Grid: skip contribution measurement when no spanned track can receive it

[nicoburns]

Objective

In resolve_intrinsic_track_sizes, the loops for "2. content-based minimums", "3. max-content minimums" (both the max-content-constraint variant and the unconditional one), "5. intrinsic maximums" and "6. max-content maximums" call item_sizer.min_content_contribution / max_content_contribution for every item in the batch, and only afterwards filter tracks via the track_is_affected predicate inside distribute_item_space_to_*. Only step 1 (minimum_contribution) is gated on the spanned tracks.

So an item spanning only tracks the step cannot apply to (e.g. 12.25rem minmax(0, 1fr)) still gets measured under a min-content constraint. That measurement is a full subtree measure and is then discarded. On Wikipedia's Vector skin (.mw-page-container-inner { grid-template-columns: 12.25rem minmax(0, 1fr) }) this measures the entire article under MinContent, inside which .mw-body's minmax(0, 59.25rem) column legitimately collapses to 0, so every inline layout in the article is line-broken at width 0 (one word per line) for nothing.

This PR moves the same predicate in front of the measurement:

for item in batch.iter_mut() {
    if !item.spans_track_matching(axis, axis_tracks, has_min_or_max_content_min_track_sizing_function) {
        continue;
    }
    let space = item_sizer.min_content_contribution(item, axis_tracks);
    ...
}

Track sizing results are unchanged (the distribution step would have been a no-op for these items). All generated fixtures pass. New test grid_item_spanning_only_unaffected_tracks_is_not_measured: 3 measurements on main → 1 (the final layout) with this change.

Measured on a Blitz headless load of the Barack Obama Wikipedia article: zero-width break_lines passes 916 → 3, first layout 845–862 ms → 592–626 ms, relayout 737–758 ms → 473–502 ms, live heap after layout 88.6 MB → 63.0 MB, layout geometry unchanged.

Second commit: expand_flexible_tracks max-content measurement

Skipping the step-2/3 measurements exposed a latent bug: expand_flexible_tracks (11.7, indefinite free space) measured each item's max-content contribution with Size::NONE for both grid area and available space (there was a TODO about it). max_content_contribution_cached ignores the sizes passed to it, so previously the cache was always pre-filled by steps 2/3 (which measure with the real grid-area size) and the Size::NONE never took effect. With the skip, the flex step became the first measurer for items spanning e.g. 50px / minmax(10px, 1fr), text was measured at an unconstrained width, and 1fr rows came out one line tall — WPT css/css-grid/grid-definition/flex-content-resolution-rows-002.html went 21/21 → 18/21 in Blitz.

Fix: route the flex step through IntrinsicSizeMeasurer::max_content_contribution like every other contribution step (this also includes the item's margins, as the other steps do). New Chrome-generated fixture grid_fr_row_max_content_uses_column_width (fails on main + first commit alone, passes now). WPT css-grid/css-flexbox/CSS2 in Blitz: no regressions vs Blitz main, +2 newly passing.

Context

Blink does the same: GridTrackSizingAlgorithm::IncreaseTrackSizesToAccommodateGridItems collects sets_to_grow via IsContributionAppliedToSet and continues if empty before calling contribution_size(...); additionally ResolveIntrinsicTrackSizes only considers items with IsSpanningIntrinsicTrack. Spec: css-grid-2 §12.5 scopes each step to "items spanning tracks with a min/max track sizing function of …".

Per-step predicates used:

• step 2: min sizing fn is min-content/max-content
• step 3 (max-content constraint): min sizing fn is auto (and max not min-content) or max-content
• max-content minimums (all cases): min sizing fn is max-content
• step 5: max sizing fn has no definite value
• step 6: max sizing fn is max-content-alike (or percentage with indefinite container)

Related: DioxusLabs/blitz#835 repins Taffy to this branch and clamps negative inline available widths (a downstream symptom of the same discarded pass).

Link to Devin session: https://dioxus.staging.devinenterprise.com/sessions/3a0b8cc210c041d8955c841fb2a76a8e
Open in Devin Desktop: https://dioxus.staging.devinenterprise.com/desktop/session/3a0b8cc210c041d8955c841fb2a76a8e?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