[dborup-agent]

Relates to #596

Plan and change

Port of upstream PR 2274 (read only), adapted to this fork rather than cherry-picked.

• makeColumnsResizable appends an absolutely positioned .col-resize-handle to each header. A static th is not a containing block, so the handle was positioned against the page instead of its header.
• Fix (public/app.js, 2 lines): before appending a handle, set position: relative on the header only if its computed position is static. Sticky, relative, absolute and fixed headers are left untouched.
• Fork adaptations: tests live in the repo root here; the makeColumnsResizable measures once and credits colspan rows to column 0 (packets Details column too narrow) #258 sandbox needed a getComputedStyle stub; the E2E uses this fork's theme-refresh event to wait for the final analytics render.

Tests (red before, green after)

• test-frontend-helpers.js: static headers get relative (measured and saved widths); positioned headers are preserved; last header untouched.
• test-table-fluid-e2e.js: analytics nodes tab at 1250px and 375px; every handle's offsetParent is its header, its box sits on the header's right edge, and the page does not overflow.
• Mutants: (1) fix removed, (2) unconditional position = 'relative'. Both fail the unit test (2 failures each); with the fix removed both E2E cases fail.

Performance

One getComputedStyle read per header at setup, once per table. No new requests, listeners or settings.

[dborup-agent]

Rapport — CS-pve-agent3 PR#605 #596 — head be802fb

Status: Ready for review (draft); all CI jobs green, no flaky-test rerun needed.

Requirement: resize handles anchored to their table headers (port of upstream 2274, adapted)

Check	Test	Red before	Green after	Mutant
Static headers get position: relative (measured and saved widths), last header untouched	test-frontend-helpers.js	2 failures [T]	731 passed, 0 failed [T]	Fix removed: 2 failures. Unconditional relative: 2 failures [T]
Already positioned headers (sticky/relative/absolute/fixed) are not overridden	test-frontend-helpers.js	n/a (passes before; guards the conditional)	pass [T]	Unconditional relative mutant fails it [T]
Handles sit on the header's right edge, offsetParent is the header, no page overflow, desktop 1250px and mobile 375px	test-table-fluid-e2e.js (local server, fixture DB)	2 failed [T]	43 passed, 0 failed [T]	Fix removed: both viewport cases fail [T]
Existing column-width behaviour unchanged	test-issue-258-column-widths.js, -e2e.js	n/a	15/15 and 22/22 pass [T]	n/a

Adaptation notes [A]: the fix is two lines in public/app.js; the #258 sandbox gained a getComputedStyle stub; tests live in the repo root as in this fork. The upstream drag/reload-restore and sticky-Nodes E2E cases were not ported; drag and saved-width behaviour is already covered by the existing #258 E2E.

Other gates

• sh test-all.sh: 273 files passed, 0 failed [T]
• scripts/check-xss-sinks.sh --diff origin/master: clean, exit 0 [T]
• No new map[string]interface{}, no Go changes, cmd/server untouched [T]
• No hardcoded colors; no workflow files changed, so the fork guards are unchanged [T]
• Go suites were not run locally because no Go code changed; they ran in CI [A]

CI per job (run on head be802fb) [T]

All passed: Go Build & Test, Go Libraries & CLI, Go Ingestor Tests, Go Server Tests (1/3, 2/3, 3/3), JavaScript Unit Tests & Lint, Script & Container Unit Checks, Verify Go Test Assignments, Validate Docker Image, Playwright E2E Tests (1/3, 2/3, 3/3), Verify E2E Shards & Coverage, Build & Publish Docker Image. Skipped by design: Deploy Staging, Release Artifacts, Publish Badges & Summary. Known flaky test #271 did not fail.

Remaining

• test-table-fluid-e2e.js is not referenced by name in deploy.yml; I did not confirm how CI picks it up, so the new E2E cases may not run there [K]. The Playwright shards passed either way.
• Browser validation was done headless against a local server only; no screenshot taken [A].

[dborup-agent]

Review — CS-pve-agent1 PR#605 — head be802fb

Dom: APPROVE med nits

Independent, read-only review. Head SHA was identical before and after [T]. Verified on the merged tree of the head and origin/master (edfeea99); the merged tree is byte-identical to the head for all files [T].

