rspec-conductor Support

docs/Parallelism.md

@@ -285,8 +285,9 @@ SimpleCov coordinates with parallel test runners through a small pluggable adapt
   uses its `ParallelTests.first_process?` / `ParallelTests.wait_for_other_processes_to_finish` APIs for precise worker
   coordination. Activates only when the native `parallel_tests` pid-file contract is present.
 - **`GenericAdapter`** — catch-all for any runner that follows the `TEST_ENV_NUMBER` / `PARALLEL_TEST_GROUPS` env-var
-  convention but doesn't ship a Ruby API (parallel_rspec, knapsack-style splitters, custom CI sharding scripts).
-  Activates when `TEST_ENV_NUMBER` is set and no more-specific adapter is.
+  convention but doesn't ship a Ruby API (parallel_rspec, [rspec-conductor](https://github.com/markiz/rspec-conductor)
+  1.0.7 or later, knapsack-style splitters, custom CI sharding scripts). Activates when `TEST_ENV_NUMBER` is set and no
+  more-specific adapter is.
 
 Adapters are tried in registration order; the first whose `active?` returns `true` is chosen. With both built-ins, this
 means parallel_tests users get the precise gem-based path and parallel_rspec (or any env-var-only runner) gets the

features/parallel_tests.feature

@@ -22,7 +22,7 @@ Feature:
       SimpleCov.start
       """
     When I open the coverage report generated with `bundle exec parallel_rspec -n 2 spec`
-    Then I should see the line coverage results for the parallel tests project
+    Then I should see the line coverage results for the parallel fixture project
 
   # Note it's better not to test this in the same scenario as before.
   # Merging of results might kick in and ruin this.
@@ -34,7 +34,7 @@ Feature:
       SimpleCov.start
       """
     When I open the coverage report generated with `bundle exec rspec spec`
-    Then I should see the line coverage results for the parallel tests project
+    Then I should see the line coverage results for the parallel fixture project
 
   @branch_coverage
   Scenario: Running the project with normal rspec and branch coverage
@@ -47,7 +47,7 @@ Feature:
       end
       """
     When I open the coverage report generated with `bundle exec rspec spec`
-    Then I should see the branch coverage results for the parallel tests project
+    Then I should see the branch coverage results for the parallel fixture project
 
   @branch_coverage
   Scenario: Running the project with normal rspec and branch coverage
@@ -60,7 +60,7 @@ Feature:
       end
       """
     When I open the coverage report generated with `bundle exec parallel_rspec -n 2 spec`
-    Then I should see the branch coverage results for the parallel tests project
+    Then I should see the branch coverage results for the parallel fixture project
 
   Scenario: Coverage violations aren't printed until the end
     Given I install dependencies

