-
Alexandre Duret-Lutz authored
Newer Jupyter version are able to capture the system's stdout and stderr to display it in the notebook. This is done asynchronously, with a thread polling those file descriptor. While this will help us debug (finaly we can see the tracing code we put in C++) this causes two issues for testing. One is the asynchronous behaviour, which makes it very hard to reproduce notebooks. The second issue is that older version of Jupyter used to hide some of the prints from the notebook, so it is hard to accommodate both. In the case of the ltsmin-pml notebook, loading the PML file from a filename used to trigger a compilation silently (with output on the console, but not in the notebook). The newer version had the output of that compilation spread into two cells. * python/spot/ltsmin.i: Work around the issue by triggering the compilation from Python, and capturing its output explicitly, so it work with all Jupyter versions. Also adjust to use the more recent and simpler subprocess.run() interface, available since Python 3.5. * tests/python/ltsmin-pml.ipynb: Adjust expected output. * tests/python/ipnbdoctest.py (canonicalize): Adjust patterns.
186d2063