Skip to content

perf(tests): one BPMN editor journey; the full new-file sweep leaves the PR gate - #6709

Merged
delchev merged 1 commit into
masterfrom
perf/it-consolidate-editors
Aug 13, 2026
Merged

perf(tests): one BPMN editor journey; the full new-file sweep leaves the PR gate#6709
delchev merged 1 commit into
masterfrom
perf/it-consolidate-editors

Conversation

@delchev

@delchev delchev commented Aug 13, 2026

Copy link
Copy Markdown
Contributor

What

Two consolidations in the ui/smoke IT set — the third installment of the IT-speed effort (#6702 sharding, #6703/#6708 boot-per-class):

1. The three BPMN editor ITs become one editor session. BpmnEditorLoadsIT, BpmnEditorPropertyPopupIT and BpmnEditorIT each paid a Dirigible boot, a browser and ~30s of Oryx editor bring-up to look at the same freshly created .bpmn file — three sequential stations of one user journey. BpmnEditorIT now asserts all three in order (editor boot → property-popup lifecycle → rename + save + "Published"), keeping every assertion of all three classes. Measured: 59s for the merged journey vs ~200s of CI time for the three classes.

2. The exhaustive new-file sweep leaves the per-PR smoke gate. CreateNewFileIT loops all 20 NewFileOptions, each opening its own editor — the single longest IT (431s on CI) — and was tagged smoke, so every PR paid for it. The PR gate now runs the new CreateNewFileSmokeIT: one representative option per heavyweight editor family (Monaco/TypeScript, BPMN modeler, EDM modeler, form builder), 59s locally vs 187s for the sweep. The full sweep keeps running on master + nightly in the ui shard, so no coverage is lost — it just stops taxing every PR.

Honesty note: closing each editor tab between options was tried first and measured — 183.8s vs 186.9s baseline, i.e. no effect (the cost is the per-editor bring-up itself, not iframe accumulation) — so that change was discarded and the sweep body stays untouched.

CLAUDE_BPMN.md / CLAUDE_UI.md updated for the merged class.

Impact

  • ui shard (master/nightly): ~2.5 min less (two fewer boots + editor bring-ups).
  • PR smoke gate: ~5 min less (431s sweep → ~135s slice, extrapolating the local 2.3× CI factor).

Verification

  • Merged BpmnEditorIT: green, 58.65s (isolated Maven repo, fresh state).
  • CreateNewFileSmokeIT: green, 58.63s.
  • CreateNewFileIT (untouched body, tag removed): green in the same setup earlier (186.9s baseline run).
  • mvn formatter:validate green.

🤖 Generated with Claude Code

…the PR gate

Two consolidations in the ui/smoke IT set:

BpmnEditorIT, BpmnEditorLoadsIT and BpmnEditorPropertyPopupIT each paid
a Dirigible boot, a browser and ~30s of editor bring-up to look at the
same freshly created .bpmn file - three sequential stations of one user
journey. They are now ONE editor session in BpmnEditorIT (boot
assertions -> property-popup lifecycle -> rename + save + published),
keeping every assertion of all three classes. Measured: 59s for the
merged journey vs ~200s of CI time for the three classes.

CreateNewFileIT (the full 20-option sweep - every option opens its own
editor, making it the single longest IT at 431s on CI) was tagged
smoke, so every PR paid for it. The PR gate now runs the new
CreateNewFileSmokeIT - one representative option per heavyweight editor
family (Monaco, BPMN modeler, EDM modeler, form builder), 59s locally
vs 187s for the sweep - while the exhaustive sweep keeps running
nightly/master in the ui shard. (Closing each editor tab between
options was tried first and measured: 183.8s vs 186.9s baseline - the
cost is the per-editor bring-up itself, not iframe accumulation, so the
sweep stays as-is.)

CLAUDE_BPMN.md / CLAUDE_UI.md updated for the merged class.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
@delchev
delchev merged commit 7ac7ce1 into master Aug 13, 2026
9 checks passed
@delchev
delchev deleted the perf/it-consolidate-editors branch August 13, 2026 16:29
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant