Problem. At real data volume the v2 site view is mostly empty space. Verified live against Tower on :4100 after #104 merged — 22 workspaces, 60 builders — the page renders 2846px tall for content that would fit in a fraction of it.
Root cause. .plot-grid (apps/v2/src/site.css:59) is a CSS grid with default align-items: stretch, so every plot in a row is stretched to the height of the tallest one.
In the live render, row 2 contains CODEV-1455 (13 builders) and ENTRIQ (27 builders). The other two cells in that row — CODEV1 and DVARR, with 0 and 3 builders — are stretched to roughly 1100px tall each, almost entirely blank. The same happens in every row with one tall plot.
Why this matters. 4 builders are gate-waiting out of 60. Those 4 are the only rows on the page that need a human, and they are scattered down 2846px of mostly-empty plots. The footer says 22 WORKSPACES · 60 BUILDERS · 4 GATE-WAITING while the thing you have to scroll for is the 4.
This is the requirement the program was given in the user's own words: "We have to have solid usability and use of screen realestate."
Likely fix. align-items: start on .plot-grid so each plot is only as tall as its content. One line. Worth checking whether a masonry-style column flow packs better than a row grid once plots are free-height, since plot heights vary by two orders of magnitude here.
Not a spec 83 defect. Spec 83's scope was "render the hierarchy live, nothing else," and it does that correctly — containment holds, rust appears only on gate stamps, counts read as machine totals, heldMail marks render. The density problem is only visible at realistic volume, which is exactly why it took a live pass to find: the Playwright fixture has one workspace and could never have shown it.
Related. ENTRIQ shows 27 offline builders and CODEV-1455 shows 13, most of them orphaned worktrees per #100. Fixing #100 would remove a large share of the rows driving this, but it would not fix the stretching — one tall plot is enough.
Testing. Render a fixture with plots of wildly different heights (0, 3, 13, 27 builders) at 1440 and assert no plot's rendered height exceeds its content height. A single-workspace fixture cannot catch this.
Problem. At real data volume the v2 site view is mostly empty space. Verified live against Tower on :4100 after #104 merged — 22 workspaces, 60 builders — the page renders 2846px tall for content that would fit in a fraction of it.
Root cause.
.plot-grid(apps/v2/src/site.css:59) is a CSS grid with defaultalign-items: stretch, so every plot in a row is stretched to the height of the tallest one.In the live render, row 2 contains
CODEV-1455(13 builders) andENTRIQ(27 builders). The other two cells in that row —CODEV1andDVARR, with 0 and 3 builders — are stretched to roughly 1100px tall each, almost entirely blank. The same happens in every row with one tall plot.Why this matters. 4 builders are gate-waiting out of 60. Those 4 are the only rows on the page that need a human, and they are scattered down 2846px of mostly-empty plots. The footer says
22 WORKSPACES · 60 BUILDERS · 4 GATE-WAITINGwhile the thing you have to scroll for is the 4.This is the requirement the program was given in the user's own words: "We have to have solid usability and use of screen realestate."
Likely fix.
align-items: starton.plot-gridso each plot is only as tall as its content. One line. Worth checking whether a masonry-style column flow packs better than a row grid once plots are free-height, since plot heights vary by two orders of magnitude here.Not a spec 83 defect. Spec 83's scope was "render the hierarchy live, nothing else," and it does that correctly — containment holds, rust appears only on gate stamps, counts read as machine totals,
heldMailmarks render. The density problem is only visible at realistic volume, which is exactly why it took a live pass to find: the Playwright fixture has one workspace and could never have shown it.Related.
ENTRIQshows 27 offline builders andCODEV-1455shows 13, most of them orphaned worktrees per #100. Fixing #100 would remove a large share of the rows driving this, but it would not fix the stretching — one tall plot is enough.Testing. Render a fixture with plots of wildly different heights (0, 3, 13, 27 builders) at 1440 and assert no plot's rendered height exceeds its content height. A single-workspace fixture cannot catch this.