-
-
Notifications
You must be signed in to change notification settings - Fork 43
Refreshing when stopped on a breakpoint #220
Comments
It's also worth keeping an eye on the kernel and session architecture refactoring work happening in JupyterLab: jupyterlab/jupyterlab#7252 For reference this is where the |
It seems that actaully the workaround is not possible since the web socket that allows to communicate with the kernel is not created until a first web socket receives a Therefore the only mean to fix that is to send the |
We discussed another way to detect whether a kernel supports debugging without having to rely on the debug info reply. One workaround could be to send one of the debug requests, for example One possible issue with this is that other kernels such as Also we currently use the Line 197 in dc94d28
Instead we could check whether the kernel is available with Finally there is an issue about sending the |
Moving to 0.2.0 to check if this is a viable workaround. |
As pointed out by @jtpio in #217, refreshing the page when the code execution is stopped on a breakpoint does not work. The reason is that after refreshing, the client sends a
kernel_info_request
to the kernel and waits for its reply. This request is sent on the shell, however the thread executing on the shell inxeus_python
is stopped because of the previous breakpoint. Therefore, the client never gets its answer and loops on creating web sockets.We can workaround this issue by restarting the debugger when refreshing the page. This would resume the execution of the thread on the
shell
without loosing the breakpoints.A lont-term solution would be to have the
kernel_info_request
sent on thecontrol
instead of having it on theshell
but that would require changes in Jupyter Lab core.cc @afshin @KsavinN
The text was updated successfully, but these errors were encountered: