Saxon » SaxonC

I have now tried to get rid of all the web api or web framework stuff and I have tried to use a simple Python multithreading app with SaxonC HE 11.2, it also gets that EXCEPTION_ACCESS_VIOLATION:

import threading
import time

from saxonc import *

exitFlag = 0

class myThread (threading.Thread):
   def __init__(self, threadID, name, counter, saxon_proc):
      threading.Thread.__init__(self)
      self.threadID = threadID
      self.name = name
      self.counter = counter
      self.saxon_proc = saxon_proc
      self.xquery_processor = saxon_proc.new_xquery_processor()
   def run(self):
      print ("Starting " + self.name)
      example(self.name, self.counter, 5, self.xquery_processor)
      print ("Exiting " + self.name)

def example(threadName, delay, counter, xquery_processor):
   while counter:
      if exitFlag:
         threadName.exit()
      time.sleep(delay)
      print(xquery_processor.run_query_to_string(query_text = '"Hello from SaxonC: " || current-dateTime()'))
      counter -= 1


with PySaxonProcessor(license = False) as saxon_proc:
    # Create new threads
    thread1 = myThread(1, "Thread-1", 1, saxon_proc)
    thread2 = myThread(2, "Thread-2", 2, saxon_proc)

    # Start new Threads
    thread1.start()
    thread2.start()
    thread1.join()
    thread2.join()
    print ("Exiting Main Thread")

Output:

SaxonProc constructor(l) called
SaxonProc constructor: jvm exists! jvmCreatedCPP=1
SaxonProc constructor(l) called
New processor created, Processor: -29383024
New processor created, Processor: -29383024
Starting Thread-1
Starting Thread-2
DEBUG: SaxonCAPI: param-name: qs, type of Value= java.lang.String
qs
<?xml version="1.0" encoding="UTF-8"?>Hello from SaxonC: 2022-03-05T11:35:18.791+01:00
DEBUG: SaxonCAPI: param-name: qs, type of Value= java.lang.String
qs
<?xml version="1.0" encoding="UTF-8"?>Hello from SaxonC: 2022-03-05T11:35:19.691+01:00
DEBUG: SaxonCAPI: param-name: qs, type of Value= java.lang.String
qs

JET RUNTIME HAS DETECTED UNRECOVERABLE ERROR: system exception at 0x0000000000a6730e

That output on Windows at least shows that calling SaxonC works twice before crashing, on Linux it looks to go down earlier:

python3 Python3SimpleThreadingSaxonCHE112Ex1.py
Starting Thread-1
Starting Thread-2

JET RUNTIME HAS DETECTED UNRECOVERABLE ERROR: system exception at 0x00007ff275ebed49
Please, contact the vendor of the application.
Core dump will be written to "/mnt/wslg/dumps/core.%e"
Extra information about error is saved in the "jet_err_8068.txt" file.

JET RUNTIME HAS DETECTED UNRECOVERABLE ERROR: system exception at 0x00007ff275ebed49
Please, contact the vendor of the application.
Core dump will be written to "/mnt/wslg/dumps/core.%e"

Signal 11 (SIGSEGV)
siginfo: si_signo 11, si_errno 0, si_code 1, si_addr 0x000000000000000f

Version Information:

  Java version: 1.8.0_181
  Excelsior JET 15.30 Enterprise edition
  JET Profile: OpenJDK version: 1.8.0_181; JET update level: 6; CPU architecture: amd64
  Runtime: Server
  CPU features: cmov mmx sse sse2 sse3 ssse3 sse4.1 sse4.2 avx avx2 fma f16c lzcnt popcnt bmi1 bmi2 adx cx8 cx16 movbe
  Application was deployed

Options and system properties:

  -Djet.jit.disable.resolution=
  -Djet.gc.heaplimit=0
  -Djet.stack.trace=

Entry point type: Invocation API

Command line: "python3" "Python3SimpleThreadingSaxonCHE112Ex1.py"

OS:

Ubuntu 20.04.4 LTS
Linux 5.10.60.1-microsoft-standard-WSL2 #1 SMP Wed Aug 25 23:20:18 UTC 2021 x86_64
glibc 2.31

JET-compiled Components:

/usr/lib/libsaxonhec.so dll, version info: jet-1530-mp1 (ent, en)

Is there any unit test trying to use SaxonC 11 with Python and multithreading?

