T128: column flex automatic minimum size measures content with the item height as auto (Blink IntrinsicBlockSize)

[thejackshelton]

What changed

T128: in a column flex container, an item's automatic minimum size (css-flexbox-1 §4.5 content size suggestion) is now measured with the item's own height treated as auto, as Blink does (flex_layout_algorithm.cc at 145.0.7632.6, BSD: lines 914-923 and 1117-1120, LayoutResult::IntrinsicBlockSize). Before, the engine measured content with the item's height applied, so an empty height: 11px item could not shrink below 11px where Chrome shrinks it to 0 (found by CALC-2).

• packages/layout/src/flex.ts: columnIntrinsicBlockSize lays the item out with height: auto for the content suggestion. A percentage height, min-height, max-height (or flex-basis inside a column flex item) on a child of such an item is refused as percent-height-flex, since Blink resolves it against the item's set height, which an auto-height measurement cannot reproduce.
• Blink's flex-shrink 0 shortcut (flex_layout_algorithm.cc lines 1078-1084): an item that cannot shrink takes its specified size without measuring content. The used size is the same; it avoids 4 needless percent-height-flex refusals.
• New fixture values-calc-flex-column-auto-min (values group, ltr and rtl): fixed column containers shorter than their items, px and calc() heights, empty and non-empty content, padding/border, max-height, nested column flex, and a row control. Chrome captures from pnpm run parity:capture and pnpm run parity:dpr-capture.
• Regenerated outputs in their own commit (bdec838), whose message names every command. Every existing vector and capture is byte-identical; emitted CSS changes only in the compilation-digest header.

What passed

Caught up with master before review: merges of origin/master up to the train 1 freeze base 1f17b6e (#64 SIZE-ar, #65, #66 INL1a breaks, #67 REPL-a engine, #68 T116). The REPL-a merge conflicted in flex.ts buildItem: master's ratio and replaced-item branches stay, and T128's column suggestion (columnIntrinsicBlockSize) applies to a sized box item; a replaced item keeps contentMain. docs/ports.json's flex_layout_algorithm.cc entry lists master's ratio.ts port and T128's two flex.ts ports. /tmp/heavy-lease.sh pnpm regen rebuilt the generated outputs in their own commit (ca22af3, fixed point after 3 passes).

On ca22af3:

• pnpm typecheck: pass.
• Targeted: engine.test.ts, vectors.test.ts, values.test.ts, chrome-ports.test.ts, regen.test.ts: 590 pass, 1 skipped (the DRAGON_REGEN_CHECK suite).
• /tmp/heavy-lease.sh pnpm test on ca22af3: queued behind the other members' regens; the result follows in a comment.
• Pre-PR audit (AGENTS.md step 3) of flex.ts, engine.test.ts, the fixture and values.ts: no instance found. The percentage check covers only the item's direct children by design: a deeper descendant's percentage resolves against an intermediate box, and a stretched-size basis is already refused as percent-height-flex by the real layout pass.

Train 1 (freeze base f78f198 / 1f17b6e)

This PR is a member of merge train 1 (AGENTS.md step 8), in this order: #59 T128, #60 CALC-2, #61 T130, T129, seld-r1. The train merges each member onto the previous position, so members that touch the same lines are stacked, with their overlaps resolved and reviewed in the member branches rather than in the train: CALC-2 merges T128, T130 merges CALC-2, and T129 merges T130. Until an earlier member lands, GitHub shows its changes in this diff too. seld-r1 sits on master alone; its merges onto the earlier positions were simulated clean, and every position's reviewed-path patch id equals its member's clean head (git merge-tree chain plus patchIdOver), so each position should be vouched at landing.

The device lanes, and the regen of each position, run in the train itself: one device run per position. Master now has T116 (#68), so device-pixels is iOS 57 / Android 86 on master; each position's device run is compared against that.

Random corpus: 136 geometry moves and 35 new percent-height-flex refusals, all on cases without a Chrome reference; refusing is accepted (follow-up T136).

Changed tests and checks

• packages/layout/test/engine.test.ts: new describe block. It pins the c1 and c2 fixture cases in LU, the percent-height-flex refusal, and the fact that a row item's percentage flex-basis child is not a height.
• packages/parity/src/fixture-groups/values.ts: adds both('values-calc-flex-column-auto-min').
• No tolerance, check or test is loosened or removed.

🤖 Generated with Claude Code

Note

Add column flex automatic minimum size support via columnIntrinsicBlockSize

• Adds columnIntrinsicBlockSize in flex.ts: for column flex items with a specified height, it clones the item with an auto height, lays out contents, and uses the resulting content-box height as the content size suggestion.
• Updates buildItem in flex.ts: non-shrinkable column items with a specified main size now use that size directly as the minimum.
• Rejects percentage height, min/max height, and applicable percentage flex-basis inputs inside column intrinsic measurement with the percent-height-flex unsupported code.
• Adds engine tests in engine.test.ts and a parity fixture with generated snapshots, break vectors, and expected pixels in LTR and RTL at multiple DPRs; regenerates Swift and Kotlin layout artifacts.
• Registers the values-calc-flex-column-auto-min case in the iOS, Android, and web support profiles.
• Behavioral Change: column flex items with specified heights now size differently than before — content-based minimum sizing replaces the generic content measurement path in buildItem; percentage-height configurations return percent-height-flex unsupported results.

^{Macroscope summarized 2d47e5d.}

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — The PR changes existing column-flex sizing in the production layout engine and adds a new unsupported outcome for percentage-height descendants, with an additional intrinsic layout pass. These compatibility and runtime effects extend beyond a simple isolated fix and merit human review.

Notes:

• No code objects were reviewed. Approvability was decided on eligibility alone.

^{You can add or adjust custom eligibility rules. Learn more.}

Dismissing prior approval to re-evaluate 2d47e5d