sync client and server Adds broadcast_enable parameter to client and server, default value is False. When true it will treat unit_id 0 as broadcast and execute requests on all server slave contexts and not send a response and on the client side will send the request and not try to receive a response. * #368 Fixes minor bug in broadcast support code * Fixed erronous CRC handling If the CRC recieved is not correct in my case my slave got caught in a deadlock, not taking any new requests. This addition fixed that. * Update Changelog * Fix test coverage * Fix #387 Transactions failing on 2.2.0rc2. * Task Cancellation and CRC Errors Alternate solution for #356 and #360. Changes the RTU to make the transaction ID as the unit ID instead of an ever incrementing number. Previously this transaction ID was always 0 on the receiving end but was the unique transaction ID on sending. As such the FIFO buffer made the most sense. By tying it to the unit ID, we can recover from failure modes such as: - - Asyncio task cancellations (eg. timeouts) #360 - Skipped responses from slaves. (hangs on master #360) - CRC Errors #356 - Busy response * Cherry pick commit from PR #367 , Update changelog , bump version to 2.2.0rc4 * #389 Support passing all serial port parameters to asynchronous server * Fix BinaryPayloadDecoder and Builder wrt to coils * Misc updates, bump version to 2.2.0 * ReportSlaveIdResponse now tries to get slave id based on server identity for pymodbus servers * Update missing bcrypt requirement for testing * Fix docs (#407) * Fix document generation * Formatting updates in Changelog * Remove pycrypto dep (#411) It has not been needed by Twisted for a long time, and has been unmaintained for a long time. * Fix --upgrade option in install dependencies (#413) * Fix document generation * Formatting updates in Changelog * Fix --upgrade option in install dependencies * Padding for odd sized responses (#425) If the response is odd size the buffer needs to be padded with an additional byte. * README update: REPL stands for Read Evaluate **Print** Loop (#426) * Drop python 3.4 support (#440) Python 3.4 is EoL and has an easy upgrade path to 3.5+. Support was dropped in Twisted 19.7.0, which is causing Travis to fail pymodbus tests for 3.4. * Re-enable travis python 3.7 builds (#441) * Update __init__.py (#436) * Use SPDX identifier to specify the exact license type (#427) * asyncio server implementation (#400) * #357 Support registration of custom requests * #368 Fixes write to broadcast address When writing to broadcast address (unit_id=0) there should be no response according to the Modbus spec. This fix changes expected_response_length to 0 when writing to unit_id=0. This will break any existing code that is improperly using unit_id 0 for a slave address. * Bump version to 2.2.0 Fix #366 Update failures in sql context Update Changelog Fix major minor version in example codes * Fix #371 pymodbus repl on python3 * 1. Fix tornado async serial client `TypeError` while processing incoming packet. 2. Fix asyncio examples. 3. Minor update in factory.py, now server logs prints received request instead of only function cod * [fix v3] poprawa sprawdzania timeout * Release candidate for pymodbus 2.2.0 * Fix #377 when invalid port is supplied and minor updates in logging * #368 adds broadcast support for sync client and server Adds broadcast_enable parameter to client and server, default value is False. When true it will treat unit_id 0 as broadcast and execute requests on all server slave contexts and not send a response and on the client side will send the request and not try to receive a response. * #368 Fixes minor bug in broadcast support code * Fixed erronous CRC handling If the CRC recieved is not correct in my case my slave got caught in a deadlock, not taking any new requests. This addition fixed that. * Update Changelog * Fix test coverage * Fix #387 Transactions failing on 2.2.0rc2. * Task Cancellation and CRC Errors Alternate solution for #356 and #360. Changes the RTU to make the transaction ID as the unit ID instead of an ever incrementing number. Previously this transaction ID was always 0 on the receiving end but was the unique transaction ID on sending. As such the FIFO buffer made the most sense. By tying it to the unit ID, we can recover from failure modes such as: - - Asyncio task cancellations (eg. timeouts) #360 - Skipped responses from slaves. (hangs on master #360) - CRC Errors #356 - Busy response * Cherry pick commit from PR #367 , Update changelog , bump version to 2.2.0rc4 * native asyncio implementation of ModbusTcpServer and ModbusUdpServer * preliminary asyncio server examples * move serial module dependency into class instantiation * unittests for asyncio based server implementation * induce exception in execute method by mock patching the request object's execute method * move serial module dependency into class instantiation * added asynctest depency to requirements-tests.txt * add unittest skip condition for unsupported targets, remove failing assertion from unsupported targets, use lower asynctest version * remove logger setLevel call since doing so may override library consumers' already set log level * remove async def/await keywords from unittest so that the ast can be loaded in py2 even if the test is to be skipped * Add option to repl allowing Modbus RTU framing on a TCP socket (#447) * repl: Allow Modbus RTU framing on a TCP socket * repl: Update README for framing option * Fix asynci server test failures on python3.6 and below * Bump version to 2.2.0rc1, update six requirements and Changelog * Support multiple Python versions to fix test error from PR #400 (#444) * client/sync.py: Fix missing serial module dependency The serial.connect failed in PR riptideio#400 with "NameError: name 'serial' is not defined" [1]: self = <ModbusSerialClient at 0x7fcda4009b00 socket=None, method=ascii, timeout=3> def connect(self): """ Connect to the modbus serial server :returns: True if connection succeeded, False otherwise """ if self.socket: return True try: > self.socket = serial.Serial(port=self.port, timeout=self.timeout, bytesize=self.bytesize, stopbits=self.stopbits, baudrate=self.baudrate, parity=self.parity) E NameError: name 'serial' is not defined pymodbus/client/sync.py:476: NameError This patch moves the serial import back to the head. [1] https://travis-ci.org/riptideio/pymodbus/jobs/566009109 Fixes: commit e6da559 asyncio server implementation (#400) * server/asyncio.py: Create server with appropriate args and environment If Python is older than 3.7, the create_server will fail like PR riptideio#400 with "unexpected keyword argument 'start_serving'" [1] which is new in Python version 3.7: self.server_factory = self.loop.create_server(lambda :self.handler(self), *self.address, reuse_address=allow_reuse_address, reuse_port=allow_reuse_port, backlog=backlog, > start_serving=not defer_start) E TypeError: create_server() got an unexpected keyword argument 'start_serving' pymodbus/server/asyncio.py:400: TypeError This patch creates server according to Python environment. [1] https://travis-ci.org/starnight/pymodbus/jobs/584178484 Fixes: commit e6da559 asyncio server implementation (#400) * Create asyncio task with appropriate method and environment If Python is older than 3.7, the asyncio.create_task will fail like PR riptideio#400 with "AttributeError: module 'asyncio' has no attribute 'create_task'" [1] which is new in Python version 3.7 [2]: @asyncio.coroutine def testTcpServerCloseActiveConnection(self): ''' Test server_close() while there are active TCP connections''' data = b"\x01\x00\x00\x00\x00\x06\x01\x01\x00\x00\x00\x01" server = yield from StartTcpServer(context=self.context,address=("127.0.0.1", 0),loop=self.loop) > server_task = asyncio.create_task(server.serve_forever()) E AttributeError: module 'asyncio' has no attribute 'create_task' test/test_server_asyncio.py:205: AttributeError This patch creates task according to Python environment. [1] https://travis-ci.org/starnight/pymodbus/jobs/584193587 [2] https://docs.python.org/3/library/asyncio-task.html#creating-tasks Fixes: commit e6da559 asyncio server implementation (#400) * server/asyncio.py: Fix format string for older Python If Python is older than 3.6, f-Strings will fail like PR riptideio#400 with "SyntaxError: invalid syntax" [1] which is new in Python version 3.6 with PEP 498 -- Literal String Interpolation [2]: test/test_server_asyncio.py:14: in <module> from pymodbus.server.asyncio import StartTcpServer, StartUdpServer, StartSerialServer, StopServer, ModbusServerFactory E File "/home/travis/build/starnight/pymodbus/pymodbus/server/asyncio.py", line 424 E _logger.warning(f"aborting active session {k}") E ^ E SyntaxError: invalid syntax This patch fixes the format string with traditional format string syntax. [1] https://travis-ci.org/starnight/pymodbus/jobs/584427976 [2] https://www.python.org/dev/peps/pep-0498/ Fixes: commit e6da559 asyncio server implementation (#400) * test: Make assert_called_once() test only with Python 3.6+ If Python is older than 3.6, unittest.mock.assert_called_once() will fail like PR riptideio#400 with "AttributeError: assert_called_once" [1] which is new in Python version 3.6 [2]: > self.loop.create_server.assert_called_once() test/test_server_asyncio.py:76: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ self = <CoroutineMock name='mock.create_server' id='139638313234784'> name = 'assert_called_once' def __getattr__(self, name): if name in {'_mock_methods', '_mock_unsafe'}: raise AttributeError(name) elif self._mock_methods is not None: if name not in self._mock_methods or name in _all_magics: raise AttributeError("Mock object has no attribute %r" % name) elif _is_magic(name): raise AttributeError(name) if not self._mock_unsafe: if name.startswith(('assert', 'assret')): > raise AttributeError(name) E AttributeError: assert_called_once /opt/python/3.5.6/lib/python3.5/unittest/mock.py:585: AttributeError This patch skips the tests if they are not in Python 3.6+. [1] https://travis-ci.org/starnight/pymodbus/jobs/584431003 [2] https://docs.python.org/3/library/unittest.mock.html#unittest.mock.Mock.assert_called_once Fixes: commit e6da559 asyncio server implementation (#400) * test: Make serve_forever() test only with Python 3.7+ If Python is older than 3.7, asyncio.base_events.Server.serve_forever will fail like PR riptideio#400 with "AttributeError: <class 'asyncio.base_events.Server'> does not have the attribute 'serve_forever'" [1] which is new in Python version 3.7 [2]: @asyncio.coroutine def testTcpServerServeNoDefer(self): ''' Test StartTcpServer without deferred start (immediate execution of server) ''' > with patch('asyncio.base_events.Server.serve_forever', new_callable=asynctest.CoroutineMock) as serve: test/test_server_asyncio.py:81: ... if not self.create and original is DEFAULT: raise AttributeError( > "%s does not have the attribute %r" % (target, name) ) E AttributeError: <class 'asyncio.base_events.Server'> does not have the attribute 'serve_forever' This patch skips the tests if they are not in Python 3.7+. [1] https://travis-ci.org/starnight/pymodbus/jobs/584212511 [2] https://docs.python.org/3/library/asyncio-eventloop.html#asyncio.Server.serve_forever Fixes: commit e6da559 asyncio server implementation (#400) * Add TLS feature for Modbus synchronous (#446) * Add TLS feature for Modbus synchronous Modbus.org released MODBUS/TCP Security Protocol Specification [1], which focuses variant of the Mobdbus/TCP protocol utilizing Transport Layer Security (TLS). This patch enables the Modbus over TLS feature as ModbusTlsClient with the Python builtin module ssl - TLS/SSL wrapper for socket objects. [1]: http://modbus.org/docs/MB-TCP-Security-v21_2018-07-24.pdf * Implement MODBUS TLS synchronous server Since we have the MODBUS TLS synchronous client, we can also have the MODBUS TLS synchronous server. * Fix #461 - Udp client/server , Fix #401 - package license with source, #457 Fix typo's in docstrings, #455-Support float16 * Fix examples, Merge #431 * #401 Move license to root folder from docs
dhoomakethu
added a commit
that referenced
this issue
Sep 11, 2020
* Rebase to dev3.7 * Adding 3.7 to travis configuration * Updated documentation to resolve warnings introduced with the longer names Updated requirements-docs.txt to include missing modules * Fixed reference to deprecated asynchronous * Adding gmp disable to fix pypy build issues * Adding gmp disable to fix pypy build issues * Removing travis python 3.7 configuration Commenting out python3.7 from Travis while waiting for support. You can run teh 3.7 tests with tox without issues * Adding asserts for Payload Endianness * Fixing example of Payload. Same Endianness for builder and decoder. * Fix Sql db slave context validate and get methods - #139 * #353 - debugging, Add debug logs to check size of avaialble data in read buffer * #353 Provide an option to disable inter char timeouts * #353 Bump version, update changelog * check self.socket (#354) * check self.socket self.socket might be None at this point * Update pymodbus/client/sync.py Co-Authored-By: mpf82 <[email protected]> * Fix typo (#378) * Pymodbus 2.2.0 (#375) * #357 Support registration of custom requests * #368 Fixes write to broadcast address When writing to broadcast address (unit_id=0) there should be no response according to the Modbus spec. This fix changes expected_response_length to 0 when writing to unit_id=0. This will break any existing code that is improperly using unit_id 0 for a slave address. * Bump version to 2.2.0 Fix #366 Update failures in sql context Update Changelog Fix major minor version in example codes * Fix #371 pymodbus repl on python3 * 1. Fix tornado async serial client `TypeError` while processing incoming packet. 2. Fix asyncio examples. 3. Minor update in factory.py, now server logs prints received request instead of only function cod * [fix v3] poprawa sprawdzania timeout * Release candidate for pymodbus 2.2.0 * Fix #377 when invalid port is supplied and minor updates in logging * #368 adds broadcast support for sync client and server Adds broadcast_enable parameter to client and server, default value is False. When true it will treat unit_id 0 as broadcast and execute requests on all server slave contexts and not send a response and on the client side will send the request and not try to receive a response. * #368 Fixes minor bug in broadcast support code * Fixed erronous CRC handling If the CRC recieved is not correct in my case my slave got caught in a deadlock, not taking any new requests. This addition fixed that. * Update Changelog * Fix test coverage * Fix #387 Transactions failing on 2.2.0rc2. * Task Cancellation and CRC Errors Alternate solution for #356 and #360. Changes the RTU to make the transaction ID as the unit ID instead of an ever incrementing number. Previously this transaction ID was always 0 on the receiving end but was the unique transaction ID on sending. As such the FIFO buffer made the most sense. By tying it to the unit ID, we can recover from failure modes such as: - - Asyncio task cancellations (eg. timeouts) #360 - Skipped responses from slaves. (hangs on master #360) - CRC Errors #356 - Busy response * Cherry pick commit from PR #367 , Update changelog , bump version to 2.2.0rc4 * #389 Support passing all serial port parameters to asynchronous server * Fix BinaryPayloadDecoder and Builder wrt to coils * Misc updates, bump version to 2.2.0 * ReportSlaveIdResponse now tries to get slave id based on server identity for pymodbus servers * Update missing bcrypt requirement for testing * Fix docs (#407) * Fix document generation * Formatting updates in Changelog * Remove pycrypto dep (#411) It has not been needed by Twisted for a long time, and has been unmaintained for a long time. * Fix --upgrade option in install dependencies (#413) * Fix document generation * Formatting updates in Changelog * Fix --upgrade option in install dependencies * Padding for odd sized responses (#425) If the response is odd size the buffer needs to be padded with an additional byte. * README update: REPL stands for Read Evaluate **Print** Loop (#426) * Drop python 3.4 support (#440) Python 3.4 is EoL and has an easy upgrade path to 3.5+. Support was dropped in Twisted 19.7.0, which is causing Travis to fail pymodbus tests for 3.4. * Re-enable travis python 3.7 builds (#441) * Update __init__.py (#436) * Use SPDX identifier to specify the exact license type (#427) * asyncio server implementation (#400) * #357 Support registration of custom requests * #368 Fixes write to broadcast address When writing to broadcast address (unit_id=0) there should be no response according to the Modbus spec. This fix changes expected_response_length to 0 when writing to unit_id=0. This will break any existing code that is improperly using unit_id 0 for a slave address. * Bump version to 2.2.0 Fix #366 Update failures in sql context Update Changelog Fix major minor version in example codes * Fix #371 pymodbus repl on python3 * 1. Fix tornado async serial client `TypeError` while processing incoming packet. 2. Fix asyncio examples. 3. Minor update in factory.py, now server logs prints received request instead of only function cod * [fix v3] poprawa sprawdzania timeout * Release candidate for pymodbus 2.2.0 * Fix #377 when invalid port is supplied and minor updates in logging * #368 adds broadcast support for sync client and server Adds broadcast_enable parameter to client and server, default value is False. When true it will treat unit_id 0 as broadcast and execute requests on all server slave contexts and not send a response and on the client side will send the request and not try to receive a response. * #368 Fixes minor bug in broadcast support code * Fixed erronous CRC handling If the CRC recieved is not correct in my case my slave got caught in a deadlock, not taking any new requests. This addition fixed that. * Update Changelog * Fix test coverage * Fix #387 Transactions failing on 2.2.0rc2. * Task Cancellation and CRC Errors Alternate solution for #356 and #360. Changes the RTU to make the transaction ID as the unit ID instead of an ever incrementing number. Previously this transaction ID was always 0 on the receiving end but was the unique transaction ID on sending. As such the FIFO buffer made the most sense. By tying it to the unit ID, we can recover from failure modes such as: - - Asyncio task cancellations (eg. timeouts) #360 - Skipped responses from slaves. (hangs on master #360) - CRC Errors #356 - Busy response * Cherry pick commit from PR #367 , Update changelog , bump version to 2.2.0rc4 * native asyncio implementation of ModbusTcpServer and ModbusUdpServer * preliminary asyncio server examples * move serial module dependency into class instantiation * unittests for asyncio based server implementation * induce exception in execute method by mock patching the request object's execute method * move serial module dependency into class instantiation * added asynctest depency to requirements-tests.txt * add unittest skip condition for unsupported targets, remove failing assertion from unsupported targets, use lower asynctest version * remove logger setLevel call since doing so may override library consumers' already set log level * remove async def/await keywords from unittest so that the ast can be loaded in py2 even if the test is to be skipped * Add option to repl allowing Modbus RTU framing on a TCP socket (#447) * repl: Allow Modbus RTU framing on a TCP socket * repl: Update README for framing option * Fix asynci server test failures on python3.6 and below * Bump version to 2.2.0rc1, update six requirements and Changelog * Support multiple Python versions to fix test error from PR #400 (#444) * client/sync.py: Fix missing serial module dependency The serial.connect failed in PR riptideio#400 with "NameError: name 'serial' is not defined" [1]: self = <ModbusSerialClient at 0x7fcda4009b00 socket=None, method=ascii, timeout=3> def connect(self): """ Connect to the modbus serial server :returns: True if connection succeeded, False otherwise """ if self.socket: return True try: > self.socket = serial.Serial(port=self.port, timeout=self.timeout, bytesize=self.bytesize, stopbits=self.stopbits, baudrate=self.baudrate, parity=self.parity) E NameError: name 'serial' is not defined pymodbus/client/sync.py:476: NameError This patch moves the serial import back to the head. [1] https://travis-ci.org/riptideio/pymodbus/jobs/566009109 Fixes: commit e6da559 asyncio server implementation (#400) * server/asyncio.py: Create server with appropriate args and environment If Python is older than 3.7, the create_server will fail like PR riptideio#400 with "unexpected keyword argument 'start_serving'" [1] which is new in Python version 3.7: self.server_factory = self.loop.create_server(lambda :self.handler(self), *self.address, reuse_address=allow_reuse_address, reuse_port=allow_reuse_port, backlog=backlog, > start_serving=not defer_start) E TypeError: create_server() got an unexpected keyword argument 'start_serving' pymodbus/server/asyncio.py:400: TypeError This patch creates server according to Python environment. [1] https://travis-ci.org/starnight/pymodbus/jobs/584178484 Fixes: commit e6da559 asyncio server implementation (#400) * Create asyncio task with appropriate method and environment If Python is older than 3.7, the asyncio.create_task will fail like PR riptideio#400 with "AttributeError: module 'asyncio' has no attribute 'create_task'" [1] which is new in Python version 3.7 [2]: @asyncio.coroutine def testTcpServerCloseActiveConnection(self): ''' Test server_close() while there are active TCP connections''' data = b"\x01\x00\x00\x00\x00\x06\x01\x01\x00\x00\x00\x01" server = yield from StartTcpServer(context=self.context,address=("127.0.0.1", 0),loop=self.loop) > server_task = asyncio.create_task(server.serve_forever()) E AttributeError: module 'asyncio' has no attribute 'create_task' test/test_server_asyncio.py:205: AttributeError This patch creates task according to Python environment. [1] https://travis-ci.org/starnight/pymodbus/jobs/584193587 [2] https://docs.python.org/3/library/asyncio-task.html#creating-tasks Fixes: commit e6da559 asyncio server implementation (#400) * server/asyncio.py: Fix format string for older Python If Python is older than 3.6, f-Strings will fail like PR riptideio#400 with "SyntaxError: invalid syntax" [1] which is new in Python version 3.6 with PEP 498 -- Literal String Interpolation [2]: test/test_server_asyncio.py:14: in <module> from pymodbus.server.asyncio import StartTcpServer, StartUdpServer, StartSerialServer, StopServer, ModbusServerFactory E File "/home/travis/build/starnight/pymodbus/pymodbus/server/asyncio.py", line 424 E _logger.warning(f"aborting active session {k}") E ^ E SyntaxError: invalid syntax This patch fixes the format string with traditional format string syntax. [1] https://travis-ci.org/starnight/pymodbus/jobs/584427976 [2] https://www.python.org/dev/peps/pep-0498/ Fixes: commit e6da559 asyncio server implementation (#400) * test: Make assert_called_once() test only with Python 3.6+ If Python is older than 3.6, unittest.mock.assert_called_once() will fail like PR riptideio#400 with "AttributeError: assert_called_once" [1] which is new in Python version 3.6 [2]: > self.loop.create_server.assert_called_once() test/test_server_asyncio.py:76: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ self = <CoroutineMock name='mock.create_server' id='139638313234784'> name = 'assert_called_once' def __getattr__(self, name): if name in {'_mock_methods', '_mock_unsafe'}: raise AttributeError(name) elif self._mock_methods is not None: if name not in self._mock_methods or name in _all_magics: raise AttributeError("Mock object has no attribute %r" % name) elif _is_magic(name): raise AttributeError(name) if not self._mock_unsafe: if name.startswith(('assert', 'assret')): > raise AttributeError(name) E AttributeError: assert_called_once /opt/python/3.5.6/lib/python3.5/unittest/mock.py:585: AttributeError This patch skips the tests if they are not in Python 3.6+. [1] https://travis-ci.org/starnight/pymodbus/jobs/584431003 [2] https://docs.python.org/3/library/unittest.mock.html#unittest.mock.Mock.assert_called_once Fixes: commit e6da559 asyncio server implementation (#400) * test: Make serve_forever() test only with Python 3.7+ If Python is older than 3.7, asyncio.base_events.Server.serve_forever will fail like PR riptideio#400 with "AttributeError: <class 'asyncio.base_events.Server'> does not have the attribute 'serve_forever'" [1] which is new in Python version 3.7 [2]: @asyncio.coroutine def testTcpServerServeNoDefer(self): ''' Test StartTcpServer without deferred start (immediate execution of server) ''' > with patch('asyncio.base_events.Server.serve_forever', new_callable=asynctest.CoroutineMock) as serve: test/test_server_asyncio.py:81: ... if not self.create and original is DEFAULT: raise AttributeError( > "%s does not have the attribute %r" % (target, name) ) E AttributeError: <class 'asyncio.base_events.Server'> does not have the attribute 'serve_forever' This patch skips the tests if they are not in Python 3.7+. [1] https://travis-ci.org/starnight/pymodbus/jobs/584212511 [2] https://docs.python.org/3/library/asyncio-eventloop.html#asyncio.Server.serve_forever Fixes: commit e6da559 asyncio server implementation (#400) * Add TLS feature for Modbus synchronous (#446) * Add TLS feature for Modbus synchronous Modbus.org released MODBUS/TCP Security Protocol Specification [1], which focuses variant of the Mobdbus/TCP protocol utilizing Transport Layer Security (TLS). This patch enables the Modbus over TLS feature as ModbusTlsClient with the Python builtin module ssl - TLS/SSL wrapper for socket objects. [1]: http://modbus.org/docs/MB-TCP-Security-v21_2018-07-24.pdf * Implement MODBUS TLS synchronous server Since we have the MODBUS TLS synchronous client, we can also have the MODBUS TLS synchronous server. * Fix #461 - Udp client/server , Fix #401 - package license with source, #457 Fix typo's in docstrings, #455-Support float16 * Fix examples, Merge #431 * #401 Move