NAK errors #7
Comments
|
It looks like it is failing when it tries to open the file to save to: Does that path exist on the machine that the Logic software is running on There's no good way for the script to do platform detection of the remote On Fri, Oct 30, 2015 at 14:26 AnalogIO [email protected] wrote:
|
|
I'm testing on a Mac so the '~/Desktop/' folder does exist. I don't have a file already called test.logicdata at this path but I assume that is ok. |
|
I'm testing this some more, here is the latest output Setting to sample rate to at least digitial 4 MS/s, analog 100 S/s Starting a capture It looks to me like the 'IS_PROCESSING_COMPLETE' cmd is resulting in a NAK error. Is it possible that if processing is not complete it returns a NAK? |
|
Internally, the Saleae software has some (previously not documented) requirements: The IS_PROCESSING_COMPLETE command is supposed to NAK when capture processing is not complete, this is in the documentation too. Generally you should wait in a loop until the function returns ACK. |
|
ok so now I am able to start captures via python but I am unable to export the data as csv. I am using the following code: The logic captures folder does exist, test.csv does not. The commands execute just fine but no file can be found after execution. |
|
Pipe the output of the Logic binary to a file when you launch it, and then run the script. After the script completes, post a copy of the log file. That might contain useful information about the export process. Make sure you're using 1.2.5, it prints out much more detailed debug data when socket based export commands are not properly formed. Also, the data export command is crazy complicated. I created a new command, EXPORT_DATA2, which formally covers all possible export combinations, where the original EXPORT_DATA command didn't actually handle all possible export options properly. Here is the C# source for the export_data2 wrapper: Here are the tests we scripted to test all significant export option variants: The API is a 1 to 1 direct mapping to the internal settings struct that is configured with the normal socket GUI. unfortunately it's not possible to use the GUI to create a socket command, but at least you can see what settings need to be set from the GUI based on your capture settings, and then try to replicated then in the API. The socket export command is very sensitive to the type of data you captured. Like this: Even if you're just exporting digital, it still matters if you capture analog or not. If you tell me what analog and digital channels you had active, and what kind of file you want (CSV, just digital, etc) I can help form the correct command. You can also use the examples from the ExportUnitTests.cs file to find an export mode that closest matches your situation. |
|
So I've been trying to pipe out that data, this is what I put into the terminal to launch but not having any luck: |
|
I've never tried this using open before. Here is how I normally capture the log on OSX: |
|
That worked much better... What I am trying to export is digtal only channels 0,1,2 |
|
Thanks for the log. Unfortunately, it looks like there isn't any more information about the cause of the issue here. What would really help is a copy of the text string that the python code sent to the Logic application. It looks like the python code is already logging it. Right before the message is sent to the socket, this is called: Also, from your log, I can see that you captured both digital and analog channels, but only selected digital channels for export. There are special requirements for this, and I think it's one of the edge cases that caused us to write EXPORT_DATA2 to replace EXPORT_DATA, which this python wrapper is currently using. As soon as I have the log from the python script and which channels were active in the capture, I can figure out what's going wrong. |
|
Here's the latest log: https://gist.github.com/analog-io/bd1a52474cd68fddb371 I was not able to grab the log from python but I did step debug the code in pyCharm and extracted this string that was constructed from: 'EXPORT_DATA, /Users/lgbeno/Desktop/test.csv, digital_channels, 4, 5, 6, 7, all_time, csv, headers, comma, time_stamp, combined, row_per_change, hex' |
|
For the future, the library uses the standard Python logging facility. Before calling saleae.demo() import logging, saleae
logging.basicConfig(level=logging.DEBUG)
saleae.demo()I can also add something to pick up a DEBUG environment variable. |
|
@analog-io, which analog and digital channels were enabled when the capture was taken? I'll need that to reproduce your capture settings correctly, as that information is not in the software output. |
|
Thanks, just those channels were open in Logic. |
|
@Marcus10110 has there been an update on this? |
|
Sorry for the long delay on this! I did find a bug in the Logic software - the EXPORT_DATA function only checks for the "ROW_PER_CHANGE"/"ROW_PER_SAMPLE" parameter if "SEPARATE" is specified instead of "COMBINED". Then, also incorrectly, the software still requires the display radix, even if "SEPARATE" is specified. Basically, the correct behavior is that the display radix parameter (hex/dec/bin/etc) should only be required if "COMBINED" is specified. The good news is that when the density parameter ("ROW_PER_CHANGE"/"ROW_PER_SAMPLE") is not parsed, it uses the previously specified value, so you can set this manually once in the export settings in the GUI, and then it should get preserved. Plus ROW_PER_CHANGE is the default mode in the software. The command that works: Because we've deprecated the EXPORT_DATA command, I'm not sure if I should fix it in the Saleae software. The good news is that this is not broken in EXPORT_DATA2, which replaced EXPORT_DATA. Unfortunately, EXPORT_DATA2 hasn't been ported from our C# example application to this Python library, and I won't have the time to take care of this. The ideal solution would be to update this library to no longer use the EXPORT_DATA function, and use the new one, but unfortunately I don't have the time right now to do that. A short term solution would be to work around this my supporting the bug above. |
