test(bdd): add C++ SDK BDD tests for basic messaging by seokjin0414 · Pull Request #3554 · apache/iggy

seokjin0414 · 2026-06-23T09:50:27Z

What

Adds BDD coverage for the C++ SDK by implementing the shared
bdd/scenarios/basic_messaging.feature with cucumber-cpp and wiring it into the
Docker-based BDD harness and CI, the same way the other SDKs run. This is the
second part of #2965; the messaging FFI it relies on landed in #3046.

How it works

cucumber-cpp only supports the Cucumber wire protocol, so the step definitions
(bdd/cpp/features/step_definitions) build with CMake into a wire server, and a
Ruby Cucumber runner (cucumber + cucumber-wire, pinned to the versions
cucumber-cpp itself tests against) drives the feature file over the wire against a
real iggy-server. The step logic and assertions are all C++ and exercise the
SDK's FFI (get_streams, create_stream, create_topic, send_messages,
poll_messages). The Iggy C++ SDK is built with cargo and linked into the wire
server as a staticlib.

Heads-up for reviewers

Because cucumber-cpp is wire-protocol-only, the C++ BDD image pulls in a Ruby
runtime (cucumber + cucumber-wire). Flagging this explicitly in case it
changes how you'd like the harness shaped.

Notes

Everything lives under bdd/cpp, next to the other SDK BDD suites; foreign/cpp
is unchanged apart from the reserve() fix to the existing test.

The TearDown fixture and the magic-number cleanup discussed on Discord are tracked
separately by @slbotbm.

Refs #2965

slbotbm · 2026-06-23T10:14:01Z

@seokjin0414 the test should be in he root bdd/ folder, not inside of foreign/cpp

slbotbm · 2026-06-23T10:51:26Z

ignore the above, i didn't read the description😅

codecov · 2026-06-23T11:48:47Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 74.08%. Comparing base (7257940) to head (78c3ba4).

Additional details and impacted files

@@             Coverage Diff              @@
##             master    #3554      +/-   ##
============================================
- Coverage     74.10%   74.08%   -0.03%     
  Complexity      937      937              
============================================
  Files          1258     1258              
  Lines        131485   131485              
  Branches     107354   107398      +44     
============================================
- Hits          97435    97405      -30     
+ Misses        30963    30941      -22     
- Partials       3087     3139      +52

Components	Coverage Δ
Rust Core	`74.74% <ø> (+0.01%)`	⬆️
Java SDK	`62.44% <ø> (ø)`
C# SDK	`71.40% <ø> (-0.73%)`	⬇️
Python SDK	`88.88% <ø> (ø)`
PHP SDK	`84.29% <ø> (ø)`
Node SDK	`91.35% <ø> (ø)`
Go SDK	`40.14% <ø> (ø)`
see 33 files with indirect coverage changes

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

seokjin0414 · 2026-06-23T23:29:30Z

@slbotbm
no worries! ended up moving it to bdd/cpp anyway. i kept it out of foreign/cpp by building the wire server with CMake (FetchContent cucumber-cpp + the cxx staticlib) instead of Bazel, so it's self-contained like the other bdd dirs and
foreign/cpp now only carries the reserve() fix.
can switch it to a Bazel target under foreign/cpp if you'd prefer.

slbotbm · 2026-06-24T04:56:04Z

@seokjin0414 I didn't understand why did you use cmake instead of bazel. I think it would be possible to use bazel in bdd/cpp. you could define iggy_cpp as a bazel dependency in bdd/cpp/MODULE.bazel, and then locally define the path to iggy_cpp as ../../foreign/cpp.

slbotbm · 2026-06-24T17:56:55Z

/author

seokjin0414 · 2026-06-24T23:52:58Z

@slbotbm the cmake route was just to keep the test self-contained in bdd/cpp without touching foreign/cpp — i didn't realize local_path_override could pull the sdk in cleanly. it can, so i switched to bazel as you suggested: bdd/cpp is its own module now, cucumber-cpp from the registry + iggy_cpp via local_path_override(../../foreign/cpp). cmake is gone.

the only change to foreign/cpp is one line — making the iggy-cpp target public so the harness can depend on it. build + the full wire e2e (17 steps) pass.

slbotbm

Let's add a .clang-format as well that copies the one from foreign/cpp

You can use /ready when your PR is ready for review again

slbotbm · 2026-06-26T05:40:32Z

Let's also commit the ruby lockfile as well.

seokjin0414 · 2026-06-28T15:39:19Z

@slbotbm
also addressed the two non-inline ones: added bdd/cpp/.clang-format (copied from foreign/cpp), and committed bdd/cpp/Gemfile.lock for the ruby side. rebased onto master — CI is green.

seokjin0414 · 2026-06-28T15:45:08Z

/ready

seokjin0414 · 2026-06-30T23:30:11Z

/ready

slbotbm

Some comments, otherwise looks lgtm

slbotbm · 2026-07-01T14:57:15Z

+
+WORKDIR /workspace/bdd/cpp
+COPY bdd/cpp/Gemfile bdd/cpp/Gemfile.lock ./
+RUN gem install bundler --no-document \


Let's pin bundler to a specific version

pinned to 2.6.9 (matches BUNDLED WITH in Gemfile.lock).

slbotbm · 2026-07-01T15:13:03Z

+
+#include "world.hpp"
+
+using cucumber::ScenarioScope;


Let's remove this line and explicitly use cucumber::ScenarioScope whever it is called.

dropped the using alias in both step files; every call site now spells out cucumber::ScenarioScope.

slbotbm · 2026-07-01T15:23:47Z

+    EXPECT_EQ(streams.size(), static_cast<std::size_t>(0));
+}
+
+WHEN("^I create a stream with name \"(.*)\"$") {


Let's change the regex from (.*) to ([^"]{1,255}) to reflect actual stream name restrictions

Also do for topic as well.

tightened the name captures to ([^"]{1,255}) for both stream and topic (create + should-have-name steps).

slbotbm · 2026-07-01T15:27:30Z

+WHEN("^I create a topic with name \"(.*)\" in stream ([0-9]+) with ([0-9]+) partitions$") {
+    REGEX_PARAM(std::string, topic_name);
+    REGEX_PARAM(int, stream_id);
+    REGEX_PARAM(int, partitions_count);
+    ScenarioScope<bdd::GlobalContext> context;
+
+    context->client->create_topic(bdd::make_numeric_identifier(static_cast<std::uint32_t>(stream_id)), topic_name,
+                                  static_cast<std::uint32_t>(partitions_count), "none", 0, "never_expire", 0,
+                                  "server_default");


Instead of passing strings like "server_default" and "never_expire" here, I think we can use the functions and classes in iggy.hpp

can also be done for the rest of the functions. This will help us understand what the sdk is not providing right now.

switched create_topic to CompressionAlgorithm::none() / Expiry::never_expire() / MaxTopicSize::server_default(), and poll to PollingStrategy::offset(). the consumer kind and partitioning are still raw strings — iggy.hpp has no typed wrapper for those yet, so they mark exactly the gaps you mentioned.

slbotbm · 2026-07-01T15:37:11Z

+THEN("^all messages should be sent successfully$") {
+    ScenarioScope<bdd::GlobalContext> context;
+
+    EXPECT_GT(context->sent_count, static_cast<std::uint32_t>(0));


can we tighten this to expect the exact message count?

now asserts the exact count: the send step records how many messages it dispatched and this step checks sent == requested instead of just > 0.

The oversized-payload test grew its 64 MB buffer one byte at a time and a comment claimed cxx::Vec had no public reserve API. It does, so reserve the capacity once before filling it. Signed-off-by: seokjin0414 <sars21@hanmail.net>

The C++ SDK had no BDD coverage. Add cucumber-cpp step definitions under bdd/cpp that drive the shared basic_messaging.feature through the C++ FFI client. cucumber-cpp uses the wire protocol, so bdd/cpp is a self-contained Bazel module: it builds the step definitions into a wire server (pulling cucumber-cpp from the Bazel registry and the Iggy C++ SDK from foreign/cpp via local_path_override) that a Ruby Cucumber runner connects to. The SDK's iggy-cpp target is made public so the harness can depend on it. A per-file CUKE_OBJECT_PREFIX keeps the generated step classes from clashing across translation units. The Gemfile and cucumber.wire comment styles are registered with the license checker. Signed-off-by: seokjin0414 <sars21@hanmail.net>