features/rspec_conductor.feature

@@ -0,0 +1,98 @@
+@rspec @disable-bundler @process_fork
+
+Feature:
+
+  rspec-conductor is a queue-based parallel spec runner. It forks its workers
+  from a server process, gives each worker one spec file at a time, and marks
+  every worker with the parallel_tests environment convention: TEST_ENV_NUMBER
+  per worker plus PARALLEL_TEST_GROUPS for the worker count (since
+  rspec-conductor 1.0.7). SimpleCov's generic parallel adapter picks that up
+  without any configuration, so each worker records under a unique suite name
+  and the first worker waits for its siblings and builds the merged report.
+
+  # `--workers 2` pins the worker count for a fast, deterministic run. Left to
+  # default, rspec-conductor spawns one worker per core. Workers that get no
+  # spec file still load spec_helper at spawn and record an empty resultset,
+  # so the merged totals come out the same either way.
+  #
+  # The scenarios assert on the generated report files rather than on the
+  # usual "Coverage report generated" line. The conductor server closes each
+  # worker's output pipes once the worker has sent its run summary, before
+  # at_exit hooks run, so nothing SimpleCov prints from at_exit can reach
+  # the output. The report itself is written fine.
+
+  Background:
+    Given I'm working on the project "parallel_tests" with the Gemfile from "rspec_conductor"
+
+  Scenario: Running it through rspec-conductor produces the same results as a normal rspec run
+    Given I install dependencies
+    And SimpleCov for RSpec is configured with:
+      """
+      require 'simplecov'
+      SimpleCov.start
+      """
+    When I successfully run `bundle exec rspec-conductor --workers 2 spec`
+    Then the following files should exist:
+      | coverage/index.html      |
+      | coverage/.resultset.json |
+    When I open the coverage report
+    Then I should see the line coverage results for the parallel fixture project
+
+  # In default mode the first worker's TEST_ENV_NUMBER is "" like
+  # parallel_tests. With --first-is-1 it is "1" instead, and the workers count
+  # 1..N. Both spellings must resolve to a single reporting worker.
+  Scenario: Running with --first-is-1 merges just the same
+    Given I install dependencies
+    And SimpleCov for RSpec is configured with:
+      """
+      require 'simplecov'
+      SimpleCov.start
+      """
+    When I successfully run `bundle exec rspec-conductor --workers 2 --first-is-1 spec`
+    Then the following files should exist:
+      | coverage/index.html      |
+      | coverage/.resultset.json |
+    When I open the coverage report
+    Then I should see the line coverage results for the parallel fixture project
+
+  @branch_coverage
+  Scenario: Running with branch coverage enabled
+    Given I install dependencies
+    And SimpleCov for RSpec is configured with:
+      """
+      require 'simplecov'
+      SimpleCov.start do
+        enable_coverage :branch
+      end
+      """
+    When I successfully run `bundle exec rspec-conductor --workers 2 spec`
+    Then the following files should exist:
+      | coverage/index.html      |
+      | coverage/.resultset.json |
+    When I open the coverage report
+    Then I should see the branch coverage results for the parallel fixture project
+
+  # Each worker on its own covers only a slice of the project, so a threshold
+  # check against a single worker's partial result would fail. Only the
+  # reporting worker, after merging every sibling's slice, may enforce it.
+  #
+  # The conductor server swallows a worker's output printed from at_exit and
+  # ignores its exit status once its run summary is in, so asserting on the
+  # command's output or exit code proves nothing here. Instead the scenario
+  # leans on .last_run.json, which SimpleCov writes only when the threshold
+  # check passes. Finding it filled with the merged 81.48 shows the check ran
+  # against the merged result, where any single worker's slice would have
+  # failed and left the file unwritten.
+  Scenario: Coverage thresholds are enforced against the merged result, not per worker
+    Given I install dependencies
+    And SimpleCov for RSpec is configured with:
+      """
+      require 'simplecov'
+      SimpleCov.start do
+        minimum_coverage 81.48
+      end
+      """
+    When I successfully run `bundle exec rspec-conductor --workers 2 spec`
+    Then the file "coverage/.last_run.json" should contain "81.48"
+    When I open the coverage report
+    Then I should see the line coverage results for the parallel fixture project

features/step_definitions/parallel_tests_steps.rb → features/step_definitions/parallel_fixture_steps.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-Then "I should see the line coverage results for the parallel tests project" do
+Then "I should see the line coverage results for the parallel fixture project" do
   steps %(
     Then I should see the groups:
       | name      | coverage | files |
@@ -15,7 +15,7 @@
   )
 end
 
-Then "I should see the branch coverage results for the parallel tests project" do
+Then "I should see the branch coverage results for the parallel fixture project" do
   steps %(
     Then I should see the groups:
       | name      | coverage | files |

features/step_definitions/simplecov_steps.rb

@@ -159,3 +159,15 @@ def with_cross_worker_install_lock
 
   step 'I cd to "project"'
 end
+
+# For fixtures that reuse another project's sources wholesale and differ only
+# in which gems they pull in (their test_projects directory holds nothing but
+# a Gemfile and Gemfile.lock).
+Given "I'm working on the project {string} with the Gemfile from {string}" do |project_name, gemfile_project|
+  step "I'm working on the project \"#{project_name}\""
+
+  gemfile_source = File.expand_path("../../test_projects/#{gemfile_project}", __dir__)
+  cd(".") do
+    FileUtils.cp(Dir.glob("#{gemfile_source}/Gemfile*"), ".")
+  end
+end

lib/simplecov/color.rb

@@ -45,6 +45,10 @@ def enabled?(stream = $stderr)
       return true  if env_set?("FORCE_COLOR")
 
       stream.tty?
+    rescue IOError
+      # A parallel runner can close a worker's stdio before its at_exit
+      # hooks run (rspec-conductor does). A closed stream is not a tty.
+      false
     end
 
     def for_percent(percent)

lib/simplecov/exit_codes.rb

@@ -18,6 +18,10 @@ module ExitCodes
     # warning logs. `print_errors false` remains the intended opt-out.
     def self.print_error(message)
       $stderr.puts message # rubocop:disable Style/StderrPuts
+    rescue IOError
+      # A parallel runner can close a worker's stderr before its at_exit
+      # hooks run (rspec-conductor does). The violation still has to set
+      # the exit status even when its explanation cannot be printed.
     end
   end
 end

lib/simplecov/formatter/base.rb

@@ -32,6 +32,11 @@ def initialize(silent: false, output_dir: nil)
       # Subclasses call this at the end of their `format`.
       def emit_status(result)
         $stderr.puts output_message(result) unless @silent # rubocop:disable Style/StderrPuts
+      rescue IOError
+        # A parallel runner can close a worker's stderr before its at_exit
+        # hooks run (rspec-conductor does). Losing the status line must not
+        # abort the exit tasks that follow report generation, i.e. the
+        # threshold checks and the .last_run.json write.
       end
 
       # Subclasses override to prepend a marker (e.g. "JSON ") to the

lib/simplecov/parallel_adapters.rb

@@ -20,8 +20,10 @@ module SimpleCov
   #     (precise sync + first-process detection via the gem's own API).
   #   - `GenericAdapter` — env-var-only detection for runners that follow
   #     the parallel_tests `TEST_ENV_NUMBER` convention but don't ship a
-  #     Ruby API (parallel_rspec, custom CI sharding, knapsack-style
-  #     splitters). See https://github.com/simplecov-ruby/simplecov/issues/1065.
+  #     Ruby API (parallel_rspec, rspec-conductor 1.0.7+, custom CI
+  #     sharding, knapsack-style splitters). See
+  #     https://github.com/simplecov-ruby/simplecov/issues/1065 and
+  #     https://github.com/simplecov-ruby/simplecov/issues/1156.
   #
   # Users can plug in additional adapters:
   #

lib/simplecov/parallel_adapters/generic.rb

@@ -7,14 +7,15 @@ module ParallelAdapters
     # Catch-all adapter for parallel test runners that follow the
     # `TEST_ENV_NUMBER` / `PARALLEL_TEST_GROUPS` env-var convention but
     # don't ship a Ruby API for SimpleCov to hook (parallel_rspec,
-    # knapsack-style splitters, custom CI sharding scripts). Activates
-    # when `TEST_ENV_NUMBER` is set; doesn't require any specific gem to
-    # be loaded.
+    # rspec-conductor since 1.0.7, knapsack-style splitters, custom CI
+    # sharding scripts). Activates when `TEST_ENV_NUMBER` is set;
+    # doesn't require any specific gem to be loaded.
     #
     # Heuristic for `first_worker?`: the worker whose `TEST_ENV_NUMBER`
-    # is `""` (parallel_tests/parallel_rspec convention) or `"1"`
-    # (zero-based runners that start at 1). Any other value is treated
-    # as a non-first worker.
+    # is `""` (parallel_tests/parallel_rspec convention, rspec-conductor's
+    # default) or `"1"` (runners that number from 1, like rspec-conductor
+    # with `--first-is-1`). Any other value is treated as a non-first
+    # worker.
     #
     # `wait_for_siblings` is inherited from Base as a no-op — without a
     # runner-provided API the only synchronization available is polling

spec/color_spec.rb

@@ -84,6 +84,13 @@
       expect(described_class.enabled?(stdout_not_tty)).to be false
     end
 
