fix(desktop): settings row switches end-align — tag-qualified selector matched nothing by jackwener · Pull Request #663 · apache/maka

jackwener · 2026-07-09T11:09:15Z

Maintainer report: 每日回顾 switches sat mid-page while every other settings control right-aligns, plus 'why is only this one different — our governance is failing'. The diagnosis proves the point sharper than the symptom:

The end-align rule already existed (.settingsRow > button[role="switch"] { justify-self: end }, shipped in an earlier detail round) — but Base UI renders the Switch root as a , so the tag-qualified selector matched zero elements, silently, for weeks. No contract could catch it because we pinned the selector's existence, not its effect.

Fixes:

Tag-agnostic rule .settingsRow > [role="switch"] — CDP-measured: switch right-gap 461px → 20px (the row's right rail) on settings-daily-review, geometry still symmetric 3/21↔21/3
New contract: pins the rule in tag-agnostic form and bans button[role=…] tag-qualified role selectors across settings CSS — role is the contract, the rendered tag never is. This kills the whole rot class, not just this instance.

Note: the thumb-gap detail in the user's screenshot was the rem-arbitrary-value bug already fixed in #659 (their build predated it); re-measured symmetric here.

Desktop 2285/2285 (new contract +1), check-dead-css clean.

…r matched nothing 每日回顾 (and every settingsRow-based) switch sat at the value column's start while inputs/selects hug the right rail. The end-align rule existed but was written as `button[role="switch"]` — Base UI renders the Switch root as a SPAN, so the selector silently matched zero elements and no test could tell. Governance failure class: selector rot with no effect-level pin. - Rule is tag-agnostic now (.settingsRow > [role="switch"]); CDP measured rowRight 461px → 20px on settings-daily-review - Contract pins the rule in tag-agnostic form AND bans tag-qualified role selectors in settings CSS outright (role is the contract, the rendered tag never is)

…n doesn't match its own documentation (#664) #663's new contract bans the literal pattern in settings CSS; the explanatory comment contained it verbatim and failed the suite on main. Comment rephrased; 2285/2285 green.

jackwener · 2026-07-09T11:09:15Z

Maintainer report: 每日回顾 switches sat mid-page while every other settings control right-aligns, plus 'why is only this one different — our governance is failing'. The diagnosis proves the point sharper than the symptom:

The end-align rule already existed (.settingsRow > button[role="switch"] { justify-self: end }, shipped in an earlier detail round) — but Base UI renders the Switch root as a , so the tag-qualified selector matched zero elements, silently, for weeks. No contract could catch it because we pinned the selector's existence, not its effect.

Fixes:

Tag-agnostic rule .settingsRow > [role="switch"] — CDP-measured: switch right-gap 461px → 20px (the row's right rail) on settings-daily-review, geometry still symmetric 3/21↔21/3
New contract: pins the rule in tag-agnostic form and bans button[role=…] tag-qualified role selectors across settings CSS — role is the contract, the rendered tag never is. This kills the whole rot class, not just this instance.

Note: the thumb-gap detail in the user's screenshot was the rem-arbitrary-value bug already fixed in #659 (their build predated it); re-measured symmetric here.

Desktop 2285/2285 (new contract +1), check-dead-css clean.

…r matched nothing 每日回顾 (and every settingsRow-based) switch sat at the value column's start while inputs/selects hug the right rail. The end-align rule existed but was written as `button[role="switch"]` — Base UI renders the Switch root as a SPAN, so the selector silently matched zero elements and no test could tell. Governance failure class: selector rot with no effect-level pin. - Rule is tag-agnostic now (.settingsRow > [role="switch"]); CDP measured rowRight 461px → 20px on settings-daily-review - Contract pins the rule in tag-agnostic form AND bans tag-qualified role selectors in settings CSS outright (role is the contract, the rendered tag never is)

…n doesn't match its own documentation (#664) #663's new contract bans the literal pattern in settings CSS; the explanatory comment contained it verbatim and failed the suite on main. Comment rephrased; 2285/2285 green.

