Repository navigation
feat: chain the caller's EXIT trap into server teardown - #26
Merged
Merged
Conversation
run_mcp_server captures the EXIT handler it displaces and _server_teardown runs it after the SDK's own cleanup, on clean exit and trapped signals alike. Direct-call shape only; a subshell-isolated call keeps today's behavior. Handler stdout goes to stderr, a handler failure is logged rather than fatal. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…din reads Review findings: a posix-mode unset trap reported as - was eval'd; the handler could stall shutdown, so it now gets the cancel hook's grace and a /dev/null stdin; a signal recorded during any teardown pass re-raises; the capture drops its temp-file round-trip for the plain substitution. Suites share a server-script writer in tests/test_helper. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Cross-call tool state travels through a server-owned exported file; the section states the $$/BASHPID distinction and the detached-tool caveat. The shutdown section carries the chained-handler rules and the corrected cost of the subshell wrap. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Martin Bens (SpiGAndromeda)
added a commit
that referenced
this pull request
Sep 15, 2026
Classifies the chained EXIT handler from #26 as minor. The contract's major triggers are a renamed or resignatured function, a change to what a function writes to stdout, and a tightened validator; the chaining is none of them. A consumer's EXIT handler that `run_mcp_server` previously dropped now runs during teardown, so a consumer whose handler was silently discarded sees it execute after the upgrade. The entry also records the handler's stdin and subshell rules, which README.md and docs/architecture.md already state. Co-Authored-By: Claude <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #25.
run_mcp_serverreplaced any EXIT trap a consumer had installed, so a server script's own cleanup never ran in the direct-call shape. The server now captures the handler it displaces and_server_teardownruns it after the SDK's own cleanup, on a clean exit and on a trapped signal alike. A subshell-isolated call keeps its old behavior: the parent's trap stays the parent's.The handler's contract, documented in README §Cancelling and shutting down:
/dev/nullrun_mcp_serverreturns on a clean exitrun_mcp_serverdirectly is chainedAlso in this change: a posix-mode unset trap (
trap -- - EXIT) is never eval'd as a handler; a signal recorded during any teardown pass is re-raised at the pass's end, so the shell dies by the signal on the clean-exit route too; and the README states the subshell wrap's cost precisely — a pid-only signal loses the SDK teardown, while the parent's EXIT trap still runs.