+    it "returns false for a closed stream (a parallel runner can close worker stdio before at_exit runs)" do
+      read, write = IO.pipe
+      read.close
+      write.close
+      expect(described_class.enabled?(write)).to be false
+    end
+
     it "falls through when SimpleCov has no color config (standalone CLI)" do
       # In the standalone `exe/simplecov` process, `simplecov/color` is
       # loaded without `simplecov/configuration`, so `SimpleCov` doesn't

spec/exit_codes_spec.rb

@@ -0,0 +1,17 @@
+# frozen_string_literal: true
+
+require "helper"
+
+RSpec.describe SimpleCov::ExitCodes do
+  describe ".print_error" do
+    it "writes the message to stderr" do
+      expect(capture_stderr { described_class.print_error("boom") }).to eq("boom\n")
+    end
+
+    it "stays quiet when stderr was closed before at_exit (e.g. by rspec-conductor)" do
+      allow($stderr).to receive(:puts).and_raise(IOError, "closed stream")
+
+      expect { described_class.print_error("boom") }.not_to raise_error
+    end
+  end
+end

spec/formatter/base_spec.rb

@@ -38,6 +38,13 @@ def result_with(coverage_statistics)
     expect(Warning).not_to have_received(:warn)
   end
 
+  it "loses the status line quietly when stderr was closed before at_exit (e.g. by rspec-conductor)" do
+    result = result_with(line: statistics(8, 2))
+    allow($stderr).to receive(:puts).and_raise(IOError, "closed stream")
+
+    expect { formatter.send(:emit_status, result) }.not_to raise_error
+  end
+
   it "does not build or emit a status when silent" do
     silent_formatter = described_class.new(silent: true)
     result = result_with(line: statistics(8, 2))

test_projects/rspec_conductor/Gemfile

@@ -0,0 +1,16 @@
+# frozen_string_literal: true
+
+source "https://rubygems.org"
+
+gem "rspec"
+gem "rspec-conductor"
+
+# when the tests are executed the project is in tmp/aruba/project
+# which is a different nesting from its usual place
+if File.exist?("../../simplecov.gemspec")
+  gem "simplecov", path: "../.."
+else
+  # rubocop:disable Bundler/DuplicatedGem
+  gem "simplecov", path: "../../.."
+  # rubocop:enable Bundler/DuplicatedGem
+end

test_projects/rspec_conductor/Gemfile.lock

@@ -0,0 +1,46 @@
+PATH
+  remote: ../..
+  specs:
+    simplecov (1.1.0)
+
+GEM
+  remote: https://rubygems.org/
+  specs:
+    diff-lcs (1.6.2)
+    rspec (3.13.2)
+      rspec-core (~> 3.13.0)
+      rspec-expectations (~> 3.13.0)
+      rspec-mocks (~> 3.13.0)
+    rspec-conductor (1.0.11)
+      rspec-core (>= 3.8.0)
+    rspec-core (3.13.6)
+      rspec-support (~> 3.13.0)
+    rspec-expectations (3.13.5)
+      diff-lcs (>= 1.2.0, < 2.0)
+      rspec-support (~> 3.13.0)
+    rspec-mocks (3.13.8)
+      diff-lcs (>= 1.2.0, < 2.0)
+      rspec-support (~> 3.13.0)
+    rspec-support (3.13.7)
+
+PLATFORMS
+  arm64-darwin-25
+  ruby
+
+DEPENDENCIES
+  rspec
+  rspec-conductor
+  simplecov!
+
+CHECKSUMS
+  diff-lcs (1.6.2) sha256=9ae0d2cba7d4df3075fe8cd8602a8604993efc0dfa934cff568969efb1909962
+  rspec (3.13.2) sha256=206284a08ad798e61f86d7ca3e376718d52c0bc944626b2349266f239f820587
+  rspec-conductor (1.0.11) sha256=6958c98d4bf07cff6827229160debca81c8f13dd550dd8e1db02e63e3c092618
+  rspec-core (3.13.6) sha256=a8823c6411667b60a8bca135364351dda34cd55e44ff94c4be4633b37d828b2d
+  rspec-expectations (3.13.5) sha256=33a4d3a1d95060aea4c94e9f237030a8f9eae5615e9bd85718fe3a09e4b58836
+  rspec-mocks (3.13.8) sha256=086ad3d3d17533f4237643de0b5c42f04b66348c28bf6b9c2d3f4a3b01af1d47
+  rspec-support (3.13.7) sha256=0640e5570872aafefd79867901deeeeb40b0c9875a36b983d85f54fb7381c47c
+  simplecov (1.1.0)
+
+BUNDLED WITH
+  4.0.18