Summary
Several test-parity/known_failures.json entries are categorized ci-env: they fail on the Linux CI runner for environmental reasons, not Perry compiler bugs. They had no tracking issue (issue: null). This is their shared home.
These are not fixed by changing Perry — they need CI-side fixtures or runner setup.
Entries
Need a TCP/TLS echo-server fixture (pass locally with a server running; CI has none):
test_issue_422_socket_connect
test_net_min
test_net_socket
test_net_upgrade_tls (needs the TLS-upgrade companion server)
test_sock_write_map
Need an npm install in CI:
test_ramda_sum (parity harness can't npm install ramda)
Runner-specific / decorator-metadata flake:
test_decorators_nest_js_common_canary (expected-output mismatch under the Linux runner; tied to NestJS decorator metadata-inheritance follow-ups — re-home to a dedicated decorator-parity issue if it turns out to be a real compiler gap rather than runner noise)
Suggested resolution
Add the echo-server lifecycle (already present in run_parity_tests.sh as start_echo_server/start_tls_upgrade_server) to the CI job, or gate these tests behind a requires-fixture marker so they don't need to live in the skip-list. Part of #793.
Summary
Several
test-parity/known_failures.jsonentries are categorizedci-env: they fail on the Linux CI runner for environmental reasons, not Perry compiler bugs. They had no tracking issue (issue: null). This is their shared home.These are not fixed by changing Perry — they need CI-side fixtures or runner setup.
Entries
Need a TCP/TLS echo-server fixture (pass locally with a server running; CI has none):
test_issue_422_socket_connecttest_net_mintest_net_sockettest_net_upgrade_tls(needs the TLS-upgrade companion server)test_sock_write_mapNeed an npm install in CI:
test_ramda_sum(parity harness can'tnpm install ramda)Runner-specific / decorator-metadata flake:
test_decorators_nest_js_common_canary(expected-output mismatch under the Linux runner; tied to NestJS decorator metadata-inheritance follow-ups — re-home to a dedicated decorator-parity issue if it turns out to be a real compiler gap rather than runner noise)Suggested resolution
Add the echo-server lifecycle (already present in
run_parity_tests.shasstart_echo_server/start_tls_upgrade_server) to the CI job, or gate these tests behind arequires-fixturemarker so they don't need to live in the skip-list. Part of #793.