jackwener · 2026-07-09T11:09:15Z

Maintainer report: 每日回顾 switches sat mid-page while every other settings control right-aligns, plus 'why is only this one different — our governance is failing'. The diagnosis proves the point sharper than the symptom:

The end-align rule already existed (.settingsRow > button[role="switch"] { justify-self: end }, shipped in an earlier detail round) — but Base UI renders the Switch root as a , so the tag-qualified selector matched zero elements, silently, for weeks. No contract could catch it because we pinned the selector's existence, not its effect.

Fixes:

Tag-agnostic rule .settingsRow > [role="switch"] — CDP-measured: switch right-gap 461px → 20px (the row's right rail) on settings-daily-review, geometry still symmetric 3/21↔21/3
New contract: pins the rule in tag-agnostic form and bans button[role=…] tag-qualified role selectors across settings CSS — role is the contract, the rendered tag never is. This kills the whole rot class, not just this instance.

Note: the thumb-gap detail in the user's screenshot was the rem-arbitrary-value bug already fixed in #659 (their build predated it); re-measured symmetric here.

Desktop 2285/2285 (new contract +1), check-dead-css clean.

…r matched nothing 每日回顾 (and every settingsRow-based) switch sat at the value column's start while inputs/selects hug the right rail. The end-align rule existed but was written as `button[role="switch"]` — Base UI renders the Switch root as a SPAN, so the selector silently matched zero elements and no test could tell. Governance failure class: selector rot with no effect-level pin. - Rule is tag-agnostic now (.settingsRow > [role="switch"]); CDP measured rowRight 461px → 20px on settings-daily-review - Contract pins the rule in tag-agnostic form AND bans tag-qualified role selectors in settings CSS outright (role is the contract, the rendered tag never is)

…n doesn't match its own documentation (#664) #663's new contract bans the literal pattern in settings CSS; the explanatory comment contained it verbatim and failed the suite on main. Comment rephrased; 2285/2285 green.

jackwener · 2026-07-09T11:09:15Z

Maintainer report: 每日回顾 switches sat mid-page while every other settings control right-aligns, plus 'why is only this one different — our governance is failing'. The diagnosis proves the point sharper than the symptom:

The end-align rule already existed (.settingsRow > button[role="switch"] { justify-self: end }, shipped in an earlier detail round) — but Base UI renders the Switch root as a , so the tag-qualified selector matched zero elements, silently, for weeks. No contract could catch it because we pinned the selector's existence, not its effect.

Fixes:

Tag-agnostic rule .settingsRow > [role="switch"] — CDP-measured: switch right-gap 461px → 20px (the row's right rail) on settings-daily-review, geometry still symmetric 3/21↔21/3
New contract: pins the rule in tag-agnostic form and bans button[role=…] tag-qualified role selectors across settings CSS — role is the contract, the rendered tag never is. This kills the whole rot class, not just this instance.

Note: the thumb-gap detail in the user's screenshot was the rem-arbitrary-value bug already fixed in #659 (their build predated it); re-measured symmetric here.

Desktop 2285/2285 (new contract +1), check-dead-css clean.

…r matched nothing 每日回顾 (and every settingsRow-based) switch sat at the value column's start while inputs/selects hug the right rail. The end-align rule existed but was written as `button[role="switch"]` — Base UI renders the Switch root as a SPAN, so the selector silently matched zero elements and no test could tell. Governance failure class: selector rot with no effect-level pin. - Rule is tag-agnostic now (.settingsRow > [role="switch"]); CDP measured rowRight 461px → 20px on settings-daily-review - Contract pins the rule in tag-agnostic form AND bans tag-qualified role selectors in settings CSS outright (role is the contract, the rendered tag never is)

…n doesn't match its own documentation (#664) #663's new contract bans the literal pattern in settings CSS; the explanatory comment contained it verbatim and failed the suite on main. Comment rephrased; 2285/2285 green.

jackwener · 2026-07-09T11:09:15Z

Maintainer report: 每日回顾 switches sat mid-page while every other settings control right-aligns, plus 'why is only this one different — our governance is failing'. The diagnosis proves the point sharper than the symptom:

The end-align rule already existed (.settingsRow > button[role="switch"] { justify-self: end }, shipped in an earlier detail round) — but Base UI renders the Switch root as a , so the tag-qualified selector matched zero elements, silently, for weeks. No contract could catch it because we pinned the selector's existence, not its effect.

Fixes:

Tag-agnostic rule .settingsRow > [role="switch"] — CDP-measured: switch right-gap 461px → 20px (the row's right rail) on settings-daily-review, geometry still symmetric 3/21↔21/3
New contract: pins the rule in tag-agnostic form and bans button[role=…] tag-qualified role selectors across settings CSS — role is the contract, the rendered tag never is. This kills the whole rot class, not just this instance.

Note: the thumb-gap detail in the user's screenshot was the rem-arbitrary-value bug already fixed in #659 (their build predated it); re-measured symmetric here.

Desktop 2285/2285 (new contract +1), check-dead-css clean.

…r matched nothing 每日回顾 (and every settingsRow-based) switch sat at the value column's start while inputs/selects hug the right rail. The end-align rule existed but was written as `button[role="switch"]` — Base UI renders the Switch root as a SPAN, so the selector silently matched zero elements and no test could tell. Governance failure class: selector rot with no effect-level pin. - Rule is tag-agnostic now (.settingsRow > [role="switch"]); CDP measured rowRight 461px → 20px on settings-daily-review - Contract pins the rule in tag-agnostic form AND bans tag-qualified role selectors in settings CSS outright (role is the contract, the rendered tag never is)

…n doesn't match its own documentation (#664) #663's new contract bans the literal pattern in settings CSS; the explanatory comment contained it verbatim and failed the suite on main. Comment rephrased; 2285/2285 green.

jackwener · 2026-07-09T11:09:15Z

Maintainer report: 每日回顾 switches sat mid-page while every other settings control right-aligns, plus 'why is only this one different — our governance is failing'. The diagnosis proves the point sharper than the symptom:

The end-align rule already existed (.settingsRow > button[role="switch"] { justify-self: end }, shipped in an earlier detail round) — but Base UI renders the Switch root as a , so the tag-qualified selector matched zero elements, silently, for weeks. No contract could catch it because we pinned the selector's existence, not its effect.

Fixes:

Tag-agnostic rule .settingsRow > [role="switch"] — CDP-measured: switch right-gap 461px → 20px (the row's right rail) on settings-daily-review, geometry still symmetric 3/21↔21/3
New contract: pins the rule in tag-agnostic form and bans button[role=…] tag-qualified role selectors across settings CSS — role is the contract, the rendered tag never is. This kills the whole rot class, not just this instance.

Note: the thumb-gap detail in the user's screenshot was the rem-arbitrary-value bug already fixed in #659 (their build predated it); re-measured symmetric here.

Desktop 2285/2285 (new contract +1), check-dead-css clean.

…r matched nothing 每日回顾 (and every settingsRow-based) switch sat at the value column's start while inputs/selects hug the right rail. The end-align rule existed but was written as `button[role="switch"]` — Base UI renders the Switch root as a SPAN, so the selector silently matched zero elements and no test could tell. Governance failure class: selector rot with no effect-level pin. - Rule is tag-agnostic now (.settingsRow > [role="switch"]); CDP measured rowRight 461px → 20px on settings-daily-review - Contract pins the rule in tag-agnostic form AND bans tag-qualified role selectors in settings CSS outright (role is the contract, the rendered tag never is)

…n doesn't match its own documentation (#664) #663's new contract bans the literal pattern in settings CSS; the explanatory comment contained it verbatim and failed the suite on main. Comment rephrased; 2285/2285 green.

jackwener · 2026-07-09T11:09:15Z

Maintainer report: 每日回顾 switches sat mid-page while every other settings control right-aligns, plus 'why is only this one different — our governance is failing'. The diagnosis proves the point sharper than the symptom:

The end-align rule already existed (.settingsRow > button[role="switch"] { justify-self: end }, shipped in an earlier detail round) — but Base UI renders the Switch root as a , so the tag-qualified selector matched zero elements, silently, for weeks. No contract could catch it because we pinned the selector's existence, not its effect.

Fixes:

Tag-agnostic rule .settingsRow > [role="switch"] — CDP-measured: switch right-gap 461px → 20px (the row's right rail) on settings-daily-review, geometry still symmetric 3/21↔21/3
New contract: pins the rule in tag-agnostic form and bans button[role=…] tag-qualified role selectors across settings CSS — role is the contract, the rendered tag never is. This kills the whole rot class, not just this instance.

Note: the thumb-gap detail in the user's screenshot was the rem-arbitrary-value bug already fixed in #659 (their build predated it); re-measured symmetric here.

Desktop 2285/2285 (new contract +1), check-dead-css clean.

…r matched nothing 每日回顾 (and every settingsRow-based) switch sat at the value column's start while inputs/selects hug the right rail. The end-align rule existed but was written as `button[role="switch"]` — Base UI renders the Switch root as a SPAN, so the selector silently matched zero elements and no test could tell. Governance failure class: selector rot with no effect-level pin. - Rule is tag-agnostic now (.settingsRow > [role="switch"]); CDP measured rowRight 461px → 20px on settings-daily-review - Contract pins the rule in tag-agnostic form AND bans tag-qualified role selectors in settings CSS outright (role is the contract, the rendered tag never is)

…n doesn't match its own documentation (#664) #663's new contract bans the literal pattern in settings CSS; the explanatory comment contained it verbatim and failed the suite on main. Comment rephrased; 2285/2285 green.

jackwener · 2026-07-09T11:09:15Z

Maintainer report: 每日回顾 switches sat mid-page while every other settings control right-aligns, plus 'why is only this one different — our governance is failing'. The diagnosis proves the point sharper than the symptom:

The end-align rule already existed (.settingsRow > button[role="switch"] { justify-self: end }, shipped in an earlier detail round) — but Base UI renders the Switch root as a , so the tag-qualified selector matched zero elements, silently, for weeks. No contract could catch it because we pinned the selector's existence, not its effect.

Fixes:

Tag-agnostic rule .settingsRow > [role="switch"] — CDP-measured: switch right-gap 461px → 20px (the row's right rail) on settings-daily-review, geometry still symmetric 3/21↔21/3
New contract: pins the rule in tag-agnostic form and bans button[role=…] tag-qualified role selectors across settings CSS — role is the contract, the rendered tag never is. This kills the whole rot class, not just this instance.

Note: the thumb-gap detail in the user's screenshot was the rem-arbitrary-value bug already fixed in #659 (their build predated it); re-measured symmetric here.

Desktop 2285/2285 (new contract +1), check-dead-css clean.

…r matched nothing 每日回顾 (and every settingsRow-based) switch sat at the value column's start while inputs/selects hug the right rail. The end-align rule existed but was written as `button[role="switch"]` — Base UI renders the Switch root as a SPAN, so the selector silently matched zero elements and no test could tell. Governance failure class: selector rot with no effect-level pin. - Rule is tag-agnostic now (.settingsRow > [role="switch"]); CDP measured rowRight 461px → 20px on settings-daily-review - Contract pins the rule in tag-agnostic form AND bans tag-qualified role selectors in settings CSS outright (role is the contract, the rendered tag never is)

…n doesn't match its own documentation (#664) #663's new contract bans the literal pattern in settings CSS; the explanatory comment contained it verbatim and failed the suite on main. Comment rephrased; 2285/2285 green.