ai-app/docs/HANDOFF.md

Handoff

Where the work in flight stands for a session picking it up cold. Keep current invariants, measurements, and failed hypotheses here; this is not a decisions log. Pruned on 2026-09-18: the frame/extent prototype's chronicle went, its four surviving findings and two failed hypotheses stayed, and the transparent-frames plan became a record of what it landed as.

Where things stand

Canonical Iris main is ca2b4b2 (#17, the headless rig). #18 split/18-position-chain is open in /home/bob/repos/iris-pr18, head e44dea3, pushed. It holds LAYOUT.md §2's position chain, leftover, the Holds retained-layout contract, region nodes, built-in alignment and size rules, fixed-point layout, a box in pixels threaded down the draw, and a report read as a fraction of the containing widget. No PR review was present when checked on 2026-09-15.

The separate frame/extent experiment lives in /home/bob/repos/iris-layout-experiment. The transparent-frames protocol is implemented on wip/transparent-frames, head 49cec82, three commits over 34cafb6 (wip/region-and-placement, the experiment as it was). The app's framework pin is unchanged, and none of this is ready to replace #18: two fuzzer seeds still settle differently warm than cold, and many and resize cost more than #18 on a deep tree. What it does, what it cost and what it found are in Transparent frames: what landed below; read that before anything else here about Span, Pad, Stack, the placement pin or measure_len, all of which it supersedes.

Transparent frames: what landed (2026-09-18)

Decided with Bryan on 2026-09-18 and implemented the same day. His rules are kept below because they are the design; the protocol section says what the code now does, and What it cost and What is not done, and why say where it stands. The step list the plan carried is gone: steps 1 to 5 are in 1956be3, the measurements that follow are in 0954770, the review pass is 49cec82, and steps 6 to 9 are the open questions at the end.

Decided by Bryan

• Containers are transparent. The frame a widget's fractions are of is forwarded from its parent through a span, a stack and a scroll unchanged, and through an inset narrowed by its margins. Any number of nested spans lay out against one frame. The reason: px already passes through, rel should behave the same way, and leftover is already the way to say "fill the containing widget", so rel(1.0) meaning that too would be two ways to say one thing.
• A frame is narrowed only by what is decided from above: a declared length on the widget (.width(rel(0.5)) on a span narrows its children's frame too), an inset's margins, the root. A box that is an answer (a row's height, a stack sized by a child) is never anything's frame.
• rel overflows on purpose when it sums past one or has pixels beside it.
• Inset and Outset take px, rel and leftover margins. Outset adds its margins to the child's report and moves the child in; Inset draws the child in a frame with the margins subtracted and reports the child's size plus what it subtracted. An inset resolves what it subtracts, so px and rel margins narrow the frame symbolically and a leftover margin is resolved in pixels from the extent after the child answers, like a span's shares. A general Pad would outset pixels and inset rel and leftover. Do not build these yet; make them a few lines each to write.
• Along its own axis a span places a child whose length is known in pixels at its final slot while measuring, so it is drawn once, and moves a child that is in the wrong place, px or rel, by translation rather than drawing it again. Across itself it may still re-place by the answer.
• Layers may later be tied to another widget (a popup near an anchor). A position is never defined by two widgets; positions compose up the tree.

The protocol

Two boxes per widget, both in the parent's frame coordinates:

• frame -- what a declared or reported fraction is a fraction of. UiRegion::FULL for a transparent parent; narrowed by a declared length, and that is the only narrowing in the code today. Its length is the same on every ask of the widget, which is the property the whole retained model rests on.
• extent -- where the drawing goes, as a part of the parent's own box.

/// What of a widget's own box a child is given, along one axis.
pub enum Part {
    /// The whole of it.
    All,
    /// Frame lengths from where the box starts, which is what a container
    /// dividing room among its children speaks: a child's report is a length
    /// of the frame, so the cursor that sums those reports is one too. A
    /// moved box re-places every child by re-adding its start, exactly.
    From(UiSpan),
    /// A part of the box in its own coordinates, which is what a container
    /// that insets one speaks: taking eleven pixels off the end needs no
    /// length, where saying the same thing in frame lengths would make the
    /// container read its own box -- and a box chosen from its own answer
    /// then feeds back into the answer.
    Of(UiSpan),
}

/// Where a child goes along one axis, as a part of this widget's box.
pub enum Place {
    /// The child's answer, aligned inside the part by the child's alignment.
    Within(Part),
    /// Exactly the part; the answer is not placed inside it again.
    Fill(Part),
}

Part::Of is not in the plan; it is what the measurements asked for, and the reasoning is in its doc comment above. Part::From is the plan's extent-relative span.

The widget's draw sees only its own box: px_len/px_size are that box's pixels and narrow the extent range, holds widens it. Primitives and masks are written in its coordinates; there is no DrawRegion and no frame-coordinate primitive. A container reads extent_len(axis), the box's symbolic length along one axis in frame units, which pins its drawing to that length and to nothing about the start -- one axis at a time, because a span dividing one of them holds for any length of the other. A span that divides room also reads frame_px_len(axis) (was region_px_len) and widens through frame_holds (was region_holds). placement(), region(), box_of, measure_len, widget_within, DrawRegion and ExtentPlacement are gone.

The parent side is one call, with a shorthand:

pub fn widget_at<'s, W: ?Sized>(
    &'s mut self,
    id: &'s StrongWidget<W>,
    frame: UiRegion,     // in this widget's frame coordinates; FULL forwards it
    place: [Place; 2],
) -> DrawResult<'s, 'a, W>;

/// The transparent default: the frame as given, the answer aligned in the box.
pub fn widget<'s, W: ?Sized>(&'s mut self, id: &'s StrongWidget<W>) -> DrawResult<'s, 'a, W> {
    self.widget_at(id, UiRegion::FULL, [Place::Within(Part::All); 2])
}

Span is the plan's, without the known-length shortcut (step 7, not done): it reads far = painter.extent_len(axis), measures each child at Place::Within(Part::From(along(cursor, far))), and places it at Place::Fill(Part::From(along(from, start))), with Place::Within(Part::All) across itself. Stack gives its sizing child Fill(All) and the rest Within(All); Scroll measures at Fill(All) and places at Fill(From(px content box)); Pad is transparent and insets by Within(Of(..)); Masked sets its mask over its own box.

A report is still LayoutLen { px, rel, leftover } per axis, a fraction of the reporting widget's frame, so it passes a transparent parent unchanged and is composed by within_len(frame.len()) through a narrowing one. A stack sized by a child that reports rel(0.5) therefore reports rel(0.5) and the fraction is applied once -- the parked wip/stack-fraction-twice defect is closed by the protocol rather than by a guard, and placed_extent is where: it takes the answer's length from the part rather than composing it into the part.

Retained state and reuse

Per widget (ActiveData), replacing region/placement/given_region/ offer_len/offer_placement:

• frame: UiRegion in the parent's frame coordinates and frame_abs in parent_move's; extent: UiRegion in the frame's own coordinates, which window_region and recomposition compose as extent.within(&frame_abs). There is no offer_frame: the frame's length is the same on every ask.
• place: [Place; 2] as last given, offer_place from the ask its answer came from, and offer_part: UiRegion -- the box that ask gave it. The box is kept and not worked out again from where the parent's own box is now: a parent drawn again in the box its answer chose gives its children boxes it never measured anything in.
• answer: Option<(Size, LayoutHolds)> from that ask; holds: LayoutHolds for the drawing, where

pub struct LayoutHolds {
    pub frame: [Holds; 2],            // frame pixel lengths
    pub extent: [Holds; 2],           // pixel lengths of the widget's own box
    pub extent_len: [Option<Len>; 2], // its symbolic length, where read
}

The placement: Option<UiRegion> pin is gone. Nothing may depend on where a box starts.

• primitives and the mask retained in the widget's own box's coordinates.

Reuse of a drawing at an ask: same layer, parent move and region-node choice; frame pixels inside frame; the box resolved from place (Fill is the part; Within is the answer aligned in the part, or the part where the answer fills) has pixels inside extent and, where pinned, the same symbolic length. A frame that moved recomposes the subtree from retained local coordinates (recompose_subtree). A box that moved re-places every child through its retained place (reposition, generalised from the old extent_children to all children because every child is now a part of it).

Dependencies composed into the parent (in_parent): a child's frame range goes through the child's frame length into the parent's frame range. A child's own-box range goes into the parent's frame range through the part's length where the part is From (a frame length), into the parent's own-box range through the part where it is Of, and straight into it where the part is All. A pin composes only for All. Answer dependencies come from children whose answer was read, drawing dependencies from every child drawn.

Local redraw (redraw) still defers where the box it is given is not as long as the box it was measured in -- see What is not done, and why.

What it cost

Widget draws / distinct widgets / update, from tests/layout_diagnostics.rs, at depth 8. Draw counts are deterministic, so these are single runs rather than medians; e44dea3 is #18's head and 34cafb6 the commit this branch starts from.

seed 1	e44dea3	34cafb6	here
cold	369/261/10.6	463/274/13.3	516/288/12.0
repaint	1	1	1
many	157/95/0.33	263/108/0.59	187/119/0.52
size	16/12/0.018	3/3	3/3/0.010
scroll	2/0.002	1	1/0.004
resize	13/13/0.019	22/15/0.032	24/76/0.090

seed 13	e44dea3	here
cold	1330/707/20.3	2940/982/28.3
many	524/159/1.09	1091/423/2.39
resize	nothing drawn	2215/510/6.56

So size and scroll keep the experiment's wins, many is better than the commit it starts from and still well short of #18, and resize on a deep tree is where it loses badly: #18 draws nothing at all at seed 13, because every fraction scales and every Holds admits the new window.

Three measured findings, each already applied:

• Lazy Within placement costs more than it saves. The plan's step 5 -- leave a child's answer to be placed at the end of the parent's draw -- puts the drawing in the part first and in the answer's box after, and where it does not hold for both that is two drawings rather than one. Seed 1's resize went from 391 widget draws to 29 with it removed. The test that pinned three draws for a numeric leaf in a span went with it.
• An inset said in frame lengths makes a container read its own box. "Less eleven pixels at the end" needs the length, and a container whose box is its own answer then depends on its own answer: Pad drew sixty-four times in one resize frame at seed 13, chasing its own width. Part::Of says it as a part of the box instead and composes without a length.
• Pin one axis at a time. extent_len pinning both made a span dividing one axis hold for one length of the other, and a resize broke every span whose cross-axis answer moved.

What is not done, and why

Step 6, redraw without the deferral, does not hold. Removing it -- asking the measuring question locally and placing the answer afterwards -- makes seeds 104 (align) and 210 (reorder) at depth 5 settle differently warm than cold. The plan said to stop and report rather than restore it; it is restored, in the form the protocol allows (info.part.size() != offered.part.size()), and this is the report.

What is underneath it. A Place is relative to the box the container is being drawn in, so the same ask expression resolves to a different box depending on which of its own boxes the container is drawing in -- and a container is routinely drawn twice, once in the box its parent measured it in and once in the box its own answer chose. So "the box this widget's answer was measured in" cannot be recovered from the ask, which is why offer_part is retained; and whether a given draw is a measurement cannot be recovered either, which is where the two open seeds live:

• seed 2, repaint, depth 5: Stack > Pad{6,14,16,15} > Span{DOWN} with a wrapping text under the span. The span answers 880 px measured in the pad's first box and 874.24 px measured in the box the pad's answer then chose, and both are fixed points of "measure in the box, place at the answer, measure again". Cold reuses the drawing and keeps the first; a repaint redraws and lands on the second.
• seed 1 at depth 4 and 108 at depth 5, reorder: the same shape through Branch, which chooses a subtree from a measured length, so the two fixed points are two different trees.

Four bookkeeping rules were tried for "which draw is a measurement": the ask's first-ness alone, the place matching the retained offer place, the box matching the box the answer was measured in, and a retained flag on the drawing. Each fixes some seeds and breaks others, which is the signal that the question is the protocol's rather than the bookkeeping's. What the baseline does instead is never act on it: a local redraw whose box is not the offer's defers to the parent, and the parent re-asks from a geometry that a cold layout also reaches.

Also not done: step 7 (Span's known-length shortcut and the cross-axis report in frame pixels), step 8 (Inset/Outset as test widgets) and the LazySpan/SizeRule work that sits on top of this.

