I often get the "broken pipe" error when I have a debug breakpoint and exactly on the line where I set the breakpoint.
No breakpoint, no error... it must be a problem in the debugger...

To reproduce the "[Errno 32] Broken pipe" error, build the attached algorithm after having set for example a breakpoint at line 142 and an enough long back-testing period (i.e. from 01/03/2002 to 01/11/2015)

I hope, that the Quantopian Team will prioritize this issue, it's a real bottleneck! I reported it for the first time in Januar.
The Quantopian platform looks very promising, but without a debugger that works properly, it is very hard to do serious development.

You've hit the issue on the head, this error is a bug in the debugger. I've gone ahead to file it internally and we'll take a look to fix it.

I'm not quite sure yet what's causing the message and I'm sorry for the troubles. Once we patch it, I'll follow up in this thread.

I've been running into several broken pipe error recently, any progress?

Unfortunately the error appears to be random, the same breakpoint will work maybe 85 percent of the time, then it will throw an intermittent error. It does appear to be related to print statements more often than not. I will set a date and time logical test then set the result to print 'OK' after the date time if statement. I set the breakpoint on the print statement to stop processing at the specified date and time. Again, the error appears to be random on the print statement breakpoint. I have been testing all day and have set other breakpoints within the code on 'regular' statements past the 'initial' date and time if statement breakpoint, for example...

if date == 140506 and time == 1320 :
print('OK')

Broken Pipe on stackoverflow

In the neighborhood of buffer, SIGPIPE, keep-alive, seems a process was already abandoned/timed-out and doomed but user doesn't find out until the breakpoint is hit later.

Im my experience this error occurs when the backtesting period is quite long, i.e. from 01/03/2002 to 01/11/2015 and when the code has several if-blocks.

If you servers run on a Linux/Unix System, maybe this could help: http://coding.derkeiler.com/Archive/Python/comp.lang.python/2004-06/3823.html

+1

I get it all the time too

Something new about this problem?
I reported it for the first time 4 months ago.
I think you should prioritize its solving: without a working debugger it's impossible to develop seriously.

The "broken pipe" problem doesn't occur anymore in my algorithm!

I had to many call to get_fundamentals(.) in before_trading_start(.), I limited them introducing a date logic (practically perform get_fundamentals(.) only once in month) and now the debugger works!

Maybe the reason of the "broken pipe" was to many data in memory to be serialized on the hd because of debugging modus.

Workaround:
If it is basically a timeout at some number of minutes (maybe built in to the browser, not sure), then ...
Suppose you have a breakpoint with a condition to break in if the date is near the end of a two year backtest, hitting 'Broken Pipe'.
Set another breakpoint on another line to break at around one half of that time, like str(get_datetime().date()) == '2014-05-01'.
When that first point is hit, it should reset the timer, to be able to make it to the next.

any updates on this?
still running into these at times.

I just ran into this issue as well while backtesting over 9 years. I needed to use the debugger to inspect a variable so I changed my backtest period to the three days I knew would populate the variable I wanted to check. That solved the problem so it does seem to be an issue with the debugger and length of backtest. Even with the backtest at three days though it hung up for almost two minutes before the breakpoint triggered and let me use the debugger which seemed odd since it was just a simple data.history call for 30 days of price data.

Just to bump up this post, the bug isn't fixed yet...

Also having this issue, and it really hurts development time of algorithms. Seems to happen with a test that has a long back-test period, or simply takes a long time.

Nope. Not fixed at all:/

@Danny Zilberg could you elaborate on the issues you are seeing. This thread is quite old and much has changed with the code and the platform. This also is not an issue that get's reported to our support team, so other than the replies above there isn't a lot for our engineering team to go on. Do you have an algo you could attach? Thanks!

Hi sorry for the late reply. I got ahead in my algorithm and thought it was an internet issue.
Now I've encountered the same problem. If I set the breakpoint at a certain date, way ahead in the backtest, the broken pipe error comes up. If not, the algorithm runs untill it reaches the error that I am trying to find.
For some reason I'm not able to attach that specific algorithm (it only shows 5 of my algorithms)
Sorry and thanks