1. Marks all integration tests as flaky on the CI and outputs when
they're flaky so we can track it.
2. Uses a custom test runner for integration tests, giving us full
control over how they're run.
3. Runs watcher tests sequentially for more stability.
1. Moves `unit` and `unit_node` tests to use file_test_runner.
1. Marks all spec, unit, and node_unit tests as flaky on the CI and
outputs when they're flaky so we can track it.
1. Reduces CI concurrency to 1 on 3 flaky test failures (for now...
would be more ideal to lower it slowly maybe, which is easily possible
with this pr)
1. Adds a 100ms delay between flaky test failures.
## Summary
Aligns Deno's Node.js compatibility layer (`node:net`, `node:http`,
`node:https`,
`node:http2`, `node:dns`) with Node.js v18.4.0+ behavior by returning
the `family`
property as a string (`"IPv4"` or `"IPv6"`) rather than a number in
`server.address()`
and socket address methods.
Node.js briefly changed `family` from string to number in v18.0.0
(nodejs/node#41431),
but reverted in v18.4.0 (nodejs/node#43054) due to ecosystem breakage
(nodejs/node#43014).
This fix ensures compatibility with npm packages that rely on the string
format, which
has been the stable API since Node.js v18.4.0.
## Changes
- Modified `ext/node/polyfills/http.ts` to add `family` property to
`address()` return
- Modified `ext/node/polyfills/internal_binding/tcp_wrap.ts` to return
string `family`
instead of number in `getsockname()`, `getpeername()`, and `#connect()`
- Modified `ext/node/polyfills/net.ts` to fix `socket.remoteFamily`
getter (no longer
needs conversion since `family` is now a string)
- Modified `ext/node/polyfills/dns.ts` and
`ext/node/polyfills/internal/dns/promises.ts` to accept string family
values (`"IPv4"`,
`"IPv6"`) in `lookup()`, matching [Node.js
behavior](https://nodejs.org/api/dns.html#dnslookuphostname-options-callback)
- Added tests in `tests/unit_node/http_test.ts`,
`tests/unit_node/http2_test.ts`,
`tests/unit_node/https_test.ts`, `tests/unit_node/dns_test.ts`, and
`tests/unit_node/net_test.ts`
## Node.js Compatibility Note
For non-IP addresses (when `isIP()` returns 0), the `family` property is
`undefined`.
This matches Node.js C++ behavior in
[`AddressToJS`](https://github.com/nodejs/node/blob/main/src/tcp_wrap.cc)
where family
is only set for `AF_INET` (`"IPv4"`) and `AF_INET6` (`"IPv6"`), and left
undefined
otherwise.
## Refs
- nodejs/node#43054
- nodejs/node@70b516e
<!--
Before submitting a PR, please read
https://docs.deno.com/runtime/manual/references/contributing
1. Give the PR a descriptive title.
Examples of good title:
- fix(std/http): Fix race condition in server
- docs(console): Update docstrings
- feat(doc): Handle nested reexports
Examples of bad title:
- fix#7123
- update docs
- fix bugs
2. Ensure there is a related issue and it is referenced in the PR text.
3. Ensure there are tests that cover the changes.
4. Ensure `cargo test` passes.
5. Ensure `./tools/format.js` passes without changing files.
6. Ensure `./tools/lint.js` passes.
7. Open as a draft PR if your work is still in progress. The CI won't
run
all steps, but you can add '[ci]' to a commit message to force it to.
8. If you would like to run the benchmarks on the CI, add the 'ci-bench'
label.
-->
---------
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Daniel Rahmanto <daniel.rahmanto@gmail.com>
## Summary
This PR fixes the behavior of `url.domainToASCII` to match Node.js by
returning an empty string when an invalid domain is passed, instead of
throwing an error.
## Changes
- Modified `op_node_idna_domain_to_ascii` in `ext/node/ops/idna.rs` to
return an empty string on error instead of propagating the error
- Updated documentation in `ext/node/polyfills/internal/idna.ts` to
clarify this behavior
- Added spec test in `tests/specs/node/url_domain_to_ascii/` to verify
the fix
## Test Plan
### Manual Testing
Tested with the example from the issue:
```javascript
import url from 'node:url';
console.log(url.domainToASCII('xn--iñvalid.com')); // Returns empty string instead of throwing
```
Also tested with various valid and invalid domains to ensure
compatibility:
- Valid domains: `example.com` → `example.com`
- Valid unicode: `münchen.de` → `xn--mnchen-3ya.de`
- Invalid punycode: `xn--iñvalid.com` → `` (empty string)
- Invalid domain: `xn--` → `` (empty string)
### Automated Tests
Added spec test that verifies the behavior matches Node.js for both
valid and invalid domains.
## Closes
Fixes#31133
---
_This PR was autogenerated and may require review and adjustments._
🤖 Generated with [Claude Code](https://claude.com/claude-code)
---------
Co-authored-by: claude[bot] <41898282+claude[bot]@users.noreply.github.com>
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Daniel Rahmanto <daniel.rahmanto@gmail.com>
## Summary
In Node.js, `tty.ReadStream` sets `isTTY` as a simple instance property
(`this.isTTY = true`), making it naturally writable. Deno's polyfill
used a getter-only property definition, preventing assignment.
Some npm packages need to override `isTTY` at runtime. For example,
Playwright's test server sets `process.stdin.isTTY = undefined` to
prevent reporters from blocking on user input during UI mode
(microsoft/playwright#37867).
This aligns stdin with the stdout/stderr fix from #26130, ensuring all
stdio streams behave consistently with Node.js.
## Refs
-
6a1a3ba045/lib/tty.js (L71)
-
3a5a32d26d/packages/playwright/src/runner/testServer.ts (L257-L266)
Co-authored-by: Claude <noreply@anthropic.com>
Closes#26300
Allows the following node compatibility tests to pass:
-
https://github.com/nodejs/node/blob/v24.2.0/test/parallel/test-dns-lookupService-promises.js
-
https://github.com/nodejs/node/blob/v24.2.0/test/parallel/test-dns-lookupService.js
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Added `dns.lookupService()` for reverse DNS lookups, supporting both
callback and Promise-based APIs.
* Returns hostname and service information for a given IP address and
port combination.
* **Tests**
* Added unit tests for new DNS lookupService functionality across
callback and Promise variants.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Previously, running this code causes segfault:
```ts
import { DatabaseSync } from "node:sqlite";
const db = new DatabaseSync(':memory:');
const stmt = db.prepare('SELECT 1 AS value');
db.close();
stmt.get(); // segmentation fault (core dumped)
```
Changes in this PR also allows the
https://github.com/nodejs/node/blob/v24.2.0/test/parallel/test-sqlite-statement-sync-columns.js
test to pass.
Fixes#31172
## Description
The `deepStrictEqual` function was using `asserts.equal()` which doesn't
properly handle boxed primitives like Number objects. Changed it to use
`isDeepStrictEqual()` from `comparisons.ts` which correctly handles
Number, String, Boolean, BigInt, and Symbol objects.
## Changes
- Modified `ext/node/polyfills/assert.ts`:
- Added import for `isDeepStrictEqual` from
`ext:deno_node/internal/util/comparisons.ts`
- Updated `deepStrictEqual` function to use `isDeepStrictEqual()`
instead of `asserts.equal()`
- Added test cases in `tests/unit_node/assert_test.ts`:
- Test that `deepStrictEqual` throws `AssertionError` for different
Number objects
- Test that `deepStrictEqual` passes for equal Number objects
## Testing
Added unit tests to verify:
- `deepStrictEqual` throws `AssertionError` for different Number objects
(e.g., `new Number(1)` vs `new Number(2)`)
- `deepStrictEqual` passes for equal Number objects (e.g., `new
Number(1)` vs `new Number(1)`)
## Related Issue
#31172 - `assert.deepStrictEqual` does not throw exception for Number
objects
---------
Co-authored-by: Daniel Rahmanto <daniel.rahmanto@gmail.com>
Introduces the emitWarning function from the process module to enable
warning capabilities within the console constructor implementation.
This fixes a bug where the `time`, `countReset` and `timeLogImpl`
functions would throw an error due to trying to call `emitWarning`, they
would throw an error for an undefined variable named `emitWarning` or if
a global function had that name` they would call that function, which is
not expected behaviour here.
---------
Signed-off-by: Jake Champion <me@jakechampion.name>
Closes#30013
The `'exit'` event in `node:worker_threads` was not consistently fired,
or in some cases emitted multiple times, diverging from Node.js
behavior.
Fixes a panic that occurred when using nested session.post() calls in
the node:inspector module. The issue manifested when callbacks from
inspector commands (like Profiler.enable) called session.post() again,
causing a re-entrant invocation of op_inspector_dispatch.
Fixes https://github.com/denoland/deno/issues/31020
Towards #29972
- `fs.read` and `fs.readSync`: accepts object option parameter.
- `fs.write` and `fs.writeSync`: negative position value defaults to the
current seek position.
- Allows several node compatibility tests to pass.
Closes#30972
Other changes:
- Re-enables ignored tests on `tests/unit_node/_fs/_fs_write_test.ts`
- Ported the polyfill to `.ts`, though defining the types are tricky
since our implementation (ported from node) mutates a lot of variables.
Discovered on #30905, though not directly addressing the main issue
Given this code:
```ts
console.log(globalThis.process.env)
```
Deno 2.5.4:
```bash
➜ deno index.ts
✅ Granted env access to "Symbol(Deno.customInspect)".
✅ Granted env access to "Symbol(Deno.privateCustomInspect)".
✅ Granted env access to "Symbol(nodejs.util.inspect.custom)".
✅ Granted env access to "constructor".
✅ Granted env access to "Symbol(Symbol.iterator)".
✅ Granted env access.
[..result..]
```
This PR:
```bash
➜ deno index.ts
✅ Granted env access to "constructor".
✅ Granted env access.
[..result..]
```
Futhermore, it allows the following tests to pass:
-
https://github.com/nodejs/node/blob/v24.2.0/test/parallel/test-process-env-deprecation.js
-
https://github.com/nodejs/node/blob/v24.2.0/test/parallel/test-process-env-ignore-getter-setter.js
Closes#30936
The `compressIPv6` function is causing this issue. Maybe it was used in
the past to compress IPv6 address returned from the op function, though
I might be wrong about this.
Both `Deno.resolveDns` and the `resolve6` polyfill use `op_dns_resolve`
op function, which returns a result that already matches with the
Node.js' `resolve6`. The incorrect result is due to our `resolve6`
polyfill applying the `compressIPv6` function to the result.
Fixes#30829
## Problem
The zlib polyfill had recursive calls between `processCallback` and
`handle.write()` that caused stack overflow when decompressing large
data.
## Solution
Defer `handle.write()` calls with `process.nextTick()` to break the
synchronous call chain.
## Changes
- `ext/node/polyfills/zlib.js`: Wrapped `handle.write()` in
`process.nextTick()`
- `tests/unit_node/zlib_test.ts`: Added test for 64MiB data
decompression
Fixes#30722
The root cause comes from the `.update()` and `.final()` methods of
`Cipheriv` and `Decipheriv`, which internally calls `.toString()` method
of Buffer. Doing so adds padding to make the output length a multiple of
4, which we don't want for this case. This PR replaces that to use
`StringDecoder` instead, making it more aligned with the Node.js
implementation.
clean up some outdated code and implement a few performance
improvements. most notably, the response from `upgradeWebSocket` is now
actually used as a normal response.
a followup pr will rework raw upgrades to a greater extent