Pad is an outset and that is visible. examples/text.rs has wtext(..).width(rel(1.0)) inside a .pad(16): under the old reading that was the padded box, and under transparent frames it is the window, so the labels now sit at the window's edges and the left one is clipped. view and minimal are byte-identical to 34cafb6; tabs moves its red square 3 px; random is a generated tree and moves where nested spans do. The plan's "a general Pad would outset pixels and inset rel and leftover" is the answer to this and is not expressible in the protocol as it stands: an inset that narrows the frame and places the child in a part of its own box needs the child's box expressed in the child's narrowed frame, and (0.5·B - 20)/(B - 20) is not rel + px. Either the frame narrows and the box goes with it (what the code does, so an inset inside a row draws its child across the whole row), or the box is a part and the frame does not narrow (Part::Of, what Pad does), or a third thing Bryan decides.

Review of 2026-09-17

A fresh read of core/src/fixed.rs, orientation/, ui/holds.rs, ui/painter.rs, ui/render_state.rs and the position widgets, outside of doing work on them, with each finding checked by a scratch test.

Verdict. The concepts are sound and stay. Fixed point on a 1/1024 grid is the right base for a layout that decides "same box or not" by equality. Threading the pixel box down the draw, with Holds::through the exact preimage of that one multiply, is the strongest idea in the code: layout has one route to every length and the reuse test is its exact inverse. Offer, given and placed is the ordinary measure-then-arrange model. What needed work was the bookkeeping around the second ask, one boundary in Span computed by an expression other than the drawing it guards, and a rel that meant two things. The two-step residual is structural and no grid width fixes it; where it becomes visible is the shader's snap.

Decided by Bryan

• rel is a fraction of the containing widget's whole area. In a span, rel(0.5) is half the span whatever else is in it and wherever it sits. It is never a fraction of what was left after earlier children. This supersedes the 2026-09-16 reading that a report is a fraction of the box the widget was given, wherever that box is a remainder rather than the child's whole area.
• draw stays the only layout method on Widget (Bryan, 2026-09-17, reversing the same day's acceptance of a measure/draw split). Ease of writing a widget is half the reason. The real one is that a second method holding the same layout drifts from the first, which a span makes extremely easy, and the shared logic then gets pulled into helpers both call that still have to be applied carefully in each. Where the framework needs a widget's layout twice it runs the same body again with the painter in a different state, or hands it more through the painter.
• A widget's frame does not change between the ask that measures and the ask that places. Fractions are of the frame; its own box reaches the widget through the painter. This is what transparent frames implements.

A report is a fraction of the containing widget (landed, ffd79f3)

rel(0.5) is half the span whatever else is in it and wherever the child sits. A report used to come back composed through the box it was offered, and a span offers each child the room from its cursor, so a nested span taking half of what it was given took a quarter of a row whose first half was spoken for -- where the same half written as a rule on the child took half the row.

The ask carried the base separately (reports_of) for a day; transparent frames replaced it with the frame, which is the same statement made once per widget rather than once per ask. The offer is still the remainder, because a text has to wrap at the width actually there.

A span can now overflow itself without bound, which is the consequence Bryan's rule asks for: two children reporting half each take the whole row and a third starts past the end. Under the old reading total.rel could not exceed one, so Span's fixed <= 0 branches were only reachable through declared fractions; they are ordinary now, and with them boxes of negative length passed down to children.

An answer is not an answer while anything under it is dirty (0e0d4af, superseded by a0693ac)

Superseded: dirty_size_under is deleted. Once a resize goes through the settling walk the ordering makes the whole category unreachable rather than checked, which is Bryan's steer and the better answer. The defect and its reasoning are kept below because they say what the ordering is buying.

draw_inner took an answer from try_reuse, which checks only whether the widget itself is marked, where retained_answer beside it also refused one while anything the widget read a size from was dirty. A widget whose drawing happened to be reusable therefore handed back the answer it gave before that descendant changed, and nothing puts it right: the comparison that tells a reader its child's answer moved is in redraw, and a widget settled inside its parent's own draw never goes through it. The placing ask redraws the subtree, the descendant's mark is cleared there, and the parent keeps a number the tree no longer agrees with.

So dirty_size_under was not the optimization its comment claimed: it was what made an answer an answer, until the walk made the state it guarded against impossible to be in.

Found at seed 564, depth 6, shuffle-every-other, reachable only once a span could overflow itself. No hand-built tree ever reproduced it, and the seeds at depth 4 above fail for some other reason.

A text is handed back a box its own line fits in (landed, 4bd8607)

A wrapping text reported the width it used through Px::from_f32, which takes the nearest step and is under the line the shaper measured half the time. A parent that sizes itself from that report -- a stack taking a span's width, the span taking its widest child's -- then hands the text back a box its own longest line does not fit in, and a greedy break there is one line longer. Warm kept the break it had; cold made the narrower one.

Two tolerances were holding that together and both are gone:

• TextBuffer::shape answered any width within BREAK_EPSILON_PX = 0.05 of the longest line from the break in hand. Fifty steps of the grid, and a structural decision taken on a hair's breadth -- the thing the Span boundary invariant below already forbids. It is want >= layout.width() now, exactly.
• The Holds range the text declares started at the nearest step to its longest line, so it admitted boxes that line does not fit in. It starts at Px::ceil_from_f32 of it now.

Neither was the fix. The fix is the report: Size::from_px( PxVec2::ceil_from_f32(tex.size)), the step at or above what was measured, so the box that comes back fits. With it in place either tolerance could have stayed and the case passes; both are wrong on their own terms, so both went. Fixed::ceil_from_f32 is new and is the only rounding on the grid that is not to the nearest step.

The general shape, and the third time this branch has hit it: a value that comes back as a box has to be rounded away from the measurement, not to the nearest step. Rounding to nearest is right for a value being carried; it is wrong for a bound.

A frame settles strictly bottom-up (landed, a92c6ac)

The queue was already deepest-first, but a widget that could not settle where it was called redraw on its parent from inside itself, which drew a shallow widget while dirty widgets deeper in other subtrees were still pending. A parent drawing over a subtree that has not settled reads answers about to move, and the one that settles does so inside the parent's draw -- where its mark comes off and nothing compares what it now answers.

