no stacktrace printed in bokeh serve mode after import datashader

[oarcher]

While working on test code described on #959 (comment) , no stacktrace are printed in case of exception, making debug very difficult.

minimal code bterr.py

from holoviews.operation.datashader import datashade
a=1/0  # no bt printed, nothing happen in console

bokeh serve --port 5006 --address 0.0.0.0 --allow-websocket-origin='*' bterr.py

env:
bokeh 1.1.0 py37_0
holoviews 1.12.1 py_2
python 3.7.1 h0371630_7

[oarcher]

also tried with older packages versions, with same problem:

bokeh                     0.13.0                   py36_0  
holoviews                 1.8.4                    py36_1   
python                    3.6.8                h0371630_0 

[oarcher]

It seem that backtrace disappear after import xarray as xr in https://github.com/pyviz/holoviews/blob/6042f95f01a79fe46bbf0d1cd55f298587e90dd7/holoviews/operation/datashader.py#L9

with xarray=0.11.3, at https://github.com/pydata/xarray/blob/d58e0769c7d600d5c8c3c2fb52aa28856c4d3290/xarray/__init__.py#L22

But downgrading xarray to xarray=0.11.0 makes bt appear normaly ( xarray=0.11.2 broke it)

So minimal test code become :

import xarray
a=1/0  # no bt printed, nothing happen in bokeh serve console

[philippjfr]

That's truly bizarre. xarray must be messing with standard output and error streams somehow. Would be good to track this down and file an issue upstream.

[oarcher]

stderr is still ok, because

import xarray
import sys
print("this message is printed",file=sys.stderr)

works as expected.

I go further in tracing the problem, and located the pb in https://github.com/pydata/xarray/blob/624d78f48cfa54b8952219ecf37bcee4673da900/xarray/backends/locks.py#L12, so new test code become:

from dask.distributed import Lock as DistributedLock
a=1/0 # no bt printed in bokeh console

(with dask=1.2.0)

So i need advice to fill up a bug report upstream. Is it related to bokeh that is not reporting traceback in thread context, or xarray or dask ?

[philippjfr]

This is down to distributed and should be fixed in the next release.

[github-actions]

This issue has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.