A Ruby cucumber runner with cucumber-wire drives the wire server over the wire protocol against a real iggy-server. Register the cpp-bdd compose service, the run-bdd-tests entry, and the bdd-cpp CI component so it runs like the other SDK BDD suites. Signed-off-by: seokjin0414 <sars21@hanmail.net>

seokjin0414 · 2026-07-02T11:04:24Z

/ready

hubcio · 2026-07-02T12:19:12Z

+cd /workspace/bdd/cpp || exit 1
+
+set +e
+bundle exec cucumber


cucumber 7.1.0 defaults to non-strict, so an undefined step exits 0 - if the shared feature file ever gains a step without a matching C++ regex, this suite silently goes green. bundle exec cucumber --strict closes that.

fwiw rust (cucumber-rs reports an unmatched step as skipped -> exit 0) and go (godog, non-strict by default) have the same hole and would also pass silently - node/java/python/csharp runners are strict by default. so worth adding --strict here, and probably to rust/go too (separate PRs).

separately, the image bakes no .feature files and the copy on line 24 is masked with 2>/dev/null || true, so a missing mount means 0 scenarios and exit 0 even with --strict. worth a guard like ls features/*.feature >/dev/null || exit 1 before running. both fixes needed, they cover different holes.

hubcio · 2026-07-02T12:19:12Z

+
+bdd_wire_server &
+server_pid=$!
+sleep 2


fixed sleep 2 instead of waiting for port 3902. fails safe (connection refused = red), so worst case is a flake on a slow machine - a small wait-for-port loop would remove that.

hubcio · 2026-07-02T12:19:12Z

+THEN("^all messages should be sent successfully$") {
+    cucumber::ScenarioScope<bdd::GlobalContext> context;
+
+    EXPECT_EQ(context->sent_count, context->messages_to_send);


this assert can never fail - sent_count and messages_to_send are both derived from the same message_count in the WHEN step. the real signal is send_messages() throwing there. rust and python versions of this step are equally weak, so fine to keep for parity, but the sent_count/messages_to_send bookkeeping could be dropped.

hubcio · 2026-07-02T12:19:12Z

+// (rust::Vec) across steps.
+struct PolledData {
+    std::uint32_t count          = 0;
+    std::uint64_t current_offset = 0;


current_offset is write-only - set in the poll step (messaging_steps.cpp:160) but never read by any THEN. field and assignment can go.

hubcio · 2026-07-02T12:19:12Z

+      ffi (~> 1.1)
+      memoist3 (~> 1.0.0)
+
+PLATFORMS


PLATFORMS is missing x86_64-linux (the CI runner arch), so in the image ffi installs via the generic ruby platform and compiles from source. checked empirically with the pinned bundler 2.6.9: bundle install leaves the lock byte-identical (generic ruby entry already satisfies x86_64-linux, nothing gets appended), and a frozen install succeeds against the current lock as-is - so no mutation, no correctness issue. the only gain from bundle lock --add-platform x86_64-linux + a frozen install is turning future Gemfile/lock drift into a hard error instead of a silent re-resolve. minor, optional.

hubcio · 2026-07-02T12:19:12Z

+# specific language governing permissions and limitations
+# under the License.
+
+common --enable_bzlmod


the committed MODULE.bazel.lock is never enforced - bazel 9 defaults to lockfile_mode=update, so it silently drifts. foreign/cpp/.bazelrc sets --lockfile_mode=error under its ci config and CI builds with --config=ci; same setup here would keep the lock honest.

@slbotbm please confirm

we can add the required config here as well then

hubcio · 2026-07-02T12:19:12Z

+
+# Stage 2: install the Ruby Cucumber runner. build-essential is needed to compile the ffi
+# native extension and never reaches the runtime image.
+FROM ruby:3.3-slim-trixie AS deps


heads up: the ruby 3.3 pin is load-bearing - cucumber 7.1.0 crashes on ruby 3.4 (unknown keywords: :strict, :proc from the Hash.new keyword change). worth a comment here so a future base-image bump doesn't trip on it.