A widget that cannot settle defers instead: it marks its parent, stays marked, and waits in deferred until the walk reaches the parent's depth, which cannot happen before everything deeper has settled.

loop {
    let next = rsc.widgets().needs_redraw.iter().copied()
        .filter(|id| !self.deferred.contains(id))
        .max_by_key(|&id| self.depth(id));
    let Some(id) = next else { break };
    if !self.redraw(id, rsc) {
        self.deferred.insert(id);
    }
}

Bryan's, 2026-09-17, and the right answer where 0e0d4af was a check: "then that entire category of issue can't even occur".

The walk is sound on its own, and the second entry point is closed (a0693ac). By induction on depth: when a widget at depth d draws fresh, every dirty widget deeper has been popped, so each is settled or deferred, and a deferred one has marked its parent. A clean child asked by that draw therefore has a clean subtree, because anything dirty under it would have a dirty parent, and so on up to the child itself.

The entry point that was left was update drawing the root for a resize before the walk ran, top-down over a tree with dirty widgets still in it. It is closed by marking the root instead, so layout is one walk a frame and dirty_size_under is gone at both call sites. The root is marked only where the new output falls outside what its answer holds for: that range is the intersection of everything under it, so admitting the new output says the whole tree stands, and nothing above the root moved. Marking it unconditionally cost the root its own Holds -- a leaf root that scales with its box was drawn again on every resize, which two tests caught.

The rig says the cost is scheduling only: on resize, one queue pop, one local redraw and one depth read appear and one failed reuse attempt goes, and every other counter on cold, repaint, many, size, scroll and resize is identical.

A subtree that changes hands is recorded on both sides (landed, e44dea3)

A subtree can be reused whole under a different parent -- same box, same layer, same region node, clean -- and nothing in the drawing says it moved. Two things read who its parent is, and both were wrong after one of these.

• The old parent still listed it, and a parent's next draw undraws whatever is missing from that list. Two spans under one root, with the root swapping which of them it holds, drew the subtree under the new span and then erased it when the old one drew.
• Its depth was the one it had under the old parent, which is what the settling walk orders by, so a change made under it afterwards settled at the wrong point in the frame.

Both are written where draw_inner already records what the ask decided: active.parent is replaced and the old parent's children repaired, and try_reuse re-walks the subtree's depths -- only where the top of it moved, which is what makes that free in the ordinary case. Pinned by retained::a_subtree_that_changed_parents_is_not_undrawn_by_the_one_it_left and ..._settles_at_the_depth_it_moved_to; each fails without one half.

The fuzzer never re-parents (reshuffle only trades children between a span and its own spares), which is why nothing generated reached either.

A span's leftover boundary is its own inverse (landed, 53b00c6)

The decision used a rounded division where the room the children get is a floored multiply, so the boundary and the drawing it guarded were two expressions for one length. room is that length as a Len, room.to_px is the multiply, and Holds::through is its exact preimage:

let room = Len::rel_max() - Len::from_parts(total.rel, total.px);
let mut shares = false;
if total.leftover > Weight::ZERO {
    shares = room.to_px(painter.px_len(axis)) > Px::ZERO;
    let holds = match shares {
        true => Holds::from(Px::STEP..=Px::MAX),
        false => Holds::from(Px::MIN..=Px::ZERO),
    };
    painter.holds(axis, holds.through(room));
}

The three branches were the sign of 1 - rel, which through reads already. Forty lines became twelve and one div left layout. The general rule stands and is now demonstrated: derive a boundary through the inverse of the expression that draws, never by a second expression for the same length.

Two branches parked, one of them now closed

wip/stack-fraction-twice is closed by transparent frames. A stack sized by a child that reported a fraction applied that fraction twice -- half a row became a quarter -- because the placing ask resolved the fraction in the box the answer had already chosen. Under the protocol above a report is a fraction of the frame and placed_extent takes it from the part rather than composing it into the part, so it is resolved once. No oracle could see it (warm and cold shrank alike), so the test that came with the branch is what pins it.

wip/padding-outset-and-inset is superseded by the Pad question at the end of "What is not done, and why". What it got right is kept: padding goes outside what it pads, and an Inset is a separate widget. What stopped it -- a child declaring rel(0.5) under an Inset coming out 47.5 px of the 190 inside -- is the same second application of a fraction, and the protocol removes it.

Span's leftover boundary is a third expression for the room

The decision uses a rounded division, total.px.div(fixed), while the room the children get is a floored multiply, so the two disagree at the boundary. Measured with 300 px, rel(2/3) and a leftover child:

row width	leftover child	its threaded length
900.000	undrawn
900.001	drawn	0 steps
900.002	drawn	0 steps

Harmless at two steps, but the three-branch block collapses into the inverse that already exists. room is computed a few lines below the block as Len::rel_max() - Len::from_parts(total.rel, total.px), and its to_px is exactly the threaded length the leftover children share:

let room = Len::rel_max() - Len::from_parts(total.rel, total.px);
let mut shares = false;
if total.leftover > Weight::ZERO {
    shares = room.to_px(painter.px_len(axis)) > Px::ZERO;
    let holds = match shares {
        true => Holds::from(Px::STEP..=Px::MAX),
        false => Holds::from(Px::MIN..=Px::ZERO),
    };
    painter.holds(axis, holds.through(room));
}

through already handles a negative fraction and a zero one, so the fixed < 0 and fixed == 0 branches go with it. The general lesson: mul floors while div, div_int and ratio round to nearest, so a boundary derived with a division guards a drawing made with a multiply. Derive boundaries through through, or make the grid floor everywhere.

Where the residual comes from, and the snap

The e44dea3 baseline's two-step allowance came from inverse remapping and alignment composed by different routes. The frame/extent continuation below retains local widget frames as well as local primitive coordinates, recomposes in the same order as a cold draw, and removes inverse remapping. Its oracle requires exact pixel-region equality. That experiment is not yet the pinned framework; the baseline's snap decision below still applies there.

Where it does matter is snap_floor in prelude.wgsl, which adds half a layout step before flooring: that absorbs float error and not a layout step, and truncation makes "one step under an integer" the common residue. A third of 900 px is 299.999 on the grid and lands at 299 on screen, which is the pixel 08c9d5a moved tabs's arcs by. Rounding to the nearest pixel absorbs both the truncation and the two-step residual everywhere except within two steps of a half pixel, where layout never lands on purpose, and keeps integer widths for equal fractional parts:

fn snap_floor(v: vec2<f32>) -> vec2<f32> {
    return floor(v + 0.5);
}

Approved by Bryan on 2026-09-17, together with rounding on the CPU; see item 5 under "Next" for why both, and what each does not fix. The check is the reference render set plus the oracle; expect tabs to move its arcs back.

Smaller items

• The comment on the local == UiRegion::FULL shortcut in widget_at says composing through FULL "is not quite the identity in f32". On the grid it is exact; the shortcut is performance only now.
• An undrawn leftover child still contributes its gap, so a vanished child leaves a double gap.
• Nested spans pass leftover weight up, so three leftover children in one inner span beside one in another get three quarters to one quarter. No other layout system does that, and the doc's old example of two and two did not distinguish it from per-span division. Confirm it is wanted.
• Fixed::div by zero answers MIN/MAX while ratio answers ZERO; both are caller bugs under debug_assert, but the fallbacks differ.

Before transparent frames, and what survives from it

The frame/extent prototype that 34cafb6 is the head of separated a widget's fraction reference from where its drawing sits, and the transparent-frames protocol above is that idea finished. Four of its findings still hold and are why the code is shaped this way:

• An answer and a drawing each retain their dependencies. A child's size answer constrains its parent only when the parent reads that answer; the child's drawing constrains the parent's retained drawing whether its size was read or not. Painter collects the two separately (answer_under and under), which is what stopped a stack sized by one child from remeasuring because an unmeasured overlay wrapped at another width.
• A wider contract does not invalidate an existing guarantee. When a local redraw comes back with the same size under a contract covering the old one, the old one is kept; comparing whole contracts by equality doubled scroll cycles.
• No measurement is different from a measured zero: ActiveData::answer is optional, and a previously undrawn pure share must not answer with the placeholder zero it never gave.
• The settling walk takes the deepest mark from a BTreeSet keyed by depth, and what ends it is the mark set rather than the queue.

Two failed hypotheses from it, kept because they are cheap to repeat:

• The placement pin was blamed for the many gap and is not the cause. Disabling it (unsound, a bound) still redrew 487 distinct widgets a frame at seed 13 against e44dea3's 159. What the per-widget trace showed instead was local redraws deferring to their parents and the deferrals chaining to the root -- 43 deferrals in one frame, in chains such as 424 → 425 → 437 → 449 → 483 → 487 → 491, because a span handed its children its own placement as their frame and that placement changed between the measuring ask and the placed one. Transparent frames is the answer to that and it works: the frame no longer changes under a widget.
• wip/local-reask re-asked a dirty widget at its offer instead of deferring, under the old protocol, and diverged at seeds 532 and 398 of depth 6. The same shape is what "What is not done, and why" is about; the branch is superseded and can be deleted.

How layout is decided

Fixed point

Decided with Bryan on 2026-09-15. Layout decides on a grid rather than in floats.

• Fixed<SHIFT> is an i32 counting 1 / 2^SHIFT. Adding and subtracting are exact; mul drops to the step below (Bryan, 2026-09-16: truncation is preferable); div, div_int and ratio round to nearest; to_scale takes the nearest step. Two routes to one place that land on one number are the same place, so everything downstream compares for equality.
• Px is 1/1024 px, Rel is 1/2^24 of a box, Weight is 1/65536 of a share. PX_SHIFT and REL_SHIFT are the only statement of the first two; the shader's copy is prepended from them by render::module_source. Px was 1/64 first, where one rounding's residue was 0.016 px and enough to move a box. Range is +/-2.1M px and conversion to f32 is exact to 16,384 px.
• A weight is not a fraction: a list divides its room by the total of its weights, and Rel::ratio turns two weights into a share on the finer grid.
• Arithmetic wraps (4febabf, Bryan: a coordinate past the range will not draw reasonably anyway, so wrap and break clearly). Saturating cost a twelfth of layout's instructions. MIN and MAX stand in for an unbounded end and are only ever compared against; from_f32 is the one operation that clamps, and Holds keeps a saturating narrow.
• A pointer, a wheel notch, a shaped glyph advance and a window size arrive as floats and go on the grid where they arrive. Vec2 is what the GPU and the platform speak; PxVec2 is what layout decides in.
• Do not widen the grid to chase a residue. Every failure this branch saw was one value reached by two expressions, sitting on a boundary defined by the same value coming back the other way. No precision shrinks a residue that is the whole distance.

A box in pixels is one multiply from its parent's

ActiveData keeps a widget's box as lengths of its parent's box -- given_len, and offer_len for the box it was first asked about -- DrawInfo carries the pixel lengths themselves (px, offered_px), and a draw threads them down one Len::to_px at a time: the box its parent gave it, then the part of that box its own answer placed its drawing in, which placed_lens states once for both placed_box and the walk. Painter::px_size and px_len read that value, and UiRenderState::asked_px takes the same steps back up the parent chain when a local redraw starts part-way down the tree. Neither chain has a coordinate frame in it, so a region node cannot break either, and warm and cold reach every length by the same expression.

• Holds::through is the exact preimage of px + floor(rel * box): floor(rel * B) >= lo - px is rel * B >= (lo - px) << REL and floor(rel * B) <= hi - px is rel * B < (hi - px + 1) << REL, two div_towards once the sign of rel has said which bound is which. The answer is an interval even for a single length, because a floor is not invertible. The range has to contain the box a drawing was made in (the Holds assertion in draw_at, debug only) and must not contain a box the drawing does not hold for (the oracle); being the preimage makes those one statement rather than a trade-off.
• Symbolic regions are for the GPU, hit testing and remaps alone. Moves::resolve is the only walk left and it is the vertex shader's. Nothing layout decides is composed back up the move chain.
• px is not stored on ActiveData, deliberately. A resize every widget's Holds admits redraws nothing, so a stored pixel length would be stale on every widget in the tree with nothing to say so. asked_px walks up only where a widget is already being redrawn; the mean chain is 2.8 levels.
• The window is not a move entry (5b78002). A chain bottoms out in MoveIdx::NONE; the window is applied where a fraction becomes pixels, to_px(output_size) on the CPU and the uniform in the shader. A resize rewrites no retained entry and re-uploads nothing but the uniform; its cost is whatever Holds redraws.
• Failed hypothesis, kept as the shape of the mistake: an offer composed back up the chain fell back to FULL under a region node and was resolved against that node's placed box, so everything under a Scroll was re-asked at the content's width and confirmed its own answer. Pinned by unsettled::a_widget_under_a_region_node_is_asked_in_the_box_that_node_was_offered. The old chain with an allowance in through passed that case and the old chain with the exact through failed it; both halves had to land at once.

What the fuzzers tolerate

The frame/extent continuation removes AGREE_STEPS: warm and cold pixel regions must compare exactly. This is distinct from the equal-share test: when a row's grid-step count is not divisible by the number of children, individual share widths can differ while every rerun of that layout must still agree exactly. The PR #18 baseline still allows two position steps; its earlier failure at one step (resize-size, seeds 384 and 162 at depth 5) is a useful regression target for the experimental recomposition.

Retained-layout invariants

• Holds is the interval of box lengths for which a widget's drawing and reported size stay valid. Reading Painter::px_len or px_size narrows it to the length read; Painter::holds widens it. Parent validity is the intersection of what its children induce: measured children for an answer, all painted children for a drawing. The contract is trusted: a widget declaring a wrong range is a defective widget, and Iris adds no defensive work to recover from one.
• A retained drawing can be reused only when its Holds contains the new pixel box on both axes, its parent node is unchanged, its region-node choice matches the retained structure, it is on the layer it is asked for, and the widget is clean. A valid ordinary subtree moves without redrawing by recursive remap; a region node moves by one entry.
• A retained drawing belongs to the layer it was made on. A container that measures a child by drawing it measures on the layer that child will draw on -- Painter::child_layer_at -- or it pays two draws a frame.
• The first box a parent asks about is the offer; a later box chosen from the answer is the final box, not another answer. A dirty widget is re-asked in the box its parent gave it, and only where that box is as long as the offer; anything else is its parent's question, with the mark left on. Lengths and not whole boxes: what a drawing depends on is its lengths, so the same lengths elsewhere is the same question.
• An answer is reusable where its measurement contract holds. Its final drawing is checked independently and may need redrawing even while the answer stands. Drawing validity is translated back through the chosen placement for the parent's drawing contract, not intersected into the retained answer.
• A widget's frame passes through every container that only divides room, and its length is the same on every ask. What narrows it is decided above the widget: a declared length, and an inset once there is one. A box that is an answer -- a row's height, a stack sized by a child -- is never anything's frame. Painter::widget_at(child, frame, [Place; 2]) says both things about an ask: what the child's fractions are of, and what of this widget's own box the drawing takes.
• A fraction is resolved once, against the frame. A report comes back raw and is composed into the parent's frame by within_len only where the parent narrowed the frame; placed_extent takes the reported length from the part rather than composing it into the part.
• Nothing may depend on where a box starts. A container reads extent_len(axis) for the length it divides, which pins that length symbolically, and places children as parts of its own box, so moving the box re-places them without drawing anything again.
• A widget that clips to its box reports its box: Scroll and Masked report LEFTOVER on both axes, and a debug_assert holds any widget that set a mask this draw to it. Overflowing is otherwise ordinary, which is why the assertion is narrowed to mask-setters. Where content shorter than a Scroll's viewport sits is the scroll's own alignment, and its "fits at the start of any box" widening is gated on near alignment.
• A widget's own mask is not the one it inherited. ActiveData keeps both; they differ exactly where the widget called set_mask, which says whose mask a move rewrites, and a local redraw is handed the inherited one. Pinned by retained::a_masked_widget_redrawn_on_its_own_sets_its_mask_again.
• A span is as long across itself as its longest fixed child, unless a rule gives that length outright (Painter::has_exact_size), in which case it does not read its children there at all. Any relative or leftover child makes it report leftover. (The plan's step 7 changes this to the longest px-or-rel child compared in frame pixels; not done.) Do not choose between a fixed and a relative child in pixels at the span's current width: that admits multiple self-sizing fixed points, and generated seed 13 settled differently warm and cold under it. The same circularity is what a cap containing leftover would put into SizeRule::Max.
• Span's leftover/no-leftover split is a strict layout decision, not a rounding tolerance: its Holds range must use the same exact boundary as drawing. A tolerant endpoint retained zero-height children in seed 16; a boundary moved off where boxes land was needed in floats and is not on the grid. A structural decision may not be taken on a hair's breadth that two routes can disagree about. Pinned by unsettled::a_box_that_only_rounds_past_its_fixed_children_leaves_nothing_over.
• A pixel comparison is equality. A length given in pixels is that many pixels wherever it ends up, structurally: Len::within adds a part's own pixels rather than scaling them. Pinned by a_length_in_pixels_is_that_many_pixels_however_it_is_nested. A length given as a share is not: equal shares come out one or two steps apart because positions, not lengths, are what gets rounded, so the row fills and no two children leave a seam (equal_shares_differ_by_at_most_two_steps_and_fill_the_row).
• A move that keeps a box's length is a translation, and exact. A box that changed length re-expresses each part as a fraction of the new one, which rounds. This inverts the float-era rule; tests/cases/drift.rs pins that the grid does not drift either way.
• Scroll must return the answer from the first box it asked about, whether retained or fresh; returning the final placed answer advanced one fixed-point iteration (seed 86). Content that fills the viewport unscrolled is handed back as it came, because the same box written as its own length in pixels does not round alike.
• An asked-but-undrawn size dependency names the widget that asked as its parent (seed 10). Painter records size-dependency edges only when a parent reads a child's size or hint; an undrawn measured child stays recorded so a later change reaches whoever decided not to draw it.
• Dirty widgets settle deepest-first, including during resize. A deferred child leaves its parent marked, so no clean answer can hide an unsettled size dependency. dirty_size_under has been deleted.
• Declared non-leftover lengths are resolved by the parent where the widget is drawn, so a declared-length change redraws the parent. A rule wins on the axis it names and the widget under it never learns of it. A cap may not contain leftover: a cap must read the report, so rule and report are one equation, and a share puts the row's division into it -- the multiple-fixed-point failure again. A cap is pixels and a fraction, which is what Len is.
• Text shaping is retained separately from line breaking; a greedy break holds from its longest produced line through the width it was made at, reported through Painter::holds.
• Region nodes: a node holds a whole UiRegion in its parent node's coordinates, FULL is the identity, widgets opt in with .region_node() or Widgets::set_region_node, and changing it redraws the subtree once. .scrollable() sets it once; raw Scroll::new does not. A removed node's move entry stays alive until every descendant has migrated. Span and Align add no nodes.
• Alignment is one f32 per axis (Bryan, 2026-09-15), default the middle on both because the edges assume a direction. One widget keeps one length per axis; a second length needs a second widget, Wrapper via .wrapper() (Bryan, 2026-09-16, d21a215).

