Repository navigation
experimental theme: states Core marks, open parts drawn open, and picture inputs and flyouts - #12498
josephfarina wants to merge 23 commits into
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
|
AI review status for this pull request.
|
90c324b to
b5946a2
Compare
f6eb88c to
8079a75
Compare
There was a problem hiding this comment.
Changes requested. The code is the same as at the previous head, so four fixes are still needed; the test plan's CI sentence is now accurate.
Blocking:
experimental-theme-editor-show.doc.mjsdropsrequires: ['themes.compile'], so in a project whose base is not Core the editor counts as served andastryx themehelp and suggestions point ateditor new, which that base refuses;clients/cli/bases.test.mjsfails 5 tests here and passes on the base, so restore the line._how-question.mjs:25279copies a text color to each variant a fill reaches without that fill'smode, so "buttons with a navy background in dark mode and white text" also writes white text on the secondary and ghost buttons in light mode; carry each fill's mode.- The tint pass (
_how-question.mjs:25355-25374) turns every solid fill in a "tinted" clause into a 15% mix, so "cards tinted blue with a navy header" tints the header while the answer's own lines for the colored card variants stay solid; tint only the color the word modifies, on every line that writes it. .changeset/theme-round15-part-02.md:18and the description list "a light-only theme reads--mode lightas no--mode" as new, but the base branch already does this and both new tests for it pass against the base code; drop the sentence.
Approval also waits on the direction: no current record settles this round's new how readings, set behavior, reports and pictures, so an engineering owner records the decision by commenting /approve-spec <full head SHA> on the pull request, and the author may do so as an engineering owner. Three new tests also pass against the base code (the faint-link fix, the one-note-per-cause fill and the placeholder key); a case the base gets wrong would let them prove their change.
Claims and files checked
Review mode: carry-forward; reason: none
Claims (release notes and description):
- A border asked of a part that draws none goes to its inner pieces only where they draw borders. Verified:
theme-bar-border.test.mjsfails on the base code and passes here. - A role text color too faint is fixed in its own hue. Unverified: its test passes on the base code too.
- A light-only border reports nothing about dark. Verified: its test fails on the base code.
- One fill on a group is noted once per kind of cover. Unverified: its test passes on the base code too.
- A color with a pill shape keeps the selected item's state. Verified:
theme-how-plain-words.test.mjsfails on the base code. - A text color for every variant goes only where and when it reads. Partly false: verified for white button text, but a fill given for one mode drops its mode (blocking).
- A tinted color fills at low strength. Partly false: verified for a selected nav item, but it tints fills the word does not modify and leaves follow-up lines solid (blocking).
- A rule with its own verb after a part clause is theme-wide. Verified:
theme-how-list-items.test.mjsfails on the base code. --onlyon a part styled on its own is taken and said; a group still refuses it. Verified for the first; the group refusal already held on the base.- A menu radio keeps its circle, and a round ask offers the undo. Verified: both tests fail on the base code; the offered undo runs.
- A light-only theme reads its own
--modeas none. False: the base already does this, and both tests pass on the base code (blocking). - A part Core marks open is drawn open, so
--state expandedreaches a submenu row and a top nav menu. Verified:theme-open-state-drawn.test.mjsand the newlayers.test.mjscase fail on the base code. - States Core marks on a part are settable: a list item's selected, a checkbox or radio box's hover from its holder, a container's state named on its inner part. Verified:
theme-part-states-marked.test.mjsandtheme-under-input.test.mjsfail on the base code. showdraws sample inputs as the theme's text input, a flyout once beside its menu, a press at its moment, a highlighted palette and the drawer in its cell. Verified:theme-pictures-open-layers-samples.test.mjsfails on the base code; the pictures show the flyout and drawer.- Pictures show their whole page, no MobileNav cell on any page, an overlay under its opener and sub-pixel cut text whole. Verified:
theme-pictures-whole-page.test.mjsfails on the base code; the base navigation sheet has the empty MobileNav picture. - Rows said of a part are its own rows, and a part word before an unknown noun is a compound subject. Verified:
theme-rows-and-product-words.test.mjsfails on the base code. - A state's fill is measured against its own variant's text, and a part without text gets no text problem. Verified: both tests fail on the base code.
