nokia / danm

24%

master: 44%

LAST BUILD ON BRANCH fix_add_gc

branch: fix_add_gc

CHANGE BRANCH

x

Reset

• fix_add_gc
• add-copyright-headers
• akraino-install-guide-link
• alloc6_api
• api_group_migration
• binaries
• category_all
• central_ipam_fixes
• cmd_intro
• cni_del_fix
• cni_res_fix
• cnidel_steroid
• cnidel_uts
• concurrent_confman
• configure_naming
• damn-danm
• deploy-guide
• device_handling_fixes
• dis_dad
• dnet_apitype_fix
• dummy_cosmetics
• dummy_encsa
• fix_confman_core
• fix_ip_delegation
• fix_netype_bug
• fix_none_ip
• fixin_uts
• formatting
• go_12_9
• intel_dpdk
• intel_vf_wa
• ip6conf_for_none
• ipam_for_everyone
• ipam_prefix_change
• ipam_store_last_ip
• ipam_uts
• ipv6_cudo
• k8s_1.21_fix
• logo-with-name
• master
• metacni_fourdoto
• netadmit_uts
• netdel_ut
• netwatcher_fourdoto
• netwatcher_healthcheck
• netwatcher_modify
• netwatcher_nad
• networktype_separation
• new_apis
• outage
• perf_fix
• pod_id_to_ep
• prefix_fix
• readme_facelift
• route_device_fix
• route_man_fix
• rutas
• set_ndisc_noti
• slack-links
• store_client
• svcw_update_fix
• svcwatcher_fourdoto
• svcwatcher_healthcheck
• svcwatcher_ns
• svcwatcher_update_eps
• syncher_ut
• tconf_hook_ut
• tenantnet_setup
• typo-correction
• uid_weeding
• update_delegation
• v3.2.0
• v3.3.0
• v4.0.0
• v4.1.0
• v4.2.0
• v4.2.0-RC1
• v4.2.0-rc2
• v4.2.0-rc3
• v4.2.1
• v6_alloc
• v6_stat
• validate_max_netmask
• validate_new_apis
• validation_type_change
• versioning
• webhook
• wording/changes
• y_u_no_free

Build # 526

Build Type

Pull #102

travis-ci

Committed by web-flow

Commit Message

Fixing bug of not properly cleaning up after ALL interfaces when SOME of them failed during create.

The root cause is that in some cases we have returned earlier during interface creation, than starting the asynch thread.
But if some other threads have already progressed to the point of forking, they were basically prematurely reaped by the early
termination of the whole process.
Nasty.

Resolution is getting rid of ALL early return cases. In case of an early failure we simply push the result in to the syncher object, even
if the error happened in the main thread.
Only Syncher can return from the main thread, no sooner and no later than all results are available: be them pushed from the main thread, or from a fork.

Pull Request Pull Request #102: Some interfaces are not properly cleaned during CNI_ADD if some other interface creations fail

Run Details

30 of 30 new or added lines in 4 files covered. (100.0%)

551 of 2306 relevant lines covered (23.89%)

6.01 hits per line