Record of changes made to functional tests to get them working with SoD

Note, only doing local and non-batch system tests initially.

test mod details

startup

• 00-state-summary.t: adjusted suite to keep failed task in pool under SoD.

validate

• 60-groups.t: removed `spawn to max active cycles'
• 36-fail-double-runahead.t: removed 'runahead limit' (to be restored for SoD)

restart

many changes during restart implementation CHANGE NOTES TBD Note 22-hold.t is now flaky, needs investigating.

cylc-poll

Two mods:

• 01-task-failed:
  • replaced suicide trigger with failed handler task
• 12-reverse-state:
  • removed; state reversal not meaningful under SoD (if we reset a running task to succeeded, it would be removed as succeeded).

cylc-get-config

• remove two obsolete config items from 00-simple/section1.stdout
  • runahead limit (THIS NEEDS TO BE RESTORED)
  • spawn to max active cycle points

job-submission

• 01-job-nn-localhost.t: had to modify db.sqlite3 for new task_pool and task_pool checkpoints table.

message-triggers

• 02-action.t: changed test suite to a more appropriate model (to test that message triggers work immediately). The original was SoS-specific, and not very sensible.

xtriggers

• 03-sequence.t: removed "spawn to max active cycle points" (works without it in SoD)

ext-trigger

• 01-no-nudge.t: altered slighly for SoD fail handling

logging

• 02-duplication.t - replaced cylc reset with cylc spawn for SoD.

cli

• 02.t adjust task pool comparisons for new DB content.

shutdown

• 09-now2.t: restarted suite should not abort as stalled in SoD as the failure was handled.

events

• 00-suite.t: adjusted timeout sub-suite for SoD fail handling (needs a stalled suite)
• 04-timeout-ref-live.t: ditto
• 06-timeout-ref-simulation.t : ditto
• 34-task-abort.t: adjust for SoD fail handling
• 03-timeout.t: ditto
• 23-suite-stalled-handler.t: ditto, and "cylc spawn" instead of "cylc reset"
• 35-task-event-handler-importance.t: remove unnec suicide trigger
• 26-suite-stalled-dump-prereq.t: adjusted for minimal SoD task pool

periodicals

• all: swapped runahead limit for max active

cylc-remove

• 00-simple.t: adjust for SoD fail handling and pool content
• 00-simple-comat.t: deleted, tests obsolete CLI

cylc-show

• 01-clock-triggered.t: adjusted for SoD task pool content
• 03-clock-triggered-non-utc-mode.t: ditto
• 02-clock-triggered-alttz.t: REMOVED same as 01.t (no alt TZ)
• 05-complex.t: adjusted for SoD: run the suite (instead of hold) and get the foo task to call cylc show on itself.

pre-initial

• 08-conditional-messaging.t: adjust for SoD fail handling
• 10-warm-insert.t: don't need to insert and trigger separately; removed inter-cycle dependence to avoid reflow after manual trigger (does not affect the subject of the test).
• 11-warm-insert-stall.t: REMOVED. The test is the same as 10.t but does not trigger the inserted task (which therefore stalls); In SoD we can't (because there's no need for it) insert unsatisfied waiting tasks.

authentication

• 05-full-control.t: minor adjustments for SoD task pool and commands (reset -> trigger)
• (note 4 tests skipped due to "anon auth not supported")

hold-release

• replaced 02 with SoD-appropriate test

• replaced with 08-hold.t:

  • 07-hold-family-exact.t
  • 08-hold-family-inexact.t
  • 09-hold-task-exact.t
  • 10-hold-task-inexact.t:

-18-hold-cycle-globs.t: added a new task to spawn the tasks to be held (that are already present in the SoS task pool)

• 19-no-reset-prereq-on-waiting.t: ditto

• replaced with 05-release.t:

  • 05-release-task-exact.t
  • 06-release-task-inexact.t
  • 04-release-family-inexact.t
  • 03-release-family-exact.t

• directory run-hold-after removed (not used)

suite-state

• 00/01.t: fixed by modifying the "upstream" suite for SoD failed handling

cylc-trigger

• 02-filter-failed.t: adjusted for SoD fail handling
• 04-filter-names.t: ditto
• 05-filter-cycle.t: ditto
• 08-edit-run-host-select.t: ditto
• 06-reset-ready.t:(reset command not needed in SoD; renamed 06-reflow.t - makes use of REFLOW) -03-edit-run.t: SoD fail handling -07-edit-run-abort.t: modified a bit for SoD, to keep a target task in the pool for retriggering.

triggering

• 06-fam-succeed-any.t: SoD fail-handling; remove unnecessary suicide trigger
• 04-fam-fail-all.t: ditto
• 05-fam-finish-all.t: remove suicide trigger
• 07-fam-fail-any.t: ditto
• 09-fail.t: ditto
• 10-finish.t: ditto
• 16-fam-expansion.t: modify to let suite run and get bar.1 to show its own prerequisites (original requires waiting bar.1 to exist in the SoS pool).
• 17-suicide-multi.t: adjusted for SoD task pool content
• 19-and-suicide.t: ditto
• 20-and-outputs-suicide.t: ditto
• 14-submit-fail.t: Unchanged, but note this test requires suicide trigger to remove a handled submit-failed task! (should we manage like handled failed tasks?)

reload

• 22-remove-task-cycling.t: small change for task pool content
• 18-broadcast-insert.t: replace insert with trigger
• 15-state-summary.t: simplified unnecessarily complex suite
• 14-waiting.t: modified to use a partially satisfied waiting task
• 13-add-task.t: added an intermediate task to avoid the complications of modifying the downstream consequences of a live task
• 05-graphing-simple.t: ditto
• 06-graphing-fam.t: ditto
• 12-remove-task.t: ditto
• 16-remove-add-alter-task.t: ditto (note these tests are of dubious benefit for SoD; in SoS reload had to modify waiting task proxies).
• 17-graphing-change.t: changed to integer cycling; no need to release the suite, just force-spawn proxies to check orphaning behaviour on reload.
• 10-runahead.t: rh-limit -> max active, and adjust for SoD fail handling

cyclers

• 40-integer_exclusions_advanced.t: max active cycle points instead of runahead limit
• 33-integer1.t: ditto

special

• 00-sequential.t: runahead limit -> max active cycle points

runahead

• 00-runahead.t: rh-limit -> max active
• 04-no-final.t: ditto, and simplified
• 06-release-update.t: adjusted for SoD pool content
• 02-check-default-complex.t: added failed tasks in each cycle to hold max active cycle points back (the SoS version relies on a waiting task that never gets spawned in SoD)

REMOVED as SoS-specific

• spawn-max/

• task-proc-loop/: this just counts main loop iterations for a small workflow; it could just be updated for the current SoD implementation but relevance isn't clear?

• cylc-insert: should perhaps reinstate these and convert them to cylc-trigger (with auto-insert) tests?

• cylc-reset

graphql

• 02-root-queries.t: (fixed on David's advice)

cylc-scan

• 04-outputs.t: adjusted for SoD task pool content

TBD

NOTES

Stalled detection is less useful in SoD: it relies on unmet prerequisites of waiting tasks.

FLAKYTESTS

• restart/21-elapsed.t: use --stop-point not --until (the latter change the FCP, and we can't spawn past that!); and final pool test should not have post-FCP tasks.

• job-submission/19-chatty.t: adjust for SoD task pool content and lack of spawned flag.