Prevent stdio connection hang for missing server path.

[TimChild]

• Ensure that read/write streams are always closed (even if it is the process that ends first)
• Check whether the stdio process ended prematurely
• Raise an error if that was the case

Motivation and Context

The stdio connection was hanging indefinitely for a bad server file path (e.g. command uv, args run non-existing-file.py).
I think because the read/write streams were not being closed properly in that case.

This makes it so that errors due to the underlying mcp server stopping early raises a ProcessTerminatedEarlyError.

How Has This Been Tested?

Added a test (including catching if it does hang for more than 1s)

Breaking Changes

Should not be a breaking change.

Types of changes

• [ x ] Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Documentation update

Checklist

• [ x ] I have read the MCP Documentation
• [ x ] My code follows the repository's style guidelines
• [ x ] New and existing tests pass locally
• [ x ] I have added appropriate error handling
• [ x ] I have added or updated documentation as needed -- Did not update docs (but error is self-explanatory)

Additional context

[ihrpr]

Thank you for your contribution! Sorry for the delay in reviews.

This PR identified an issue where the SDK is not handling missing server paths. I'd suggest a slightly different approach for a solution though. Should we implement error handling for the server path instead?

[TimChild]

No problem!
It's been a while since I thought about this. Definitely checking the path first is an easier and more efficient way to catch the basic problem of a missing server file, but I don't remember if I was only using that as the simplest example of a faulty server.

I'm trying to remember if I had issues when I was developing an mcp server, and had some bug in the connection procedure or something and then that was causing the client to hang indefinitely.

For now, it probably makes sense to do a simpler check that the server file as you suggest, and then worry about more complicated cases if that is still an issue.

About to fly right now, and won't have a chance to come back to this for a little while. I'm happy for you to make any changes you want, otherwise I'll hopefully get to it in a few weeks.

[ihrpr]

Thank you @TimChild, I think #333 might actually resolve this as well, but we need to use the test you have here. I'll take a look!

[bhosmer-ant]

LGTM!

stale