fix: add tsx as devDependency to prevent E2E test race condition#12
Merged
fix: add tsx as devDependency to prevent E2E test race condition#12
Conversation
When 16 E2E tests run in parallel, each invokes 'npx tsx' to start the test server. Without tsx installed locally, all 16 processes race to download and install tsx into ~/.npm/_npx/ simultaneously, causing ENOTEMPTY, TAR_ENTRY_ERROR, and 'tsx: not found' failures. Adding tsx as a devDependency ensures it's available locally and npx resolves it instantly without any network/install race.
- bridge-resilience: use --isolated to prevent shared sessions.json pollution from parallel tests causing INVARIANT VIOLATION - header-security: replace hardcoded sleep with wait_for ping polling to ensure bridge is ready before assertions - proxy: replace all hardcoded sleeps with wait_for health endpoint polling; use --isolated to avoid session file contention - server-abort: add wait_for health check after server reset and ping polling after connect; use --isolated
The connect command was failing with exit code 1 when showServerDetails() couldn't reach the bridge fast enough after session creation. The session was already created successfully — the bridge just needed more time to complete MCP initialization. Now showServerDetails is best-effort on connect: if it fails, the session is still valid and the next command will trigger ensureBridgeReady which handles retries properly.
The previous commit swallowed all errors from showServerDetails, including auth errors. This caused 'mcpc <oauth-server> connect @name' to succeed even when the server requires authentication. Auth errors must propagate so users get the login hint.
…errors ensureBridgeReady() now retries up to 3 times (1s apart) when the bridge socket is temporarily unresponsive, instead of immediately restarting the bridge. This prevents false-positive restarts caused by transient event loop stalls (common under CI load). Also adds wait_for ping readiness polling after connect in 3 E2E tests (header-security, proxy, server-abort) as defense-in-depth.
The retry logic in ensureBridgeReady() added up to 12s of extra latency per failed health check, causing cascading lock contention on the shared sessions.json across 16 parallel E2E tests. Revert to single-shot check; the wait_for ping additions in the test scripts are sufficient.
Member
|
Is this good to merge? |
commit 1e35f2f Author: Dušan Vystrčil <vystrcil.dusan@gmail.com> Date: Tue Mar 3 10:25:50 2026 +0100 fix: flush stdout before process.exit() to prevent pipe truncation (#17) When mcpc output exceeds 64KB and stdout is piped (e.g. `mcpc ... --json | jq`), the output was silently truncated at exactly 65,536 bytes (the kernel pipe buffer size), producing broken JSON. Root cause: when stdout is a pipe, Node.js uses async I/O. `process.exit(0)` terminates the process before buffered data is fully written to the pipe. When stdout is a file or TTY, Node.js uses sync I/O, so data is written before exit. The fix waits for stdout to drain before calling process.exit(0). Reproduction: mcpc @apify tools-call get-actor-output datasetId:="..." --json | wc -c Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> commit c7ea6f2 Author: Jan Curn <jan.curn@gmail.com> Date: Mon Mar 2 13:15:22 2026 +0100 Updated CHANGELOG.md, improved release process commit 3861f51 Author: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com> Date: Mon Mar 2 11:19:49 2026 +0100 Add renovate.json (#13) Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com> commit ba193b8 Author: Jan Curn <jan.curn@gmail.com> Date: Sun Mar 1 23:36:16 2026 +0100 v0.1.10 commit 8400ca3 Author: Jan Curn <jan.curn@gmail.com> Date: Sun Mar 1 23:21:57 2026 +0100 Added support for `HTTPS_PROXY` / `HTTP_PROXY` / `NO_PROXY` env vars (#10) * Added support for `HTTPS_PROXY` / `https_proxy` / `HTTP_PROXY` / `http_proxy` env vars * Improvements * Improvements * Fixes * Fixes * Nits
Collaborator
Author
|
some test are still failing in CICD, but let's merge it - it is in better state than it was before. @jancurn |
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
3 participants
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.
Summary
tsxas a devDependency to fix E2E test failures in CIProblem
All 16 E2E tests run in parallel and each invokes
npx tsxto start the test MCP server. Withouttsxinstalled locally, all 16 processes race to download and install it into~/.npm/_npx/simultaneously, causing:ENOTEMPTYerrors (directory not empty during npm install)TAR_ENTRY_ERROR(corrupted tar extraction)sh: 1: tsx: not found(install failed silently)This has been the root cause of all 19 failing E2E tests on CI across PRs.
Fix
Adding
tsxtodevDependenciesensures it's available locally afternpm install. Whennpx tsxruns, it resolves the local binary instantly with no network download or install step, eliminating the race condition entirely.