- A placeholder as faint as the base drew it is no new problem. Unverified: its test passes on the base code too.
- A one-mode theme's text color on an image fill is written for both modes. Verified: two of its three tests fail on the base code.
- A tall family picture is cut between rows into pictures of about 1,600px. Verified against
cutTalland its caller in_context.mjs. - The overlays picture shows about five to a row and a panel cropped to its top. Verified against the overlays sheet styles in
_catalog.mjs. - Editor-only, with no Core, record or official-theme change. Partly false: it also drops
editor show's base-step declaration (blocking). _how-question.mjschanges by +328/-56. Verified.- It adds mood or feel phrase matches. Unverified: no added hunk reads a mood or feel ask (non-blocking).
- Each commit was gated in an integration branch with the full suite, typechecks and an answer key. Unverified: that branch cannot be read here.
- CI has not run on this head, since a stacked part gets CI only once it targets
main. Verified: CI runs only for pull requests intomainand release branches, and none ran. - Any failure found only at this head is fixed here. False while
bases.test.mjsfails; restoring the base step makes it true. - A look gate on a sample site found no new blocking finding. Unverified.
Files:
packages/cli/clients/cli/commands/experimental-theme-editor-show.doc.mjs: drops thethemes.compilebase step (blocking).packages/cli/api/theme-edit/_how-question.mjs: a paired text color loses its fill's mode; the tint pass rewrites the wrong fills (both blocking)..changeset/theme-round15-part-02.md: the one-mode--modesentence describes base behavior (blocking).packages/cli/api/theme-edit/_check.mjs: the under-input refusal's holder list is now always empty (non-blocking)..changeset/experimental-theme-pictures-agents-open-overlays.md,.changeset/experimental-theme-sheet-cells-b.md: clear.packages/cli/api/theme-edit/_catalog.mjs,_context.mjs,_edit.mjs,_parts.mjs,_preview.mjs,_settings.mjs: clear.packages/cli/foundation/render/layers.mjs,page-worker.mjs,picture-chrome.mjs,picture-playwright.mjs,picture-window.mjs: clear.theme-fix-lines-keep-role.test.mjs,theme-state-fill-text-scope.test.mjs: pass here; three of their cases also pass on the base code (non-blocking).theme-bar-border.test.mjs,theme-edit-only-escape-and-light-only-mode.test.mjs,theme-edit.test.mjs,theme-how-clause-scope.test.mjs,theme-how-draws-modes-clearly-light.test.mjs,theme-how-list-items.test.mjs,theme-how-plain-words.test.mjs,theme-how-variant-looks.test.mjs,theme-open-state-drawn.test.mjs,theme-options-written-loosely-words.test.mjs,theme-part-states-marked.test.mjs,theme-pictures-fit-cut-text.test.mjs,theme-pictures-open-layers-samples.test.mjs,theme-pictures-whole-page.test.mjs,theme-refusal-runs-label-swapped.test.mjs,theme-rows-and-product-words.test.mjs,theme-under-input.test.mjs,theme-whole-fill-text.test.mjs(all underpackages/cli/api/theme-edit/),packages/cli/foundation/render/layers.test.mjs: clear; they pass here.
8079a75 to
2035ba1
Compare
There was a problem hiding this comment.
Changes requested. The branch was rebased onto the updated base with no change to its own patch, so the same four fixes are still needed.