Verification at the current head

At 49cec82 on wip/transparent-frames: cargo fmt --all --check, clippy with -D warnings, 108 suite tests and 20 core tests in debug, the 11 generated cases, and the six rig phases in the table above. The shrinker at 400 trees of depth 5 fails, at seeds 2 (repaint) and 108 (reorder); the long oracle at 1000/6 and the 2000/4 scan have not been run since they would only find more of the same. The five reference renders were taken and compared with 34cafb6: view and minimal byte-identical, tabs 2,332 pixels, text and random as described above. Venus on the RX 7900 XT, confirmed by vulkaninfo --summary in the same session.

At e44dea3, which is what #18 would merge:

• cargo fmt --all --check, cargo clippy --workspace --all-targets -- -D warnings, cargo test --workspace: green, 92 suite tests, 19 core unit tests, 11 generated cases. Only the long runs and the profiling rigs are ignored; no known defect is.
• The release oracle at 100 seeds in 14.2 s, and 120 seeds in debug in 59 s -- the debug run exercises the Holds assertion in draw_at.
• All fifteen shrinker cases at 400 seeds of depth 5 in 57 s, the oracle at 1000 seeds of depth 6 in 143 s, and 2000 seeds at depth 4 over all fifteen cases in 260 s. The last is not routine and should be: it is the only run that has ever found anything past seed 400.
• view, minimal, random, tabs and text byte-identical at 1920x1200 against 25e456e, as is tabs under the recorded replay. random live-resized from 1920x1200 to 1280x800 is byte-identical to a cold 1280x800 render. Rendered through Venus on the host's RX 7900 XT, confirmed against vulkaninfo --summary in the same session -- an llvmpipe fallback makes the same PNG and nothing in it says so.
• All rig counters identical on cold, repaint, many, size and scroll across 25e456e. resize gains one queue pop, one local redraw and one depth read and loses one failed reuse attempt, which is the root going through the walk; drawn widgets, widget draws, draw requests and primitive writes do not move.

Everything above is verification of what was changed, not a claim that the branch is correct.

A claim about a render holds for the commit it was checked at and no further. tabs changed twice across d3b0ebf with nobody looking; take the oracle as the reference and the five renders as a spot check.

Run the long two before believing a rounding change, and run the ordinary suite in debug:

cargo test --release --test generated -- --ignored a_long_run_of_seeds_agrees
SHRINK_CASE=all SHRINK_SEEDS=400 SHRINK_DEPTH=5 \
  cargo test --release --test shrink -- --ignored --nocapture
IRIS_GENERATED_SEEDS=1000 IRIS_GENERATED_DEPTH=6 \
  cargo test --release --test generated -- --ignored a_long_run_of_seeds_agrees

Depth is what finds things, but so is breadth: every late defect before 2026-09-17 surfaced at depth 5 or 6, and the two open ones were found by running 2000 seeds at depth 4, which nothing routine does. Widen one axis at a time and record which.

Performance

Threading a box in pixels down the draw is free on cold layout and 9-13% off the retained paths (2026-09-17). Instructions:u, medians of 21 runs of binaries built in one worktree, seed 1 at depth 8, against 5b78002:

phase	before	after
cold, 200 frames	313.1M	312.9M	-0.04%
resize	408.1M	405.6M	-0.61%
many	1,924M	1,756M	-8.75%
scroll	357.3M	323.4M	-9.49%
repaint	363.3M	315.4M	-13.18%

cold and resize compare directly: all twenty-five work counters are identical. The other three do less work: repaint goes from 23 draw requests and 13 widget draws to 1 and 1, because redraw composes nothing and a widget whose box moved without changing length settles itself instead of escalating.

How to measure here

• Check the work counters before comparing two commits' times. The rig prints drawn widgets, widget draws and primitive writes; a comparison is only worth reading when they match. random.rs's Branch picks a subtree by a measured pixel length, so the fixture's shape moves with the thing measured; Edits::fixed_branches pins it for timing and the oracle keeps measured branches on purpose. A 3x this section once reported was that artifact.
• perf stat in this VM returns garbage readings for both instructions:u and cycles:u, roughly a quarter of the time, off by a factor of five to fifteen. Take medians of nine or more and report how many readings a filter kept. Cycles spread 1-3% between sets of one unchanged binary and 6.7% in the worst; instruction counts hold to 0.02% within a binary and move 0.5% across a rebuild, so build the baseline beside the thing measured and quote a delta. ex_div_busy held to 0.1%.
• What moves cycles is whether UiSpan::within inlines. It is the hottest line in layout; nm shows it as a symbol when it does not. Shrinking its body until the inliner takes it won; #[inline] on the body it had lost 1.5% cycles. Shrink it, do not annotate it.
• Holds::through divides twice per call and accounts for essentially all of a run's i64 divisions: 21.3M cycles of a 500-frame many, 2.8%. The float head divided twice there too.

Tried and rejected, with numbers

• A float reciprocal for AxisRemap::apply_scalar's division: +6% cycles. Holds::through's division has not been tried.
• Branchless shift_round: +6.7% cycles alone, and worse again with the short-circuits removed. Size, not the branch, is what keeps within out of line.
• Removing the per-child hash lookup in remap_subtree: 0.0%.
• Short-circuiting apply_scalar where the fraction is nought or one: +17%.
• Short-circuits guarding a saturating multiply stopped paying once the multiply wrapped. Re-price a short-circuit before keeping it.
• Rust does not contract a + b * c; the float head never had an FMA to compare the grid's multiply against.

Wrapping (4febabf) was -8.6% instructions and -6.6% cycles. Truncating (08c9d5a, with Fixed::scaled's zero test and within's is_full tests removed as one commit, since they are worth 61M instructions apart and 115M together) costs a share a thousandth of a pixel of its row, makes a flipped span sit a step from its mirror, and moved an antialiased edge in tabs by one pixel. See "Where the residual comes from" for what that last one is.

Rigs and reproduction

Ordinary framework verification:

cd /home/bob/repos/iris-pr18
cargo fmt --all --check
cargo clippy --workspace --all-targets -- -D warnings
cargo test --workspace

The ordinary tests are modules of one tests/suite.rs target; pick a module with cargo test --test suite layout::. profile.test uses debug = "line-tables-only", which halved the test-target rebuild.

tests/generated.rs compares a warm incremental tree with a cold tree of the same state; IRIS_GENERATED_SEED, IRIS_GENERATED_SEEDS and IRIS_GENERATED_DEPTH select what it covers. tests/shrink.rs reduces a failing tree over the same fifteen cases and the same trees -- iris::random::plan(seed, depth, &edits) and build(rsc, &plan), so a failing seed reduces directly and the oracle prints the command:

SHRINK_SEED=18 SHRINK_DEPTH=6 SHRINK_CASE=repaint-some \
  cargo test --release --test shrink -- --ignored --nocapture

The cases live in tests/scenario/mod.rs, included by both targets by #[path]; a case only one rig knows is how the two drifted apart once. Turn what the shrinker finds into a test of its own rather than leaving a seed as the record. Both fuzzers take a thread per core but one. A git bisect once named a commit that could not be the cause; read the tree rather than the bisect when that happens.

tests/layout_diagnostics.rs is the retained CPU rig: IRIS_PHASE selects cold, many, repaint, size, scroll or resize, the layout-diagnostics feature gives the explanatory counters, and an uninstrumented release binary under perf gives totals. Dump the counters with

IRIS_SEED=1 IRIS_DEPTH=8 IRIS_FRAMES=500 IRIS_PHASE=many \
  <instrumented binary> --ignored --nocapture \
  | grep -E '^ +[a-z].*[0-9.]+$' | grep -v ' ms$' | sort

and diff two runs; identical output is what says a change is free.

The float head is checked out at /home/bob/repos/iris-float-cmp, at 5ed9e87 with Edits::fixed_branches applied uncommitted. Its counters do not match the grid's and will not, so a comparison against it is a bound rather than a measurement.

The headless reference set runs one process at a time because the rig reuses one compositor; comparison worktrees need separate target directories.

./scripts/run-headless.sh tabs --mode 1920x1200@60Hz --shot /tmp/tabs.png
./scripts/run-headless.sh tabs --mode 1920x1200@60Hz \
  --resize 900x1200@60Hz --shot /tmp/resized.png
./scripts/run-headless.sh tabs --mode 1920x1200@60Hz \
  --replay /tmp/tabs.touch --shot /tmp/replay.png

The replay used for the reference check:

0 down 1728 24
80 up 1728 24
400 down 1836 1116
480 up 1836 1116
800 down 1836 1116
880 up 1836 1116

Next

In order, from the review above and Bryan's steer (2026-09-17):

1. Two questions for Bryan, both from "What is not done, and why". What a widget's answer is the answer to, when a container is drawn in two of its own boxes in one frame -- which is what the two open fuzzer seeds turn on and what step 6 needs before the deferral can go. And what Pad should be now that a frame passes through: an outset (what the code does, and examples/text.rs shows what it looks like), an inset that takes the child's box with it, or the plan's "outset pixels, inset rel and leftover", which the protocol cannot express as it stands.
2. The rest of transparent frames: step 7 (Span's known-length shortcut and the cross-axis report in frame pixels), then step 8 (Inset/Outset as test widgets, once the question above is answered). wip/local-reask is superseded and can be deleted.
3. Write ActiveData::answer in one place -- try_reuse hands back what the last drawing reported, which is a measurement only where that drawing was one.
4. Keep the DrawInfo on ActiveData; delete the copied fields and the reconstruction in redraw.
5. Round on the CPU and snap to the nearest pixel in the shader, as one change with one verification. Bryan approved the snap on 2026-09-17 (rendering may change wherever it brings the screen closer to what the user's code says: three equal sections of 1000 px need one of them rounded up), and to CPU rounding the same day. The reason for that one is different: a Rel is off by at most 2^-25 of its box, so with round-to-nearest every product whose true value is a whole number of steps is exact for boxes under about 8,000 px, where truncation leaves half of them one step short and layout then decides "does not fit" on a container the user meant to fit exactly. Use the branchless round-half-up form, (a * b + (1 << (BY - 1))) >> BY, not the sign-branching shift_round; re-derive Holds::through for round (its two shifted bounds move by half a Rel step); check with nm that UiSpan::within still inlines; expect a couple of percent of instructions and re-run the long fuzzers and the render set once for both.
6. The smaller items: the stale f32 comment, the gap of an undrawn child, confirm nested leftover weights, one zero-divisor fallback. Add to them: a span that overflows itself hands a child a box of negative length, which is ordinary now rather than a corner, and nothing states what a widget may assume about one.
7. LazySpan, the next LAYOUT.md §2 item. Region nodes cover the movable subtree case; do not restore a separate child-placement API.
8. SizeRule::{Min, Max, Clamp}, restoring the max_width/max_height builders 8220a78 deleted. The clamp boundary is a hard layout decision with an exact Holds split at the crossover, both sides in Px. Still awaiting Bryan: whether a Max narrows the box the child draws in, or only what the parent reports for it.
9. Scroll taking a direction rather than one axis.

docs/LAYOUT.md §4, §5 and the density section are stale: they name Painter::place, SetSize, desired_width, apply_rest, Len::dp, Aligned and MaxSize, none of which exist. Do not restore OnResize::Translate or OrthoSize.

Other queued work, in dependency order: UiRenderState behind Rc<RefCell<_>>; density-independent pixels; input restructuring (pointer capture, drag slop and axis, cancellation, mask-aware hit testing, timestamps); retained paints, selection, overlays and shared runtime state; generic desktop/Android hosts and reusable example/APK tooling; application-owned fonts and replaceable glyph-atlas buckets; positioned text overflow and cluster-safe ellipsis.

The archive is a reference, not a patch: it predates returned Size, the current box chain and the current length types. Recreate changes on current types and keep app/session concepts out of Iris.