Repository navigation
mcp: make waiting for the ioConn write lock cancelable - #1499
Merged
guglielmo-san merged 3 commits intoNov 6, 2026
Merged
guglielmo-san merged 3 commits into
guglielmo-san merged 3 commits into
Conversation
ioConn.Write serialized writers with a mutex and checked ctx only before waiting for it, so a write whose context was canceled while another write was in progress still waited for that write and was then sent anyway. A write blocked on a peer that stopped reading could not be interrupted at all. Queue writers on a one-slot channel so a canceled waiter returns without sending. Canceling the active write closes the connection: bytes already handed to the stream cannot be withdrawn, and a partial frame would corrupt every later message. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
github-actions
Bot
added
the
missing-issue-link
github-actions
Bot
removed
the
missing-issue-link
Contributor
|
@Tangerg let's scope this PR only to the blocking case. Exclude for now the case of the cancellation of on going call |
As discussed in review, this change no longer interrupts a write that is already in progress by closing the connection. It only stops a write from waiting for an earlier one once its context is done. writeMu keeps its name and becomes a 1-element semaphore channel, like reqHeaderMu in the net/http HTTP/2 client. A canceled waiter returns ctx.Err(), which jsonrpc2 already treats as a failure of the call rather than of the connection. The regression test uses testing/synctest instead of sleeping. Updates modelcontextprotocol#1498 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Tangerg
changed the title
Contributor
Author
|
@guglielmo-san thanks, narrowed in a78ab1f. A write canceled while it waits for another write now returns AI assistance: prepared with Claude; I reviewed it and can explain it. |
5 tasks done
guglielmo-san
approved these changes
Nov 6, 2026
Tangerg
added a commit
to Tangerg/go-sdk
that referenced
this pull request
Nov 6, 2026
Sync 6 upstream commits through df0f295: the new ParseURLElicitationRequiredError (modelcontextprotocol#1495), the tools/call retry on HeaderMismatch (modelcontextprotocol#1502), notifications without Mcp-Method (modelcontextprotocol#1504), the cancelable ioConn write lock (modelcontextprotocol#1499), invalid_token in 401 challenges (modelcontextprotocol#1507) and the ioConn frame cap that no longer depends on read boundaries (modelcontextprotocol#1510). modelcontextprotocol#1499 and modelcontextprotocol#1510 are the upstream form of the fork's transport fixes, so ioConn now matches upstream except that canceling a write already in progress still closes the connection, because the peer may hold part of the frame. Upstream's tests replace the fork's duplicates of them; the fork keeps its test for the in-progress cancellation. New pkg.go.dev links in the docs use the fork's module path.
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
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.
ioConn.Writechecked its context only before waiting forwriteMu, so a write canceled while another write was blocked kept waiting, and was sent once the lock was released.writeMuis now a 1-element semaphore channel, likereqHeaderMuin the net/http HTTP/2 client, so a waiting write also selects onctx.Done()and returnsctx.Err()without sending. jsonrpc2 already treats that error as a failure of the call rather than of the connection. Interrupting a write that is already in progress is left out, as discussed in review.Without the change, the new test hangs with the canceled write blocked in
sync.Mutex.Lock.Updates #1498
AI assistance: prepared with Claude; I reviewed the change and can explain it.