NOTE: i didnt verify it personally

seokjin0414 marked this pull request as ready for review June 23, 2026 09:57

github-actions Bot added the S-waiting-on-review PR is waiting on a reviewer label Jun 23, 2026

seokjin0414 force-pushed the 2965-cpp-bdd-tests branch from 9d6f66a to d670d13 Compare June 23, 2026 10:26

seokjin0414 force-pushed the 2965-cpp-bdd-tests branch from d670d13 to dfc8f20 Compare June 23, 2026 10:56

seokjin0414 force-pushed the 2965-cpp-bdd-tests branch from dfc8f20 to 0e70305 Compare June 23, 2026 22:56

github-actions Bot added S-waiting-on-author PR is waiting on author response and removed S-waiting-on-review PR is waiting on a reviewer labels Jun 24, 2026

seokjin0414 force-pushed the 2965-cpp-bdd-tests branch from 0e70305 to 533c9b0 Compare June 24, 2026 23:41

slbotbm suggested changes Jun 26, 2026

View reviewed changes

Comment thread bdd/cpp/BUILD.bazel

Comment thread bdd/cpp/Dockerfile Outdated

Comment thread foreign/cpp/BUILD.bazel

Comment thread bdd/cpp/.bazelrc Outdated

Comment thread bdd/cpp/.bazelrc Outdated

Comment thread bdd/cpp/MODULE.bazel Outdated

seokjin0414 force-pushed the 2965-cpp-bdd-tests branch 2 times, most recently from f168a4d to 023d49d Compare June 28, 2026 14:41

seokjin0414 requested a review from slbotbm June 28, 2026 15:45

github-actions Bot added S-waiting-on-review PR is waiting on a reviewer and removed S-waiting-on-author PR is waiting on author response labels Jun 28, 2026

slbotbm reviewed Jun 29, 2026

View reviewed changes

Comment thread bdd/cpp/Dockerfile

seokjin0414 force-pushed the 2965-cpp-bdd-tests branch from 023d49d to 61e1c76 Compare June 30, 2026 23:28

seokjin0414 requested a review from slbotbm July 1, 2026 00:39

slbotbm reviewed Jul 1, 2026

View reviewed changes

hubcio removed the S-waiting-on-review PR is waiting on a reviewer label Jul 1, 2026

hubcio added the S-waiting-on-author PR is waiting on author response label Jul 1, 2026

seokjin0414 added 3 commits July 2, 2026 10:44

test(cpp): reserve oversized payload buffer up front

b652417

The oversized-payload test grew its 64 MB buffer one byte at a time and a comment claimed cxx::Vec had no public reserve API. It does, so reserve the capacity once before filling it. Signed-off-by: seokjin0414 <sars21@hanmail.net>

seokjin0414 force-pushed the 2965-cpp-bdd-tests branch from 61e1c76 to 78c3ba4 Compare July 2, 2026 10:21

github-actions Bot added S-waiting-on-review PR is waiting on a reviewer and removed S-waiting-on-author PR is waiting on author response labels Jul 2, 2026

seokjin0414 requested a review from slbotbm July 2, 2026 11:05

hubcio requested changes Jul 2, 2026

View reviewed changes

github-actions Bot added S-waiting-on-author PR is waiting on author response and removed S-waiting-on-review PR is waiting on a reviewer labels Jul 2, 2026

Uh oh!

Conversation

seokjin0414 commented Jun 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What

How it works

Heads-up for reviewers

Notes

Uh oh!

slbotbm commented Jun 23, 2026

Uh oh!

slbotbm commented Jun 23, 2026

Uh oh!

codecov Bot commented Jun 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

seokjin0414 commented Jun 23, 2026

Uh oh!

slbotbm commented Jun 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

slbotbm commented Jun 24, 2026

Uh oh!

seokjin0414 commented Jun 24, 2026

Uh oh!

slbotbm 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!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

slbotbm commented Jun 26, 2026

Uh oh!

seokjin0414 commented Jun 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

seokjin0414 commented Jun 28, 2026

Uh oh!

Uh oh!

seokjin0414 commented Jun 30, 2026

Uh oh!

slbotbm left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

seokjin0414 commented Jul 2, 2026

Uh oh!

hubcio Jul 2, 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!

Choose a reason for hiding this comment

seokjin0414 commented Jun 23, 2026 •

edited

Loading

codecov Bot commented Jun 23, 2026 •

edited

Loading

slbotbm commented Jun 24, 2026 •

edited

Loading

slbotbm left a comment •

edited

Loading

seokjin0414 commented Jun 28, 2026 •

edited

Loading

hubcio Jul 2, 2026 •

edited

Loading

hubcio Jul 2, 2026 •

edited

Loading

slbotbm Jul 2, 2026 •

edited

Loading