Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix for segmentation fault on exit (Python) #239

Merged
merged 1 commit into from
Nov 18, 2017
Merged

Fix for segmentation fault on exit (Python) #239

merged 1 commit into from
Nov 18, 2017

Conversation

pietrodn
Copy link
Contributor

NeoPixel Python examples, when quit with Ctrl+C, crash with a SEGFAULT on the Raspberry Pi 3 model B:

~/rpi_ws281x/python $ sudo python examples/strandtest.py
Press Ctrl-C to quit.
Color wipe animations.
^CTraceback (most recent call last):
  File "examples/strandtest.py", line 92, in <module>
    colorWipe(strip, Color(255, 0, 0))  # Red wipe
  File "examples/strandtest.py", line 30, in colorWipe
    time.sleep(wait_ms/1000.0)
KeyboardInterrupt
Segmentation fault

After some debugging I fixed the problem by removing the ws.ws2811_fini(self._leds) call from the _cleanup() method in the Adafruit_NeoPixel Python class.

I also removed the useless __del__() because _cleanup() already gets called by atexit().
Now the Python samples do not SEGFAULT anymore.

This problem is not present with the compiled C samples.

Remove useless __del__ because the cleanup() function gets called by
atexit().
Remove ws.ws2811_fini() call which causes the SEGFAULT on exit.
Python does not complain about the memory leak in this way.
@jgarff jgarff merged commit d4c7e35 into jgarff:master Nov 18, 2017
@pietrodn pietrodn deleted the segfault-fix branch November 18, 2017 18:47
@zoltan-k
Copy link

mind you, the code still segfaults if the event handler is registered to clear the LEDs before exiting (pass -c as an argument)

in my own code segfault happens also if the code to clear the leds is registered as an atexit function.

Clearing the leds happens, just I'm getting a segfault.

@pietrodn
Copy link
Contributor Author

Executing sudo python example/strandtest.py -c on a Raspberry Pi 3B, I do not get the SEGFAULT. Can you reproduce the problem on strandtest.py?

@zoltan-k
Copy link

yes, also on a Pi3.
it does not happen without -c, and I don't have ws.ws2811_fini(self._leds) in the _cleanup() method, thus I must assume I am using the latest version pulled right now.

@EvgEvg
Copy link

EvgEvg commented Nov 25, 2017

@pietrodn Hello, and thanks for this fix man, I was getting segfault and now it's fixed. Could you link to some materials or briefly describe how you debug segfaults such as the one you fixed? Is it enabling core dump and then using gdb or something else? Thank you!

@pietrodn
Copy link
Contributor Author

@Zedius in the next days I will try to reproduce the problem.
@EvgEvg I just placed some print statements in the _cleanup() method to see where the program crashed and experimented with different combinations of function calls.

brindosch pushed a commit to hyperion-project/rpi_ws281x that referenced this pull request Apr 5, 2018
* Added delay to ensure reset time being met with low led counts (jgarff#157)

* Fix includes as per jgarff#178 (jgarff#189)

* clean up doco and strandtest (jgarff#187)

* updated documentation on GPIO usage
strandtest now defaults to GRB colour order
added version.h to gitignore

* updated python readme

* more README cleanup

* added notes on spi max transfer size

* Fix freeze after running for long time  (jgarff#197)

* Whitespace cleanup

* Move channel_protocol_time so it handles 32 bit Color LEDs.

* Fix a freezing issue when running for a long time.

When current_timestamp overflowed 2^64, rolling back to 0 and ws2811->render_wait_until had not, this caused an incredibly long usleep in ws2811_render.

* Update README.md (jgarff#191)

* Update README.md

* updated spi buffer size location

* Add -lrt link flag for systems with older glibc (jgarff#195)

* Fix 32-bit overflow in get_microsecond_timestamp(). (jgarff#206)

On Raspbian systems, the tv_sec member of struct timespec is 32 bits
wide.  get_microsecond_timestamp() was overflowing in 32-bit
arithmetic on this member.  This causes the time that it returns to
jump ahead by many thousands of years at uptime = 35 minutes, 48
seconds.  (While the starting point for CLOCK_MONOTONIC_RAW is
unspecified, it seems to track system uptime in practice.)

This overflow didn't cause noticeable problems for me--ws2811_render()
skips the sleep when it thinks such a huge amount time has
elapsed--but it could possibly result in rarely-occurring skipped
frames if the WS2811 reset time isn't effectively honored by
ws2811_render().

* Add TOOLCHAIN option for cross compilation (jgarff#213)

Use this option if you want to cross-compile the library.
Typical usage is:

  scons TOOLCHAIN=arm-linux-gnueabihf

* Add -Wextra compile option and adapt code to suppress warnings (jgarff#214)

The changes in code are mostly type casts between int ann unsigned.

* Add shared library rasbpian package builder.  The package is not built by
default.  To generate the package you need:

- The dpkg-deb utility.
- Use "scons deb" to generate the package.

* Merge user-settable gamma correction into master (jgarff#216)

Merge user-settable gamma correction into master

* Bug fixes for Stretch and Compute Module support (jgarff#222)

* Bug fixes

Missing aligned(4) attribute in pcm.h makes the PCM case failing under
Stretch, due to a newer version of gcc (6.x).
Also fixed a bug in check_hwver_and_gpionum, so original Model A is
recognised.

* README, using SPI on RPi3

Added note to change the core frequency to 250 MHz when using SPI on
RPi 3.

* README fix

* Add support for Compute Module 1 and 3

* Fix undefined clock_gettime symbol

* add missing memory allocation for SPI mode (jgarff#231)

* added Ctrl-C handler to turn off all LEDs on exit (jgarff#140)

* added Ctrl-C handler to turn off all LEDs on exit

* added option -c

* Fix for segmentation fault on exit for Python API (jgarff#239)

Remove useless __del__ because the cleanup() function gets called by
atexit().
Remove ws.ws2811_fini() call which causes the SEGFAULT on exit.
Python does not complain about the memory leak in this way.

* Add support for new A+ hardware.

* Change default DMA channel to safe choice (10) (jgarff#253)

add warning on README.md

* Add getBrightness for brightness feedback from the buffer (jgarff#248)

* Instructions to install the Python library (jgarff#254)

Solving issue jgarff#142

* Fix for non-root users & SPI (jgarff#257)

* Fix for non-root users can use SPI

* use DEFINE'd device strings

also add a note to the readme

* Basic Golang Examples (jgarff#256)

* Added golang example

* Added basic README for golang examples. Added os exits for errors during color wipes for golang examples

* Fix the default DMA in python code (jgarff#266)

DMA 5 causes file system errors on PI 3

* Fixed typo in README.md (jgarff#273)

Was perusing recent commits and saw this.

* clarify electrical notes (jgarff#264)

* Fixed strandtest.py (jgarff#258)

* Fixed keyboardinterrupt, indentation inconsistency, & a few other minor bugs.  Added shebang.

* Fixed KeyboardInterrupt.  Again.

* KeyboardInterrupt works now.  For sure.

* Tabs to spaces, removed redundant signal handling

* Removed unused sys module import

* Added Raspberry Pi 3 B+ (jgarff#276)

* Added Pi 3 B+

* Remove unused import
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants