Synchronise language service project updates

[drewnoakes]

Ensures a workspace will not be disposed during an update by wrapping the update in ExecuteUnderLockAsync.

Diff looks worse than it is. Recommend viewing without white space changes.

Microsoft Reviewers: Open in CodeFlow

[tmeschter]

Looks good, but I think we need to make the same change to the WriteAsync methods.

[jasonmalinowski]

@drewnoakes @tmeschter So this meant you were potentially updating a project while it was being disposed at the same time? I wonder if that explains a bunch of bugs I've been seeing that I haven't been able to track down....

[tmeschter]

@jasonmalinowski Possibly. More likely if the issues you were seeing involved ObjectDisposedExceptions.

[drewnoakes]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[MiYanni]

@drewnoakes We've been hitting concurrency issues with the agent pool all day. If you look at the graph here and add the currently running jobs + queued jobs, it'll be around 820, which is the max for the agent pool. We'll just have to wait.
https://dev.azure.com/dnceng-public/public/_settings/agentqueues?queueId=13&view=analytics

[jasonmalinowski]

@jasonmalinowski Possibly. More likely if the issues you were seeing involved ObjectDisposedExceptions.

Nothing specifically with object disposed exceptions, but we were seeing bugs where we'd potentially try creating references to projects that are missing, or things like that. If it's possible you were removing a project from the workspace and also adding/removing output paths for it at the same time, we might have gotten into those bad states.

[drewnoakes]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[drewnoakes]

@drewnoakes We've been hitting concurrency issues with the agent pool all day. If you look at the graph here and add the currently running jobs + queued jobs, it'll be around 820, which is the max for the agent pool. We'll just have to wait. https://dev.azure.com/dnceng-public/public/_settings/agentqueues?queueId=13&view=analytics

I think the failure is actually related to this PR somehow. Other PRs are building. I will investigate later as it's late here now.

FYI @tmeschter

[tmeschter]

@drewnoakes A quick review of the build output does look like it could be hanging in the unit test run. :-(