license to root folder from docs * rtu_framer: fix processing of incomplete frames (#466) * rtu_framer: fix processing of incomplete frames * rtu_framer: fix test case * Add handle local echo option * Update constants.py Added RetryOnInvalid flag and Backoff delay. * Update transaction.py Added retry on invalid data received and exponetial backoff delay between retries. * Add TLS feature for Modbus asynchronous (#470) * Add TLS feature for Modbus asynchronous client Since we have Modbus TLS client in synchronous mode, we can also implement Modbus TLS client in asynchronous mode with ASYNC_IO. * Add TLS feature for Modbus asynchronous server Since we have Modbus TLS server in synchronous mode, we can also implement Modbus TLS server in asynchronous mode with ASYNC_IO. * PR #471 Fix transaction tests * Fix failing tests * Add "Python" trove classifier Previously only generic "Python" support (without a version) was announced. * Merge PR's , bump version to 2.4.0 * closes #481, #482, #483, #484 * Closes #491 * Asyncio bug fixes (#517) * Closes #491 * 1. update requirements 2. Fix examples 3. Fix #494 - handle_local_echo 4. Fix #500 -- asyncio serial client with already running loop 5. Fix #486 - Pass serial args for asyncio serial client 6. Fix #490 - Typo in decode_data for socker_framer 7. Fix #385 - Support timeouts to break out of responspe await when server goes offline 8. Misc updates * #516 custom data block fix * Update Changelogs , bump version to 2.4.0 * #515 fix repl broadcast (#531) * 1. update requirements 2. Fix examples 3. Fix #494 - handle_local_echo 4. Fix #500 -- asyncio serial client with already running loop 5. Fix #486 - Pass serial args for asyncio serial client 6. Fix #490 - Typo in decode_data for socker_framer 7. Fix #385 - Support timeouts to break out of responspe await when server goes offline 8. Misc updates * #516 custom data block fix * Fix broadcast error with REPL client #515 * Fix #509 Wrong unit ID referenced in framers * Update documentation for serial forwarder example. Fixes #525 * Fix unit tests, support python 3.8 for tests, renamed: pymodbus/server/asyncio.py -> pymodbus/server/async_io.py and pymodbus/client/asynchronous/asyncio -> pymodbus/client/asynchronous/async_io * Ignore python3 code syntax while reporting coverage * Fix tests failing on python 3.6 and osx * Fix typo in makefile * Fix test execution errors specific to python3.6 * Osx travis issue - Fix trial 1 * Travis reverting xcode to 8.x for mac osx * Pymodbus v2.4.0 Co-authored-by: dices <[email protected]> Co-authored-by: Eric Duminil <[email protected]> Co-authored-by: Mike <[email protected]> Co-authored-by: Kim Hansen <[email protected]> Co-authored-by: Michael Corcoran <[email protected]> Co-authored-by: Andrea Canidio <[email protected]> Co-authored-by: tcplomp <[email protected]> Co-authored-by: alecjohanson <[email protected]> Co-authored-by: hackerboygn <[email protected]> Co-authored-by: Yegor Yefremov <[email protected]> Co-authored-by: Memet Bilgin <[email protected]> Co-authored-by: Sekenre <[email protected]> Co-authored-by: sanjay <[email protected]> Co-authored-by: Jian-Hong Pan <[email protected]> Co-authored-by: Steffen Vogel <[email protected]> Co-authored-by: Alexey Andreyev <[email protected]> Co-authored-by: Wild Stray <[email protected]> Co-authored-by: Lars Kruse <[email protected]>
|
Two year later, "client.inter_char_timeout = 0.05" just saved my ass. :-) |
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Versions
Pymodbus Specific
Description
Since upgrading to version 2.1.0 I cannot read the whole input registers as I could with other versions.
Instead I receive an error:
Code and Logs
Log with pymodbus version 2.0.1
Log with pymodbus version 2.1.0
This might be related to issue #328 - however my code runs fine with pymodbus version 1.5.2 - that's why I have created a new issue.
Actually, I tested pretty much all pymodbus versions from 1.2.0 onwards, but the code "breaks" with version 2.1.0
If I change the numer of registers to read (currently 18) to something smaller it works. To be more specific, everything from 1 to 6 works, for everything else I get the mentioned error.
I have also tried with a third party Modbus viewer - also no problems there when trying to read 18 registers starting at 0x1100.
The text was updated successfully, but these errors were encountered: