A diff that holds still, the run bar's state as the way to Run details, and a compaction count on its line (#552 #553 #554) #555

Merged
cmoriarty merged 5 commits from claude/console-fixes-trio into main 2026-11-18 12:04:03 -05:00
Owner

Three console fixes, each diagnosed on its issue, plus two asks from the owner's local try. OpenSpec change console-diff-run-bar-count.

braid#552: a diff holds still

An edit's diff chose two columns or one from its box's clientWidth, which loses the box's own scrollbar. A box just over 1000 px split, scrolled, read 996, went back to one column, fitted, and split again on every frame. The rows below it moved each time, so the pane flickered (run 200's openspec.propose, with a mouse connected). It reads offsetWidth now.

A spec edit at 1004 px, split and scrolling, its scrollbar taking it to 994 px

braid#553: the run bar's state is the way to Run details

  • The decode speed is gone. Run details and Inference keep the backend's numbers.
  • The state ends the row, after the time, as Running › in its colour. It is the bar's one link to Run details, and the separate Run details › is gone.
  • Needs you › opens Run details too. The inbox keeps its own place, with its count, on the top bar and on a phone's bottom bar.
  • From the local try: a second click on the state closes Run details, and its back link reads ‹ Back. Both return to the step that was open. The bar opens Run details at an address with no step, so Back used to go to the top of the run.

The run bar: the state ends the row

A run that needs you

Run details open: the state marked current, and Back

A phone at 390 px

braid#554: the compaction count sits on its line

A settled step's cell lines up its items by baseline, and the mark's SVG has none, so it stood about 5 px above the time, in the list and on the graph. It is centred on the line now, and a row with a count is no taller than one without.

The list: the count beside 1h 44m

The graph: the count beside 1h 44m

Tests

Each check below failed on main first.

Issue Check On main Here
#552 New journey ui/e2e/scrollbars.spec.ts, with Chromium drawing its scrollbars (Playwright hides them by default, which is why no journey saw this) 9 column changes in 10 frames 0
#553 the bar is one row, its state last, and the state opens Run details on the active step. Its assertion that the state leads the row changed on purpose. fails: the state is not the link passes, a second click included
#554 The compaction journey, extended to settled steps in the list and on the graph 6 px apart at most 1 px

./tools/test.sh passed in 54 s, with all 61 journeys. Screenshots are from a seeded test server.

🤖 Generated with Claude Code

Three console fixes, each diagnosed on its issue, plus two asks from the owner's local try. OpenSpec change `console-diff-run-bar-count`. ## braid#552: a diff holds still An edit's diff chose two columns or one from its box's `clientWidth`, which loses the box's own scrollbar. A box just over 1000 px split, scrolled, read 996, went back to one column, fitted, and split again on every frame. The rows below it moved each time, so the pane flickered (run 200's `openspec.propose`, with a mouse connected). It reads `offsetWidth` now. ![A spec edit at 1004 px, split and scrolling, its scrollbar taking it to 994 px](/attachments/896696ac-a2f8-4196-ab35-e9ad122b704f) ## braid#553: the run bar's state is the way to Run details - The decode speed is gone. Run details and Inference keep the backend's numbers. - The state ends the row, after the time, as `Running ›` in its colour. It is the bar's one link to Run details, and the separate `Run details ›` is gone. - `Needs you ›` opens Run details too. The inbox keeps its own place, with its count, on the top bar and on a phone's bottom bar. - From the local try: a second click on the state closes Run details, and its back link reads `‹ Back`. Both return to the step that was open. The bar opens Run details at an address with no step, so Back used to go to the top of the run. ![The run bar: the state ends the row](/attachments/e6f1e442-896a-4dd3-8fdb-300e51f9dbb1) ![A run that needs you](/attachments/8fe8a1cd-3143-459f-8ee2-1f71a4e9b0da) ![Run details open: the state marked current, and Back](/attachments/76aaaeea-b9b5-47f3-892c-fa9c32f4f95c) ![A phone at 390 px](/attachments/6b3366d9-325b-4ea7-89b1-5646b505fd49) ## braid#554: the compaction count sits on its line A settled step's cell lines up its items by baseline, and the mark's SVG has none, so it stood about 5 px above the time, in the list and on the graph. It is centred on the line now, and a row with a count is no taller than one without. ![The list: the count beside 1h 44m](/attachments/fd5f9e1a-aad3-457f-a02a-6e36ea9d365a) ![The graph: the count beside 1h 44m](/attachments/63ebb5eb-3304-44a3-8dd2-37b0511d2b5b) ## Tests Each check below failed on main first. | Issue | Check | On main | Here | |---|---|---|---| | #552 | New journey `ui/e2e/scrollbars.spec.ts`, with Chromium drawing its scrollbars (Playwright hides them by default, which is why no journey saw this) | 9 column changes in 10 frames | 0 | | #553 | `the bar is one row, its state last, and the state opens Run details on the active step`. Its assertion that the state leads the row changed on purpose. | fails: the state is not the link | passes, a second click included | | #554 | The compaction journey, extended to settled steps in the list and on the graph | 6 px apart | at most 1 px | `./tools/test.sh` passed in 54 s, with all 61 journeys. Screenshots are from a seeded test server. 🤖 Generated with [Claude Code](https://claude.com/claude-code)
A diff chooses its columns from a width its own scrollbar cannot change; the
run bar's state ends the row, drops the speed and is the Run details link; the
compaction count is centred on the line of the words beside it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
DiffBlock read clientWidth, which loses the box's own scrollbar, so a box just over 1000 px split,
scrolled, read under 1000, went back to one column, fitted and split again on every frame.
It reads offsetWidth now. The journey runs Chromium with its scrollbars drawn, which no other
journey does, and failed on the old code with 9 column changes in 10 frames.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The decode speed goes: it often read 0 while the model worked, and Run details and Inference
keep the backend's numbers. The state moves to the right end, after the time, and is the bar's one
link to Run details, Needs you included; the separate Run details link goes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A settled step's cell lines its items up by baseline, and the mark's SVG has none, so it stood on
the text's baseline and rose about 5 px above the time, in the list and on the graph.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
feat(console): a second click on the state closes Run details, and its back link reads Back (#553)
All checks were successful
test / suite (pull_request) Successful in 1m23s
77fd6c5c22
From the owner's local try. While Run details is open, the state leads where Back does. Both now
return to the step that was open before Run details: the bar opens it at an address with no step,
so Back used to go to the top of the run.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sign in to join this conversation.
No reviewers
No labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
cmoriarty/braid!555
No description provided.