Refactor: libcrmcommon: Some pcmk__substitute_secrets() best practices
Refactor: libcrmcommon: Iterate only once in pcmk__substitute_secrets()
Refactor: libcrmcommon: Functionize read/validate CIB secret hash
Refactor: libcrmcommon: Drop forward declarations from cib_secrets.c
Refactor: libcrmcommon: Simplify read_local_file()
Refactor: libpacemaker: Filter out non-regular profile files
Refactor: based, build: Drop struct dirent:d_type check
Refactor: libcrmservice: New services__copy_result() function
Refactor: libcrmservice: Store systemd job path in svc_action_private_t
Refactor: libcrmservice: Subscribe to systemd D-Bus signals
Fix: various: Correctly detect completion of systemd start/stop actions
Refactor: libcrmservice: Improve error checks in systemd_service_name()
Refactor: libcrmservice: Fail systemd rsc start if create override fails
Refactor: libcrmservice: Defunctionize create_world_readable()
Refactor: libcrmservice: Avoid gboolean/gchar in systemd_unit_exists()
Refactor: libcrmservice: Check for empty arg in invoke_unit_by_name()
Refactor: libcrmservice: Simplify systemd override path getters
Refactor: libcrmcommon: Drop check for unused flag from str_any_of()
Low: libcrmservice: Fix overrides for systemd resources
Refactor: tools: NULL-check args in cli_resource_print()
Refactor: tools: NULL-check scheduler arg of cli_resource_move()
Refactor: tools: NULL-check scheduler arg in cli_resource_fail()
Refactor: tools: NULL-check scheduler arg of cli_cleanup_all()
Doc: tools: Suppress Coverity false positives in crm_resource.c
Doc: libcrmservice: Suppress Coverity false positives
Test: libcrmcommon: Unit-test pcmk__scan_nvpair()
Refactor: libcrmcommon: Return standard code from pcmk__scan_nvpair()
Refactor: libcrmcommon: Use g_strsplit() in pcmk__scan_nvpair()
already uses to collect the remaining arguments into an array. Then after parsing all options, it logs those remaining arguments if they're invalid. However, they're valid if we're running a subcommand that uses .
Fix: build: Fix default pacemaker-remoted path on Fedora >= 42
Fix: build: Fix default pacemaker-remoted path on Fedora >= 42
Fix: build: Fix default pacemaker-remoted path on Fedora >= 42
Refactor: libcrmcommon: Avoid unnecessary allocation for sbd paths
Refactor: various: Use uint32_t for group of pcmk_sim_flags internally
Refactor: fencer: Set scheduler input, flags, and status explicitly
Refactor: scheduler: Set scheduler input, flags, and status explicitly
Fix: tools: Avoid crash in crm_simulate --profile
Low: libpacemaker: Handle scandir() error in pcmk__profile_dir()
Refactor: libpacemaker: Create scheduler object in pcmk__profile_dir()
Low: libpacemaker: Fix mem leak in pcmk__profile_dir()
Refactor: libpacemaker: Best practices for pcmk__profile_dir()
Low: tools: Fix overflow in crm_simulate --repeat
Refactor: libpacemaker: Use scandir() filter in pcmk__profile_dir()
Refactor: libpacemaker: Don't make unnecessary copy for scheduler input
Refactor: libpacemaker: Pass NULL input in pcmk__simulate()
Refactor: libpacemaker: Set input and flags explicitly in pcmk__verify()
Refactor: tools: Set scheduler flags explicitly in wait_till_stable()
Refactor: tools: update_dataset() passes nulls to pcmk__schedule_actions
Refactor: libpacemaker: Drop unused args from pcmk__schedule_actions()
Fix: libpe_status: Make cluster_status() idempotent
Refactor: libpacemaker: Unpack status in pcmk__schedule_actions()
Feature: libpacemaker: Reset scheduler object in pcmk_simulate()
Low: libcrmcommon: Fix memory leak in text_end_list()/curses_end_list()
nrwahl2 added a comment to
Patch Review.
There's a LOT that I still don't understand about the scheduler (especially with regard to colocation), but I've probably reviewed or modified more of that code than anyone who's still on the team.
nrwahl2 edited the content of
Patch Review.
nrwahl2 edited the content of
Patch Review.
Fix: libraries: Update version numbers in so names
Fix: libraries: Update version numbers in so names
Refactor: various: Drop pcmk__xe_add_node()
For posterity, re: "reverting change" -- the XML ID is now part of the struct, so from the public API's perspective that commit has not been reverted.
Low: various: Correct some printf specifiers
Low: libcrmcommon: Catch correct errors for remote connection sockets
Refactor: libcrmcommon: Improve error checking in pcmk__new_tls_session
Refactor: based: Assert on int allocation for socket
Refactor: various: Use -1 consistently for uninitialized/invalid socket
Refactor: various: Functionize getting TLS client socket
Doc: Pacemaker Development: Note compatibility with Python 3.6 or later
Low: python: Update py-version to 3.6 in pylintrc
Refactor: agents: Drop Python 3.4 workaround from fence_legacy
Doc: python: Drop mentions of python 3.4
Fix: libcib: Don't send CRM_OP_REGISTER from cib_remote client
@kgaillot I don't think the existing XSLT error handler sets config warning, FYI. If we want that, we'll want a wishlist task for it, because it looks like it's gonna require some rearranging of that might be tricky to avoid redundant logs.
Refactor: tools: Use cib_sync_call unconditionally in cibadmin
Feature: tools: Deprecate cibadmin --sync-call
Log: xml: Warn about post-transform behavior changes for ACLs with xpath
Log: xml: Warn about post-transform behavior changes for ACLs with xpath
Low: xml: Fix ACL xpath typo from 986cc5c4
Test: cts-schemas: Add test for ACL validity after other transformations
Fix: xml: Ensure ACL permissions are valid after XSL transformations
Fix: xml: Preserve ACL reference behavior for replaced constraints
Test: cts-schemas: Add test for ACL validity after other transformations
Fix: xml: Ensure ACL permissions are valid after XSL transformations
Fix: xml: Preserve ACL reference behavior for replaced constraints
@kgaillot Can you weigh in when you have a chance? I know things are especially busy.
I figure it's unwise to work on the case until we agree upon how we ought to address it -- if at all.