Blocking:
experimental-theme-editor-show.doc.mjsdropsrequires: ['themes.compile'], so in a project whose base is not Core the editor counts as served andastryx themehelp and suggestions point ateditor new, which that base refuses;clients/cli/bases.test.mjsfails 5 tests here and passes on the base, so restore the line._how-question.mjs:25279copies a text color to each variant a fill reaches without that fill'smode, so "buttons with a navy background in dark mode and white text" also writes white text on the secondary and ghost buttons in light mode; carry each fill's mode.- The tint pass (
_how-question.mjs:25355-25374) turns every solid fill in a "tinted" clause into a 15% mix, so "cards tinted blue with a navy header" tints the header while the answer's own lines for the colored card variants stay solid; tint only the color the word modifies, on every line that writes it. .changeset/theme-round15-part-02.md:18and the description list "a light-only theme reads--mode lightas no--mode" as new, but the base branch already does this and both new tests for it pass against the base code; drop the sentence.
Approval also waits on the direction: no current record settles this round's new how readings, set behavior, reports and pictures, so an engineering owner records the decision by commenting /approve-spec <full head SHA> on the pull request, and the author may do so as an engineering owner. Three new tests also pass against the base code (the faint-link fix, the one-note-per-cause fill and the placeholder key); a case the base gets wrong would let them prove their change.
Claims and files checked
Review mode: carry-forward; reason: none
Claims (release notes and description):
- A border asked of a part that draws none goes to its inner pieces only where they draw borders. Verified:
theme-bar-border.test.mjsfails on the base code and passes here. - A role text color too faint is fixed in its own hue. Unverified: its test passes on the base code too.
- A light-only border reports nothing about dark. Verified: its test fails on the base code.
- One fill on a group is noted once per kind of cover. Unverified: its test passes on the base code too.
- A color with a pill shape keeps the selected item's state. Verified:
theme-how-plain-words.test.mjsfails on the base code. - A text color for every variant goes only where and when it reads. Partly false: verified for white button text, but a fill given for one mode drops its mode (blocking).
- A tinted color fills at low strength. Partly false: verified for a selected nav item, but it tints fills the word does not modify and leaves follow-up lines solid (blocking).
- A rule with its own verb after a part clause is theme-wide. Verified:
theme-how-list-items.test.mjsfails on the base code. --onlyon a part styled on its own is taken and said; a group still refuses it. Verified for the first; the group refusal already held on the base.- A menu radio keeps its circle, and a round ask offers the undo. Verified: both tests fail on the base code; the offered undo runs.
- A light-only theme reads its own
--modeas none. False: the base already does this, and both tests pass on the base code (blocking). - A part Core marks open is drawn open, so
--state expandedreaches a submenu row and a top nav menu. Verified:theme-open-state-drawn.test.mjsand the newlayers.test.mjscase fail on the base code. - States Core marks on a part are settable: a list item's selected, a checkbox or radio box's hover from its holder, a container's state named on its inner part. Verified:
theme-part-states-marked.test.mjsandtheme-under-input.test.mjsfail on the base code. showdraws sample inputs as the theme's text input, a flyout once beside its menu, a press at its moment, a highlighted palette and the drawer in its cell. Verified:theme-pictures-open-layers-samples.test.mjsfails on the base code; the pictures show the flyout and drawer.- Pictures show their whole page, no MobileNav cell on any page, an overlay under its opener and sub-pixel cut text whole. Verified:
theme-pictures-whole-page.test.mjsfails on the base code; the base navigation sheet has the empty MobileNav picture. - Rows said of a part are its own rows, and a part word before an unknown noun is a compound subject. Verified:
theme-rows-and-product-words.test.mjsfails on the base code. - A state's fill is measured against its own variant's text, and a part without text gets no text problem. Verified: both tests fail on the base code.
- A placeholder as faint as the base drew it is no new problem. Unverified: its test passes on the base code too.
- A one-mode theme's text color on an image fill is written for both modes. Verified: two of its three tests fail on the base code.
- A tall family picture is cut between rows into pictures of about 1,600px. Verified against
cutTalland its caller in_context.mjs. - The overlays picture shows about five to a row and a panel cropped to its top. Verified against the overlays sheet styles in
_catalog.mjs. - Editor-only, with no Core, record or official-theme change. Partly false: it also drops
editor show's base-step declaration (blocking). _how-question.mjschanges by +328/-56. Verified.- It adds mood or feel phrase matches. Unverified: no added hunk reads a mood or feel ask (non-blocking).
- Each commit was gated in an integration branch with the full suite, typechecks and an answer key. Unverified: that branch cannot be read here.
- CI has not run on this head, since a stacked part gets CI only once it targets
main. Verified: CI runs only for pull requests intomainand release branches, and none ran. - Any failure found only at this head is fixed here. False while
bases.test.mjsfails; restoring the base step makes it true. - A look gate on a sample site found no new blocking finding. Unverified.
Files:
packages/cli/clients/cli/commands/experimental-theme-editor-show.doc.mjs: drops thethemes.compilebase step (blocking).packages/cli/api/theme-edit/_how-question.mjs: a paired text color loses its fill's mode; the tint pass rewrites the wrong fills (both blocking)..changeset/theme-round15-part-02.md: the one-mode--modesentence describes base behavior (blocking).packages/cli/api/theme-edit/_check.mjs: the under-input refusal's holder list is now always empty (non-blocking)..changeset/experimental-theme-pictures-agents-open-overlays.md,.changeset/experimental-theme-sheet-cells-b.md: clear.packages/cli/api/theme-edit/_catalog.mjs,_context.mjs,_edit.mjs,_parts.mjs,_preview.mjs,_settings.mjs: clear.packages/cli/foundation/render/layers.mjs,page-worker.mjs,picture-chrome.mjs,picture-playwright.mjs,picture-window.mjs: clear.theme-fix-lines-keep-role.test.mjs,theme-state-fill-text-scope.test.mjs: pass here; three of their cases also pass on the base code (non-blocking).theme-bar-border.test.mjs,theme-edit-only-escape-and-light-only-mode.test.mjs,theme-edit.test.mjs,theme-how-clause-scope.test.mjs,theme-how-draws-modes-clearly-light.test.mjs,theme-how-list-items.test.mjs,theme-how-plain-words.test.mjs,theme-how-variant-looks.test.mjs,theme-open-state-drawn.test.mjs,theme-options-written-loosely-words.test.mjs,theme-part-states-marked.test.mjs,theme-pictures-fit-cut-text.test.mjs,theme-pictures-open-layers-samples.test.mjs,theme-pictures-whole-page.test.mjs,theme-refusal-runs-label-swapped.test.mjs,theme-rows-and-product-words.test.mjs,theme-under-input.test.mjs,theme-whole-fill-text.test.mjs(all underpackages/cli/api/theme-edit/),packages/cli/foundation/render/layers.test.mjs: clear; they pass here.
2035ba1 to
988174c
Compare
9a015e4 to
4a11932
Compare
There was a problem hiding this comment.
Changes requested. The branch was rebased onto the updated base with no change to its own patch, so the same four fixes are still needed.
Blocking:
experimental-theme-editor-show.doc.mjsdropsrequires: ['themes.compile'], so in a project whose base is not Core the editor counts as served andastryx themehelp and suggestions point ateditor new, which that base refuses;clients/cli/bases.test.mjsfails 5 tests here and passes on the base, so restore the line._how-question.mjs:25279copies a text color to each variant a fill reaches without that fill'smode, so "buttons with a navy background in dark mode and white text" also writes white text on the secondary and ghost buttons in light mode; carry each fill's mode.- The tint pass (
_how-question.mjs:25355-25374) turns every solid fill in a "tinted" clause into a 15% mix, so "cards tinted blue with a navy header" tints the header while the answer's own lines for the colored card variants stay solid; tint only the color the word modifies, on every line that writes it. .changeset/theme-round15-part-02.md:18and the description list "a light-only theme reads--mode lightas no--mode" as new, but the base branch already does this and both new tests for it pass against the base code; drop the sentence.
Approval also waits on the direction: no current record settles this round's new how readings, set behavior, reports and pictures, so an engineering owner records the decision by commenting /approve-spec <full head SHA> on the pull request, and the author may do so as an engineering owner. Three new tests also pass against the base code (the faint-link fix, the one-note-per-cause fill and the placeholder key); a case the base gets wrong would let them prove their change.
Claims and files checked
Review mode: carry-forward; reason: none
Claims (release notes and description):
- A border asked of a part that draws none goes to its inner pieces only where they draw borders. Verified:
theme-bar-border.test.mjsfails on the base code and passes here. - A role text color too faint is fixed in its own hue. Unverified: its test passes on the base code too.
- A light-only border reports nothing about dark. Verified: its test fails on the base code.
- One fill on a group is noted once per kind of cover. Unverified: its test passes on the base code too.
- A color with a pill shape keeps the selected item's state. Verified:
theme-how-plain-words.test.mjsfails on the base code. - A text color for every variant goes only where and when it reads. Partly false: verified for white button text, but a fill given for one mode drops its mode (blocking).
- A tinted color fills at low strength. Partly false: verified for a selected nav item, but it tints fills the word does not modify and leaves follow-up lines solid (blocking).
- A rule with its own verb after a part clause is theme-wide. Verified:
theme-how-list-items.test.mjsfails on the base code. --onlyon a part styled on its own is taken and said; a group still refuses it. Verified for the first; the group refusal already held on the base.- A menu radio keeps its circle, and a round ask offers the undo. Verified: both tests fail on the base code; the offered undo runs.
- A light-only theme reads its own
--modeas none. False: the base already does this, and both tests pass on the base code (blocking). - A part Core marks open is drawn open, so
--state expandedreaches a submenu row and a top nav menu. Verified:theme-open-state-drawn.test.mjsand the newlayers.test.mjscase fail on the base code. - States Core marks on a part are settable: a list item's selected, a checkbox or radio box's hover from its holder, a container's state named on its inner part. Verified:
theme-part-states-marked.test.mjsandtheme-under-input.test.mjsfail on the base code. showdraws sample inputs as the theme's text input, a flyout once beside its menu, a press at its moment, a highlighted palette and the drawer in its cell. Verified:theme-pictures-open-layers-samples.test.mjsfails on the base code; the pictures show the flyout and drawer.- Pictures show their whole page, no MobileNav cell on any page, an overlay under its opener and sub-pixel cut text whole. Verified:
theme-pictures-whole-page.test.mjsfails on the base code; the base navigation sheet has the empty MobileNav picture. - Rows said of a part are its own rows, and a part word before an unknown noun is a compound subject. Verified:
theme-rows-and-product-words.test.mjsfails on the base code. - A state's fill is measured against its own variant's text, and a part without text gets no text problem. Verified: both tests fail on the base code.
- A placeholder as faint as the base drew it is no new problem. Unverified: its test passes on the base code too.
- A one-mode theme's text color on an image fill is written for both modes. Verified: two of its three tests fail on the base code.
- A tall family picture is cut between rows into pictures of about 1,600px. Verified against
cutTalland its caller in_context.mjs. - The overlays picture shows about five to a row and a panel cropped to its top. Verified against the overlays sheet styles in
_catalog.mjs. - Editor-only, with no Core, record or official-theme change. Partly false: it also drops
editor show's base-step declaration (blocking). _how-question.mjschanges by +328/-56. Verified.- It adds mood or feel phrase matches. Unverified: no added hunk reads a mood or feel ask (non-blocking).
- Each commit was gated in an integration branch with the full suite, typechecks and an answer key. Unverified: that branch cannot be read here.
- CI has not run on this head, since a stacked part gets CI only once it targets
main. Verified: CI runs only for pull requests intomainand release branches, and none ran. - Any failure found only at this head is fixed here. False while
bases.test.mjsfails; restoring the base step makes it true. - A look gate on a sample site found no new blocking finding. Unverified.