|
Thanks for the update and looking into this. Really appreciate it and understand theres a million things to do. Knowing this issue I'll either hack my local code to work or attempt to port the C# version to Python and do a pull request. Thanks again! |
|
Yes thanks @Marcus10110 that's very helpful for the EXPORT_DATA command. Hopefully someone will find time to get EXPORT_DATA2 ported soon. Unfortunately I'm also seeing an NAK error when running the I'm getting the following failure: So it seems there is some problem with "IS_PROCESSING_COMPLETE"? Additionally, one of my colleagues is getting the error below when calling Has the API for |
|
First "IS_PROCESSING_COMPLETE" Basically, the capture command ACKs right after the capture is complete, but the software is still running some housekeeping to get the capture data to the screen and the UI state updated. Alternatively, instead of adding the wait statement, the first NAK exception could just be ignored. About the second issue "CAPTURE_TO_FILE": does the capture still start in the software? It will NAK if there is a problem with the command, or if a capture error occurred and no data was recorded. Also, is the <user_path> path mentioned here absolute? It can't start with ~, our software doesn't support that. I really should fix that. I've gone through the source and added console print statements for the different conditions that cause NAKs in several of the common commands. 1.2.6 comes out this Friday, and when it does, we can check the console output for more information. |
|
cc: @Marcus10110 from Saleae These race conditions / CommandNAKedErrors are widespread throughout python-saleae and/or the Saleae Logic software. For example, if I repeatedly run get_active_channels(), I get errors the first several times, and then it happens to succeed on the 4th or 5th try. Here's an example: SUDDENLY IT WORKS! Maybe I just got lucky and the race condition didn't occur the 5th time! I've worked around this in the capture and data-export by replacing occurrences of is_processing_complete() with the following from (#9) @bacaj1 def wait_for_processing_complete(self, timeout=5):
"""Wraps self.is_processing_complete with more lenient waiting.
There is some race condition in the Saleae software!
Args:
timeout: (int) seconds before timeout
Raises:
IOError: if timed out
"""
tic = time.time()
time.sleep(0.2)
done = False
while not done:
try:
# Check for timeout in outer loop
if time.time() - tic > timeout:
self.capture_stop()
raise IOError('Saleae timeout = {}sec reached.'.format(timeout))
while not self.is_processing_complete():
# Check for timeout in inner loop
if time.time() - tic > timeout:
self.capture_stop()
raise IOError('Saleae timeout = {}sec reached!'.format(timeout))
time.sleep(0.2)
done = True
except self.CommandNAKedError:
# This exception is raised if processing of capture is not complete on the
# side of the analyzer. This is expected for the first few calls, so we
# simply ignore it until is_processing_complete returns true
passAnd for file export, I use 'os' to poll for the output file existing and having the same size in two successive samples, using a similar timeout scheme. It would be nice if sal.is_processing_complete() actually worked, and someone cleaned up what I'm assuming are race conditions / bugs in the Saleae backend. I contacted Saleae support about the issue and they suggested I cc @Marcus10110 from Saleae on this report. |
|
A similar workaround seems to work for _recv(...): def _recv(self, timeout=1):
tic = time.time()
done = False
while not done:
try:
if time.time() - tic > timeout:
raise IOError('Saleae timeout = {}sec reached.'.format(timeout))
while 'ACK' not in self._rxbuf:
if time.time() - tic > timeout:
raise IOError('Saleae timeout = {}sec reached.'.format(timeout))
self._rxbuf += self._s.recv(1024).decode('UTF-8')
log.debug("Recv >{}<".format(self._rxbuf))
if 'NAK' == self._rxbuf[0:3]:
self._rxbuf = self._rxbuf[3:]
raise self.CommandNAKedError
ret, self._rxbuf = self._rxbuf.split('ACK', 1)
done = True
except self.CommandNAKedError:
pass
return ret |
|
get_active_channels NAKs for 2 reasons:
|
|
@analog-io Were you able to hack your local code? Do you have any partial solution at least? |
|
@Marcus10110 Could you be more specific on how we are supposed to change the parameters in the GUI to make it work? Thanks |
It looks like this is partially working but I'm getting some NAK Errors and the program crashes when the analyzer starts to capture data.
Using Python 3.5.0
Saleae connected.
Press Enter to continue...
Set performance to full.
Press Enter to continue...
Connected devices:
<saleae.ConnectedDevice #1 LOGIC_PRO_8_DEVICE Logic Pro 8 (ffcda53224bbb23a) ACTIVE>
Only one Saleae device. Skipping device selection
Press Enter to continue...
Setting active channels (digital=[0, 1, 2, 3, 4], analog=[0, 1])
Press Enter to continue...
Reading back active channels:
digital=[0, 1, 2, 3, 4]
analog=[0, 1]
Press Enter to continue...
Setting to capture 2e6 samples
Press Enter to continue...
Setting to sample rate to at least digitial 4 MS/s, analog 100 S/s
Set to (100000000, 100)
Press Enter to continue...
Starting a capture
Traceback (most recent call last):
File "", line 1, in
File "/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/saleae/saleae.py", line 546, in demo
while not s.is_processing_complete():
File "/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/saleae/saleae.py", line 360, in is_processing_complete
resp = self._cmd('IS_PROCESSING_COMPLETE')
File "/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/saleae/saleae.py", line 116, in _cmd
ret = self._recv()
File "/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/saleae/saleae.py", line 107, in _recv
raise self.CommandNAKedError
saleae.saleae.CommandNAKedError
The text was updated successfully, but these errors were encountered: