fix: Support IPv6 bind addresses in http_server helper #5239

I501307 · 2026-01-27T20:57:27Z

Which issue(s) this PR fixes:
Fixes # (if there's an issue, or remove this line)

What this PR does / why we need it:
Fixes URI::InvalidURIError when binding to IPv6 addresses. The http_server
helper was generating invalid URIs like 'http://:::24231' instead of the
RFC 3986 compliant 'http://[::]:24231'.

Release Note:
http_server helper: Fix IPv6 bind address support

Fixes URI::InvalidURIError when binding to IPv6 addresses like '::' or '::1'. The http_server helper was generating invalid URIs such as 'http://:::24231' instead of the RFC 3986 compliant 'http://[::]:24231'. Changes: - Wrap IPv6 addresses in brackets when constructing URIs - Add unit tests for IPv6 localhost (::1) and wildcard (::) binding - Update CHANGELOG.md This bug affected all Fluentd versions including v1.19.1 and caused crashes when attempting to bind HTTP servers to IPv6 addresses in dual-stack environments. Resolves: Invalid URI generation for IPv6 bind addresses Signed-off-by: Jesse Awan <jesse.awan@sap.com>

I501307 · 2026-01-27T21:04:00Z

Could a maintainer please add the bug label? This fixes a crash affecting all Fluentd versions when binding to IPv6 addresses.

lib/fluent/plugin_helper/http_server/server.rb

- Updated server.rb to check if IPv6 addresses are already bracketed - Added tests for pre-bracketed addresses ([::1] and [::]) - Improved ipv6_enabled? helper to verify both binding and resolution - Updated CHANGELOG to document pre-bracketed address handling Signed-off-by: Jesse Awan <jesse.awan@sap.com>

I501307 · 2026-01-28T12:43:41Z

@Watson1978 Same issue I encountered within a plugin as well here :
https://github.com/fluent/fluent-plugin-prometheus/pull/240/changes

CHANGELOG.md

Signed-off-by: Jesse Awan <jesse.awan@sap.com>

Watson1978 · 2026-02-03T05:43:02Z

@I501307 Seems it fails the test in CI. Can you take a look at it?

Signed-off-by: Jesse Awan <jesse.awan@sap.com>

I501307 · 2026-02-03T11:02:58Z

@I501307 Seems it fails the test in CI. Can you take a look at it?

Root Cause:
The original code didn't handle pre-bracketed IPv6 addresses like [::1] or [::]. When passed to Async::HTTP::Endpoint, socket binding failed on Unix systems because brackets are only for URI formatting, not socket binding.

The Fix (2 parts):

server.rb - Normalize addresses on initialization

Strip brackets from input addresses (e.g., [::1] → ::1)
Store unbracketed address for socket binding
Add brackets only when constructing URIs

test_http_server_helper.rb - Replace flaky binding tests

Before: Tests tried to bind to IPv6 addresses → failed on CI due to IPv6 networking issues
After: Tests only validate formatting logic → no networking required, works everywhere
Verified locally:

All 19 http_server tests pass (0 errors, 0 failures)
Full test suite: 811+ tests passing before completion
The new tests should pass on all platforms (Ubuntu, macOS, Windows) across Ruby 3.2-4.0. 🚀

I501307 · 2026-02-04T15:02:14Z

@I501307 Seems it fails the test in CI. Can you take a look at it?

@Watson1978 The Windows failures are unrelated to this PR - they're in test_out_exec_filter.rb
(expecting "5500" but getting "4184"), not in the http_server_helper tests.

All http_server_helper tests pass on all platforms. Can a maintainer re-run
the Windows CI or acknowledge these are known flaky tests?

kenhys

Thanks @I501307 !!

If you afford to write, it might be safer to write with conditional networking test case additionally.

test ... do
  omit ""  unless ipv6_enabled?
  ...
end

lib/fluent/plugin_helper/http_server/server.rb

test/plugin_helper/test_http_server_helper.rb

- Rename addr_display to ip_literal per RFC 3986 terminology - Move ipv6_enabled? helper to test/helper.rb for reusability - Improve IPv6 detection with proper socket binding and address resolution checks Signed-off-by: Jesse Awan <jesse.awan@sap.com>

I501307 · 2026-02-05T15:58:17Z

Thanks @I501307 !!

If you afford to write, it might be safer to write with conditional networking test case additionally.
test ... do
  omit ""  unless ipv6_enabled?
  ...
end

@kenhys Done! I've renamed the variable to ip_literal and moved ipv6_enabled? to test/helper.rb with improved IPv6 detection logic. Thanks for the review feedback!

Watson1978

Thanks 👍🏻

Watson1978 · 2026-02-06T01:27:00Z

I've tried to fix load error of ostruct at #5249

I501307 · 2026-02-06T01:36:09Z

I've tried to fix load error of ostruct at #5249

ah great.
Ruby 4.0 moved ostruct out of default gems, and filter_record_transformer.rb is trying to load it.

kenhys

After: Tests only validate formatting logic → no networking required, works everywhere
Verified locally:

Thanks!

It works great without network, but if plugin helper implementation was changed (in the future), also
need to follow test code too.
If we forgot to follow (it might be accidentally happen), it becomes meaningless test cases there.
So I have concern about extracting internal logic directly to test code.
What do you think? @Watson1978

Watson1978 · 2026-02-06T06:06:31Z

What do you think? @Watson1978

Sorry, I didn't get to look closely at the test code.

Indeed.

It should not only test the logic itself.
it should actually run the http_server helper to verify it behaves as expected.

I need more review for test codes

Signed-off-by: Jesse Awan <jesse.awan@sap.com>

I501307 · 2026-02-06T09:24:14Z

What do you think? @Watson1978

Sorry, I didn't get to look closely at the test code.

Indeed.

It should not only test the logic itself. it should actually run the http_server helper to verify it behaves as expected.

Integration tests are now added https://github.com/fluent/fluentd/pull/5239/changes#diff-0733209104ce038bd67b2898b481ab3b6ff2634063b71948ae8a230bb3d5bcf3R23-R29 🌻

I501307 · 2026-02-06T09:24:40Z

After: Tests only validate formatting logic → no networking required, works everywhere
Verified locally:

Thanks!

It works great without network, but if plugin helper implementation was changed (in the future), also need to follow test code too. If we forgot to follow (it might be accidentally happen), it becomes meaningless test cases there. So I have concern about extracting internal logic directly to test code. What do you think? @Watson1978

done here https://github.com/fluent/fluentd/pull/5239/changes#diff-0733209104ce038bd67b2898b481ab3b6ff2634063b71948ae8a230bb3d5bcf3R23-R29 🌻

I501307 · 2026-02-06T09:25:45Z

@kenhys Please let me know if more tests are required.

test/plugin_helper/test_http_server_helper.rb

… tests Signed-off-by: Jesse Awan <jesse.awan@sap.com>

Watson1978

Thanks

I501307 · 2026-02-06T12:33:26Z

@kenhys we good to go 🚀

kenhys · 2026-02-09T05:12:40Z

test/plugin_helper/test_http_server_helper.rb

      end
    end
+
+    test 'http_server can bind and serve on IPv6 loopback ::1' do


Thanks.

Could you use data(...) test pattern to reduce redundant test code? (something like)

data( 'ipv6 loopback test' => [:ipv6_loopback_test, '::1'], 'ipv6 any test' => [:ipv6_any_test, '::'] ... ) test 'http_server can bind and serve on IPv6 address' do |test_case, addr| ... end

If above fix was done, LGTM.

ah yeah 🌻. Now tests look better
1b21008

Maybe you can trigger the tests and we are gucci.

Reduce code duplication by combining three separate IPv6 test methods into a single parametrized test using the data() pattern as suggested by reviewer. Signed-off-by: Jesse Awan <jesse.awan@sap.com>

I501307 force-pushed the fix/http-server-ipv6-uri-support branch from a279091 to eb72e2a Compare January 27, 2026 21:00

Watson1978 reviewed Jan 28, 2026

View reviewed changes

lib/fluent/plugin_helper/http_server/server.rb Outdated Show resolved Hide resolved

I501307 requested a review from Watson1978 January 28, 2026 09:52

kenhys reviewed Feb 2, 2026

View reviewed changes

CHANGELOG.md Show resolved Hide resolved

chore: Update changelog version to v1.19.2

77b380a

Signed-off-by: Jesse Awan <jesse.awan@sap.com>

I501307 force-pushed the fix/http-server-ipv6-uri-support branch from bca5979 to 77b380a Compare February 2, 2026 11:11

I501307 requested a review from kenhys February 2, 2026 11:12

fix: Normalize IPv6 addresses for socket binding

22e03d2

Signed-off-by: Jesse Awan <jesse.awan@sap.com>

I501307 force-pushed the fix/http-server-ipv6-uri-support branch from 35a521a to 22e03d2 Compare February 3, 2026 10:44

Merge branch 'master' into fix/http-server-ipv6-uri-support

a01f67d

kenhys reviewed Feb 5, 2026

View reviewed changes

lib/fluent/plugin_helper/http_server/server.rb Outdated Show resolved Hide resolved

test/plugin_helper/test_http_server_helper.rb Outdated Show resolved Hide resolved

I501307 force-pushed the fix/http-server-ipv6-uri-support branch from b703cbd to f6d21ff Compare February 5, 2026 15:56

I501307 requested a review from kenhys February 5, 2026 15:59

Watson1978 previously approved these changes Feb 6, 2026

View reviewed changes

kenhys requested changes Feb 6, 2026

View reviewed changes

Watson1978 self-requested a review February 6, 2026 06:06

tests: Add IPv6 integration tests and improve ipv6_enabled? detection

835b99e

Signed-off-by: Jesse Awan <jesse.awan@sap.com>

Remove duplicate ipv6_enabled? - use global helper from test/helper.rb

69fdc14

Signed-off-by: Jesse Awan <jesse.awan@sap.com>

I501307 requested a review from kenhys February 6, 2026 09:24

Watson1978 reviewed Feb 6, 2026

View reviewed changes

test/plugin_helper/test_http_server_helper.rb Outdated Show resolved Hide resolved

Remove redundant IPv6 formatting logic tests - covered by integration…

046cc25

… tests Signed-off-by: Jesse Awan <jesse.awan@sap.com>

I501307 force-pushed the fix/http-server-ipv6-uri-support branch from 0280d82 to 046cc25 Compare February 6, 2026 09:40

Watson1978 previously approved these changes Feb 6, 2026

View reviewed changes

Watson1978 added this to the v1.20.0 milestone Feb 6, 2026

I501307 requested a review from Watson1978 February 6, 2026 12:32

kenhys requested changes Feb 9, 2026

View reviewed changes

test: Refactor IPv6 tests using data-driven pattern

1b21008

Reduce code duplication by combining three separate IPv6 test methods into a single parametrized test using the data() pattern as suggested by reviewer. Signed-off-by: Jesse Awan <jesse.awan@sap.com>

I501307 dismissed Watson1978’s stale review via 1b21008 February 9, 2026 18:03

I501307 requested a review from kenhys February 9, 2026 18:08

I501307 mentioned this pull request Feb 9, 2026

fix: Add IPv6 support for bind parameter fluent/fluent-plugin-prometheus#240

Open

fix: Support IPv6 bind addresses in http_server helper #5239

Are you sure you want to change the base?

fix: Support IPv6 bind addresses in http_server helper #5239

Uh oh!

Conversation

I501307 commented Jan 27, 2026

Uh oh!

I501307 commented Jan 27, 2026

Uh oh!

Uh oh!

I501307 commented Jan 28, 2026

Uh oh!

Uh oh!

Watson1978 commented Feb 3, 2026

Uh oh!

I501307 commented Feb 3, 2026

Uh oh!

I501307 commented Feb 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

kenhys left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

I501307 commented Feb 5, 2026

Uh oh!

Watson1978 left a comment

Choose a reason for hiding this comment

Uh oh!

Watson1978 commented Feb 6, 2026

Uh oh!

I501307 commented Feb 6, 2026

Uh oh!

kenhys left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Watson1978 commented Feb 6, 2026

Uh oh!

I501307 commented Feb 6, 2026

Uh oh!

I501307 commented Feb 6, 2026

Uh oh!

I501307 commented Feb 6, 2026

Uh oh!

Uh oh!

Watson1978 left a comment

Choose a reason for hiding this comment

Uh oh!

I501307 commented Feb 6, 2026

Uh oh!

kenhys Feb 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

kenhys Feb 9, 2026

Choose a reason for hiding this comment

Uh oh!

I501307 Feb 9, 2026

Choose a reason for hiding this comment

Uh oh!

I501307 Feb 9, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

I501307 commented Feb 4, 2026 •

edited

Loading

kenhys left a comment •

edited

Loading

kenhys left a comment •

edited

Loading

kenhys Feb 9, 2026 •

edited

Loading