Files:
packages/cli/clients/cli/commands/experimental-theme-editor-show.doc.mjs: drops thethemes.compilebase step (blocking).packages/cli/api/theme-edit/_how-question.mjs: a paired text color loses its fill's mode; the tint pass rewrites the wrong fills (both blocking)..changeset/theme-round15-part-02.md: the one-mode--modesentence describes base behavior (blocking).packages/cli/api/theme-edit/_check.mjs: the under-input refusal's holder list is now always empty (non-blocking)..changeset/experimental-theme-pictures-agents-open-overlays.md,.changeset/experimental-theme-sheet-cells-b.md: clear.packages/cli/api/theme-edit/_catalog.mjs,_context.mjs,_edit.mjs,_parts.mjs,_preview.mjs,_settings.mjs: clear.packages/cli/foundation/render/layers.mjs,page-worker.mjs,picture-chrome.mjs,picture-playwright.mjs,picture-window.mjs: clear.theme-fix-lines-keep-role.test.mjs,theme-state-fill-text-scope.test.mjs: pass here; three of their cases also pass on the base code (non-blocking).theme-bar-border.test.mjs,theme-edit-only-escape-and-light-only-mode.test.mjs,theme-edit.test.mjs,theme-how-clause-scope.test.mjs,theme-how-draws-modes-clearly-light.test.mjs,theme-how-list-items.test.mjs,theme-how-plain-words.test.mjs,theme-how-variant-looks.test.mjs,theme-open-state-drawn.test.mjs,theme-options-written-loosely-words.test.mjs,theme-part-states-marked.test.mjs,theme-pictures-fit-cut-text.test.mjs,theme-pictures-open-layers-samples.test.mjs,theme-pictures-whole-page.test.mjs,theme-refusal-runs-label-swapped.test.mjs,theme-rows-and-product-words.test.mjs,theme-under-input.test.mjs,theme-whole-fill-text.test.mjs(all underpackages/cli/api/theme-edit/),packages/cli/foundation/render/layers.test.mjs: clear; they pass here.
988174c to
4846c24
Compare
4a11932 to
266cb8a
Compare
|
/approve-spec 4846c24 |
There was a problem hiding this comment.
Changes requested. An engineering owner approved the direction of this round at this head, and the four fixes asked for last time are in; two more are needed.
Blocking:
- A tinted fill with a text color writes that text where it cannot be read:
how "buttons tinted blue with white text"gives white text on primary, secondary and ghost in both modes on a 15% blue tint (1.38:1 on the light page), because the text pairing at_how-question.mjs:25493-25514never measures and the tint pass runs after it; tint first and measure the paired text on the tint, writing it only in the modes where it reads. - The description's bullet "… and a light-only theme reads its own --mode as none" claims a fix the base already has (
_check.mjs:4766reads a one-mode theme's own--modeas none, and both one-mode tests pass on the base code), so drop that clause or say the tests now cover it.
Three tests for claimed fixes (the faint link, the one-note-per-cause fill and the loading placeholder) also pass on the base code; a case the base gets wrong would let them prove their change. CI has not run on this head, since it runs only on pull requests into main.
Claims and files checked
Review mode: full; reason: rebase
Claims (release notes and description):
- A border asked of a bar goes on the bar unless its pieces draw borders. Verified: its test fails on the base code and passes here.
- A too-faint role text color is fixed in its own hue. Unverified: its test passes on the base code too.
- A light-only border reports nothing about dark. Verified: its test fails on the base code.
- One fill on a group is noted once per kind of cover. Unverified: its test passes on the base code too.
- A color with a pill shape keeps the selected item's state. Verified: its test fails on the base code.
- A text color for every variant goes only where and when it reads, measured on the fill the answer gives. Partly false: verified for plain fills and a fill given for dark mode, but a tint gets white text at 1.38:1 (blocking).
- A tinted color fills at low strength, on the color the tint word modifies. Verified: the tint tests fail on the base code; the header in "cards tinted blue with a navy header" stays solid.
- A rule with its own verb after a part clause applies to the whole theme. Verified: its test fails on the base code.
--onlyon a part styled on its own is taken and said; a group still refuses it. Verified for the first; the group refusal already held.- A menu radio keeps its circle, and a round ask offers the undo. Verified: both tests fail on the base code; the offered undo runs.