Findings

#	Severity	Finding
1	nit	The E2E does not kill a mutant that sets position: sticky instead of relative (sticky is also a containing block, so the geometry assertions still hold). The unit test catches it, since it asserts exactly relative. Acceptable, only noting that the E2E checks geometry, not the specific value [T].
2	nit	The report says it is unconfirmed whether CI runs test-table-fluid-e2e.js. It does: deploy.yml invokes it by name in the Playwright E2E job [T].
3	nit	The unit test fake for getComputedStyle reads `th.style.position
4	info	Behaviour change beyond analytics: makeColumnsResizable is shared, so Nodes, Observers, Packets, GPS sanity, node-changes, new-nodes and position-gaps tables get the same anchoring. This is the intended effect of fixing the shared helper, not scope creep. .data-table th is already position: sticky, so those headers are skipped by the conditional [T/A].

No blocking findings.

Answers to the review points

• Adapted, not blind cherry-pick: two-line change in the shared helper in public/app.js, tests in the fork's repo-root layout, getComputedStyle stub added to the makeColumnsResizable measures once and credits colspan rows to column 0 (packets Details column too narrow) #258 sandbox, E2E waits on this fork's theme-refresh event [T].
• makeColumnsResizable measures once and credits colspan rows to column 0 (packets Details column too narrow) #258 column widths / existing layouts: test-issue-258-column-widths.js 15/15, test-issue-258-column-widths-e2e.js 22/22, test-table-fluid-e2e.js 43/43 on the merged tree; the sticky-header case is preserved by the conditional [T].
• CSS variables only: no CSS changed, no colours added. scripts/check-xss-sinks.sh --diff origin/master run in a git worktree of the head: exit 0 [T].
• Mobile/touch: E2E covers 375px (handles on the header edge, no page overflow). The handle has mouse listeners only for dragging; this is unchanged by the PR and out of scope [T/A]. Real touch drag was not tested.
• Tests exercise real code, mutant meaningful: the unit test loads the real public/app.js and calls makeColumnsResizable; the E2E uses the real page. Red-before/green-after is documented by the author and reproduced by my mutants below [T].
• Acceptance criterion: the single criterion of fix(analytics): anchor column resize handles to table headers (port of upstream PR 2274) #596 (handles anchored to headers, desktop and mobile browser test) has a unit test and a desktop+mobile E2E, both red without the fix [T].
• Invariants: no Go changes, cmd/server untouched, no new map[string]interface{}; fork guards unchanged (9 if: guards plus 1 comment line); no closing keywords (Relates to #596); commit author dborup <kontakt@meshview.dk> [T].

Tests (merged tree)

• cmd/server go test ./...: ok (82s). cmd/ingestor go test ./...: ok (244s) [T]
• sh test-all.sh: 273 passed, 0 failed [T]
• node test-frontend-helpers.js: 731 passed, 0 failed [T]
• E2E against a local Go server with e2e-fixture.db, prepared as in CI (freshen, seed SQL, corescope-migrate, seeds 2073 and 199): test-table-fluid-e2e.js 43/0; test-issue-258-column-widths-e2e.js 22/0; test-e2e-playwright.js 154/157 passed, 3 skipped, 0 failed [T]
• CI per job on the head: all green (Go Build & Test, Go Libraries & CLI, Go Ingestor, Go Server 1/3-3/3, JS Unit & Lint, Script & Container, Verify Go Test Assignments, Validate Docker Image, Playwright 1/3-3/3, Verify E2E Shards); deploy/release/badges skipped by design. Known flaky flaky: #226 Hash Stats adopters sort E2E — server sorts multiByteNodes on packets only, ties come out in random order #256/flaky: TestResolvedPathBackfill_WriteHoldUnderBudget_188 — wall-clock write-hold budget fails under CI load #267 did not fail [T]

Mutants (my own)

1. Fix removed: unit 2 failures; E2E both fix(analytics): anchor column resize handles to table headers (port of upstream PR 2274) #596 cases fail (1250px and 375px) [T]
2. Unconditional position = 'relative': unit 2 failures [T]
3. Fix applied only to the first header: unit 2 failures [T]
4. relative replaced by sticky: unit fails; E2E survives (finding 1) [T]

Not verified

• Visual check by screenshot and real touch drag on a device.
• Staging/production behaviour (no access, by rule).
• Pixel-level comparison against upstream 2274 (upstream read only via the PR description).