Flexbox: use hypothetical main sizes for column intrinsic contributions

[nicoburns]

Objective

Make column flex container intrinsic main sizing match what browsers actually ship, and replace the misleading comment about the row/column asymmetry with the real explanation.

As established in #351 (comment), browsers use two different item-contribution models depending on the axis:

• Row (inline axis): the current spec's §9.9.3 intrinsic item contributions (content-based, capped/floored by flex base size depending on grow/shrink), as shipped in Chrome 136.
• Column (block axis): the 2015-era rule — each item contributes its outer hypothetical main size (flex base size clamped by min/max main sizes, including the automatic minimum size). Grow/shrink factors and content (except via min-height: auto) are irrelevant.

Taffy previously approximated the column behavior with a content-based contribution floored by the flex basis:

// before (column branch)
(inner_main_size.max(item.flex_basis) + margin).maybe_clamp(style_min, style_max)

// after (column)
let hypothetical_inner_size = item
    .hypothetical_inner_size.main(dir)  // flex_basis clamped by resolved min (incl. auto min) / transferred sizes
    .maybe_min(style_max)
    .max(item.resolved_minimum_main_size);
hypothetical_inner_size + margin

This coincides with browsers whenever the automatic minimum size floors the item at its content size, but diverges when it doesn't. Distinguishing case: column container, item with flex-basis: 200px; min-height: 0, content 300px high — Chrome gives the container height 200, taffy previously gave 300.

Context

• The column branch now short-circuits before the content-measurement match, so for column containers the intrinsic pass no longer measures item content at all (the auto minimum size was already computed and cached in determine_flex_base_size). The row path is unchanged apart from de-nesting the now-row-only match.
• Three new gentest fixtures cover the distinguishing cases: min-height: 0 disabling the auto-min floor (expect 200), the default min-height: auto floor (expect 300), and max-height capping the flex basis (expect 100). All expectations are Chrome-generated.
• The old comment attributing the asymmetry to "the automatic block size of a block-level flex container is its max-content size" is replaced with the actual history: Chromium migrated only the row axis to the new algorithm after web-compat experiments (crbug 40077556), and the spec is expected — though not certain — to be updated to match (csswg-drafts#8884).
• Follows on from Flexbox: apply main-axis margin after flex-basis floor in intrinsic contributions #1164; relates to Implement "§9.9 Intrinsic Sizes" of the Flexbox Specification #351.

Feedback wanted

hypothetical_inner_size as stored by determine_flex_base_size is not clamped by the style max main size (only by the resolved minimum and aspect-ratio-transferred sizes), so the column branch applies .maybe_min(style_max) plus a re-floor by the resolved minimum (min wins over max). If that max-clamp should instead live in determine_flex_base_size itself, happy to move it.

Link to Devin session: https://dioxus.staging.devinenterprise.com/sessions/f379d8c9a11e428a92fe6198951c5e79
Open in Devin Desktop: https://dioxus.staging.devinenterprise.com/desktop/session/f379d8c9a11e428a92fe6198951c5e79?variant=devin-insiders
Requested by: @nicoburns

[staging-devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR that start with 'DevinAI' or '@devin'.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring