This repository was archived by the owner on Apr 14, 2022. It is now read-only.
Repository navigation
This repository was archived by the owner on Apr 14, 2022. It is now read-only.
Server remains open for requests after unhandled "noncritical" exceptions #448
Description
Activity
The IsCriticalException call is also strange, since it's defined as:
I almost feel like all exceptions other than a cancellation should be rethrown. And, if we're even able to catch any of these "critical" exceptions, there's no reason to not at least make an attempt to record something. Special casing a list of essentially uncatchable exceptions to choose to not catch seems strange.
IsCriticalmeans unable to recover, even unable to send telemetry or write logs, i.e. unconditional shutdown.We should report exception through telemetry and re-throw, letting process crash so it can be orderly restarted by the client. Otherwise client will be talking to the dead server.
Most probably will fix #452 and friends
Fixed in #498.
https://github.com/Microsoft/python-language-server/blob/4a00ee80cb2d7b3e736329e82b2ed441ae70b5ca/src/Analysis/Engine/Impl/Intellisense/AnalysisQueue.cs#L60-L71
When an unhandled exception occurs, and it's deemed "noncritical", the error is logged and then the AnalysisQueue gets disposed. However, the users of the queue have no idea this is happening and will continue to use it. Specifically, the language server will get requests which only lead to ObjectDisposedException.
I'm not sure what the correct behavior is here, but if an exception occurs and we are choosing to dispose of the queue, then I feel like either a new queue needs to be created or the server needs to shut down.
#446 related.