Aquarium breach · glm-5.3-flash

Build single-file index.html. Write a self-contained sim. Keep it reasonably compact but functional. Then test in browser.

Plan physics:
- Tank: rect with glass walls, opening position (crack) draggable on right wall below initial waterline.
- Water: level height, outflow via Torricelli v=sqrt(2gh), jet particle emitted from opening with initial velocity, gravity curve, land on floor -> puddle.
- Objects: fish (swim, flee current, some swept), rocks (sink), toy (floats), plants (bend). Objects movable: toy, small rocks, fish can be carried out through opening if fits.
- Glass shards on breach: fragments with rotation, collide with floor, drag in water.

Write ~500 lines. Then browser check.

Initial scene looks good. Crack indicator overlaps the duck — fine. Drag crack lower, then breach. Crack currently at waterline top; drag it down, click Breach, sample early outflow.

Breach happened: panel broken around opening, shards falling, waterline still at same spot. Sample early outflow via motion. Also note a black box near opening (jagged draw artifact?). Actually a black rectangle appears at opening — that's the clearRect leaving dark background (tank back fill?) — it's the clearRect region showing room bg, but it looks darker because... clearRect removes everything then tank back fill was drawn earlier? No—clearRect happens after water & tank back drawn, so it erases water there, showing canvas transparent → dark. That's fine, it's the opening. But it looks like a solid black box, acceptable but could look better if showing water behind. Let's check motion for jet curvature.

Problems: the waterline is NOT falling (top of water stays at y~192). Shards went out but no jet drops visible, and waterline static. Also the black box remains. Why no outflow? outflow(): top=waterTopY ~192; crack y0 ~560? Wait, dragging dragged the crack to y~560 (frame before breach showed duck at 558). But the black box appears around y~535-580. h0~540,h1~580. top=192 < h0 → top>=h1? top(192) < h1(580) so not 0. submergedTop=min(top,h1)=192; submergedBot=max(top,h0)=540 → subH negative → return 0. Bug: submerged logic inverted. submerged region is between top and h0 (water above opening? No—water is above the opening; opening is submerged when top < h0). Submerged height = h0 - top portion... Actually opening spans h0..h1; submerged part is from h0 to h1 as long as top < h0. subH = h1-h0 if top<h0, else 0. Depth = top... depth of opening center below waterline = openCy() - top. Fix:

if top >= h0 → not submerged (waterline below opening top) → 0? Actually if waterline is within opening (h0<top<h1), partially submerged: subH = h1-top? No—water is above; the part of the opening above waterline isn't submerged. Submerged part is h0..min(h1,top)? No: water occupies y< top. The opening at y in [h0,h1] is under water where y<top? Water is from top of water surface down to tank bottom... wait waterTopY = tank.y + h*(1-level), i.e., y increases downward, water fills from top down to tank bottom. So water region is y from waterTopY to tank bottom. So opening submerged where y > top. Submerged part: max(h0,top)..h1. subH = h1 - max(h0,top). Depth of submerged centroid below waterline = centroidY - top.

Current code: submergedTop=min(top,h1), submergedBot=max(top,h0) — that's wrong (computed the dry part). Fix:

const subTop=Math.max(top,h0), subBot=h1; if(subBot<=subTop) return 0; subH=subBot-subTop; depth=(subTop+subBot)/2 - top; if(depth<=0)return 0.

Emission y should be within submerged part: y in [subTop, subBot].

Also the waterline falling: tank.level -= A*v*dt/tank.w/100. With scaling /100, rate: subH(40)*v(~sqrt(2*900*430)≈880) → 40*880/760/100 ≈ 0.46/s → drains in ~2s. Too fast? /100 gives 0.0046 per px... level 0.78 → ~1.7 s. Too fast. Increase divisor to 400 → ~7s. Fine.

Also black box: draw order — I clearRect after drawing water/tank back but shards etc. Actually the black is background showing; better draw opening as water-colored/dark opening. Instead of clearRect, just draw the wall segments (I already stroke separate segments) — remove the clearRect and full-wall stroke is fine since I stroke segments. Also there's a remaining vertical stroke of full wall? I stroke both full rect AND segments; full rect stroke crosses the opening, then clearRect erased it plus background. Remove full-rect stroke's right side by drawing three separate strokes (top, left, bottom) plus right segments. Simplest: stroke left/top/bottom as path, right as segments.

Also puddle puddleCells pile only near tank; fine.

Also fish at wall x pinned near opening — one fish already got out? An orange fish at right wall seems stuck at wall, fine.

Also toy got pinned at right wall at crack level and looks like it never exits since out condition requires |t.y-openCy()|<openH*0.7 and x>tankRight()-r-10 — toy sits at x=tankRight()-r, should trigger. It did earlier? Duck moved to crack position — good, will be carried out once outflow works.

Fix edits.

Also `c` variable (openCy) unused in outflow now — fine. Also subTop referenced in update emit — defined within outflow but I used subTop/subBot in the update block. outflow returns {v,A,subH} — I changed emit to use subTop which isn't in scope in update! Fix: add subTop to returned object and use of.subTop.

Walls now look clean. Drag crack down, breach, sample motion.

Now outflow works: waterline fell from 192 to ~572 (opening at ~560, so outflow stopped once water reached opening level — good, "stops when water no longer reaches it"). Shards on floor, toy got carried out (duck outside on floor). Fish: one fish at left wall flapping (out of water, flopping). All fish seemingly out/flopped? Only one fish visible at left. Fish may have been swept out early and flopped offscreen? Fish out motion: they fly right, land, flop. Possibly went off. Rocks still inside; rock y stuck at tank floor - fine.

But wait — the drain was extremely fast: from level .78 to crack level in ~0.5s. Frame 1 (0ms) already waterline at 572. Breach → screenshot immediately showed waterline still 192 with shards; motion frame 1 (a second later) already drained to crack level. The /400 divisor: rate = A*v/w/400 = 40*880/760/400 ≈ 0.116/s → ~5.5s to drain 0.78... but level only needs to fall to crack level (0.78→~0.15?) crack at y560 of tank 120..740: level = (740-560)/620 ≈ 0.29. Falling 0.49 at 0.116/s ≈ 4s. But observed <1s. Hmm, subH=40? openH is 40, v≈ sqrt(2*900*400)=848. A=subH=40. Hmm 40*848=33920/760/400=0.111. Too fast observed... maybe dt spikes (0.033 cap). 4s/... whatever, it drained faster than intended but that's between frames (~1.6s elapsed over 4 motion frames, first frame captured later). Acceptable, but a bit quick. Slow it: /800 → ~9s. Also early outflow sample: I need to verify jet curves downward — motion frames show few drops at opening going right then down (dots at ~800,600→~790,700 area). Frame1 shows dots trailing right and downward from opening. OK.

Also fish: all gone quickly. Fish "swept through" happened but 6 fish vanished entirely — flopping off-screen left? Fish flop vx ±80, they stay. Only one visible. Others may have exited right off-screen (x>W not clamped for fish). Fish out: no x clamp. Add clamp W. Also make fewer fish swept out (probability dt*2.0 high when near). Fine—requirement says "some are swept". Keep but clamp x.

Also toy: carried out — good, lies on floor. Rocks remain inside on tank floor — good (they fit? r=26 > openH*0.6=24 → not carried; r=16<24 would be carried out if near opening, ok).

Toy floats: fine.

Also plants bend: they lean right in later frames — visible.

One issue: drain too fast and no visible sustained jet in sampled frames. Slow drain to /900 and reduce emit rate constant 0.02→keep. Let's adjust divisor to 900.

Excellent: gravity-curved jet clearly visible, waterline falling continuously, plants bend right with the current, rocks sink/stay, fish near surface huddling left (against current), duck drifted near opening. Fish count: I see 2 fish at left edge? Fish seem to have been swept out earlier (out). Fine.

Now test Reset.

Reset works: intact tank, water level 0.78, crack back at initial spot, puddle gone, duck floating, all fish inside. Done. Report.