Now tried to run the Linux Python multithreading stuff in gdb, log is

gdb) run Python3SimpleThreadingSaxonCHE112Ex1.py
Starting program: /usr/bin/python3 Python3SimpleThreadingSaxonCHE112Ex1.py
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
[New Thread 0x7ffff3e21700 (LWP 11271)]
[New Thread 0x7ffff3620700 (LWP 11272)]
[Thread 0x7ffff3620700 (LWP 11272) exited]
[New Thread 0x7ffff2e0b700 (LWP 11273)]
[New Thread 0x7ffff2bed700 (LWP 11274)]
[New Thread 0x7ffff298d700 (LWP 11275)]
[New Thread 0x7ffff296f700 (LWP 11276)]
[New Thread 0x7ffff2951700 (LWP 11277)]
[New Thread 0x7ffff2933700 (LWP 11278)]
[New Thread 0x7ffff2915700 (LWP 11279)]
[New Thread 0x7ffff28f7700 (LWP 11280)]
[New Thread 0x7ffff28d9700 (LWP 11281)]
[New Thread 0x7ffff28bb700 (LWP 11282)]
[New Thread 0x7ffff289d700 (LWP 11283)]
[New Thread 0x7ffff21dc700 (LWP 11284)]
[New Thread 0x7ffff20fa700 (LWP 11285)]
[New Thread 0x7ffff2018700 (LWP 11286)]
[New Thread 0x7ffff1dc3700 (LWP 11287)]
[New Thread 0x7ffff1ae1700 (LWP 11288)]
[New Thread 0x7ffff3620700 (LWP 11289)]
Starting Thread-1
[New Thread 0x7ffff17ff700 (LWP 11290)]
Starting Thread-2

Thread 20 "python3" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7ffff3620700 (LWP 11289)]
0x00007ffff3e7fd49 in ?? () from /usr/lib/libsaxonhec.so

With the DEBUG flag enabled under Linux I get the following exception java.lang.StackOverflowError:

python3 Python3SimpleThreadingSaxonCHE112Ex1.py
SaxonProc constructor(l) called
SaxonProc constructor: jvm exists! jvmCreatedCPP=1
SaxonProc constructor(l) called
New processor created, Processor: 550045968
New processor created, Processor: 550045968
Starting Thread-1
Starting Thread-2
Exception in thread "main" java.lang.StackOverflowError

Hi Martin,

Thank you for your posts. I will look carefully at your multithreading examples ASAP. Python web apps that work with multithreading is under tested.

Tips

• Not to use Context manager (i.e. with) in multi-threaded applications as the release method may be called before a thread is done. This also applies to in Flask.

• In Flask applications it is best to cache the PySaxonProcessor object so that it can be used throughout the application.

I have tried the multithreading code without with, it is

import threading
import time

from saxonc import *

exitFlag = 0

class myThread (threading.Thread):
   def __init__(self, threadID, name, counter, xquery_processor):
      threading.Thread.__init__(self)
      self.threadID = threadID
      self.name = name
      self.counter = counter
      self.xquery_processor = xquery_processor
   def run(self):
      print ("Starting " + self.name)
      example(self.name, self.counter, 5, self.xquery_processor)
      print ("Exiting " + self.name)

def example(threadName, delay, counter, xquery_processor):
   while counter:
      if exitFlag:
         threadName.exit()
      time.sleep(delay)
      print("%s : %s" % (threadName, xquery_processor.run_query_to_string(query_text = '"Hello from SaxonC: " || current-dateTime()')))
      counter -= 1


saxon_proc = PySaxonProcessor(license = False)

# Create new threads
thread1 = myThread(1, "Thread-1", 1, saxon_proc.new_xquery_processor())
thread2 = myThread(2, "Thread-2", 2, saxon_proc.new_xquery_processor())

# Start new Threads
thread1.start()
thread2.start()
thread1.join()
thread2.join()
print ("Exiting Main Thread")

saxon_proc.release()

but (tested on Windows) fails with the same

JET RUNTIME HAS DETECTED UNRECOVERABLE ERROR: system exception at 0x0000000000a6730e
Please, contact the vendor of the application.
Crash dump will be written to "C:\Users\marti\source\repos\PythonMultiThreadingSaxonCHE112NoWithTest1\PythonMultiThreadingSaxonCHE112NoWithTest1\jet_dump_1504.dmp"

Exception 0xC0000005 (EXCEPTION_ACCESS_VIOLATION) at 0x0000000000a6730e (C:\Program Files\Saxonica\SaxonC HE 11.2\libsaxonhec.dll+0x66730e)
Failed to read memory at 0x000000a2b1240000

after calling each thread's method once:

Starting Thread-1
Starting Thread-2
Thread-1 : <?xml version="1.0" encoding="UTF-8"?>Hello from SaxonC: 2022-03-06T12:30:23.362+01:00
Thread-2 : <?xml version="1.0" encoding="UTF-8"?>Hello from SaxonC: 2022-03-06T12:30:24.339+01:00

JET RUNTIME HAS DETECTED UNRECOVERABLE ERROR: system exception at 0x0000000000a6730e

Thanks Martin for reporting this issue. I have reproduced it on my linux machine. Created the following bug issue to keep track of its progress: https://saxonica.plan.io/issues/5373

Hi Martin,

The bug issue #5373 is now resolved, I have since turned my attention to running SaxonC on Flask. I am now getting a response with the simple example app you posted here: https://saxonica.plan.io/boards/4/topics/8523?r=8608#message-8608

After a few requests the server crashes as follows:

JET RUNTIME HAS DETECTED UNRECOVERABLE ERROR: runtime error
Thread 33B13 ["Thread-1"] is terminated without notifying the JVM. Probably, "DetachCurrentThread" function was not called
Please, contact the vendor of the application.
Core dump will be piped to "/usr/share/apport/apport %p %s %c %d %P %E"
Extra information about error is saved in the "jet_err_211711.txt" file.

Aborted (core dumped)

I did expect this kind of error. I am wondering where we can add the call to detach_current_thread.

Hi,

thanks for looking into this. I am afraid I have no immediate idea on where/how to use your resolution of that multithreading bug and the addition of the detatch_current_thread to write a well working flask app, mainly because I don't know details as to when a request results in a new thread handling it or a request terminates a thread with flask, partly of course because I will probably need to run/test the API addition to understand its use reasonable well.

The general advice was also to make sure the PySaxonProcessor is stored/cached but even there I have not made progress as it seems so far that all kind of stores/caches the framework provides wants cacheable/storable objects to be kind of JSON serializable.

I am not sure how to proceed, if you want to release Saxon 11.3 perhaps it is better to leave this issue open and at least have a new release were basic multithreading using Saxon in Python works in a controlled way before looking for further improvements to find ways to control it inside of a more complex web application.

I will see whether I can write some test code over the weekend to get a better understanding of where/when Flask creates a new thread or terminates an existing one, will also look at the documentation whether it illuminates about the use of threads. Before that I don't have a good suggestion as to where to call detach_current_thread, I don't know whether I will have a good suggestion after doing some tests, we'll see.

I have resolved the issue of calling the detach_current_thread. I found this article helpful: https://flask.palletsprojects.com/en/2.0.x/appcontext/

In particular the details about teardown_X:

teardown_X() closes or otherwise deallocates the resource if it exists. It is registered as a teardown_appcontext() handler.

In the example I added the following:

@app.teardown_appcontext
def teardown_saxonc_thread(exception):
    proc.detach_current_thread

All requests now run without crashing.