- A part Core marks open is drawn open, so
--state expandedreaches a submenu row and a top nav menu. Verified: its tests fail on the base code. - States Core marks on a part are settable, a checkbox or radio box's hover is shown from its holder, and a container's state names the inner part. Verified: those tests fail on the base code.
showdraws sample inputs as the theme's text input, a flyout once beside its menu, a press at its moment, a highlighted palette and the drawer in its cell. Verified: those tests fail on the base code, and the pictures show the flyout and the palette.- Pictures show their whole page, no empty mobile navigation cell, an overlay under its opener and slightly cut text whole. Verified: those tests fail on the base code; the navigation family is 2 pictures instead of 4 and the toolbar menu sits under its opener.
- Rows said of a part are its own rows, and a part word before an unknown noun starts a compound subject. Verified: those tests fail on the base code.
- A state's fill is measured against its own variant's text. Verified: those tests fail on the base code.
- A placeholder as faint as the base drew it is no new problem. Unverified: its test passes on the base code too.
- A one-mode theme's text color on an image fill is written for both modes. Verified: two of its three tests fail on the base code.
- The overlays picture shows about five to a row with a panel cropped to its top. Verified against the overlays sheet styles.
- A tall family picture is cut between rows into pictures of about 1,600px. Verified against the picture cutter.
- Editor-only, with no Core, record or official-theme change. Verified: every file is in the CLI or a release note, and
editor showkeeps its base step. - A light-only theme reads its own
--modeas none (description bullet). False: the base already does this (blocking). _how-question.mjschanges by +441/-58. Verified.- It adds mood or feel phrase matches. Unverified: no added hunk reads a mood or feel ask (non-blocking).
- Each commit was gated in an integration branch with the full suite and an answer key. Unverified: that branch cannot be read here.
- CI has not run on this head. Verified: no workflow ran.
- Any failure found only at this head is fixed here. Verified for the pull request's tests and the base-gate tests, which all pass.
- A look gate on a sample site found no new blocking finding. Unverified.
Files:
packages/cli/api/theme-edit/_how-question.mjs: a paired text color is never measured and is tinted under (blocking).packages/cli/api/theme-edit/_check.mjs: the under-input refusal's holder list is always empty (non-blocking).packages/cli/api/theme-edit/theme-fix-lines-keep-role.test.mjs,theme-state-fill-text-scope.test.mjs: three cases pass on the base code (non-blocking)..changeset/theme-round15-part-02.md,.changeset/experimental-theme-pictures-agents-open-overlays.md,.changeset/experimental-theme-sheet-cells-b.md: clear.packages/cli/api/theme-edit/_catalog.mjs,_context.mjs,_edit.mjs,_parts.mjs,_preview.mjs,_settings.mjs: clear.packages/cli/clients/cli/commands/experimental-theme-editor-show.doc.mjs: clear (keeps its base step).packages/cli/foundation/render/layers.mjs,page-worker.mjs,picture-chrome.mjs,picture-playwright.mjs,picture-window.mjs: clear.theme-bar-border.test.mjs,theme-edit-only-escape-and-light-only-mode.test.mjs,theme-edit.test.mjs,theme-how-clause-scope.test.mjs,theme-how-draws-modes-clearly-light.test.mjs,theme-how-list-items.test.mjs,theme-how-plain-words.test.mjs,theme-how-tint-and-mode-pairs.test.mjs,theme-how-variant-looks.test.mjs,theme-open-state-drawn.test.mjs,theme-options-written-loosely-words.test.mjs,theme-part-states-marked.test.mjs,theme-pictures-fit-cut-text.test.mjs,theme-pictures-open-layers-samples.test.mjs,theme-pictures-whole-page.test.mjs,theme-refusal-runs-label-swapped.test.mjs,theme-rows-and-product-words.test.mjs,theme-under-input.test.mjs,theme-whole-fill-text.test.mjs(all underpackages/cli/api/theme-edit/),packages/cli/foundation/render/layers.test.mjs: clear; they pass here.
4846c24 to
b32fb8a
Compare
There was a problem hiding this comment.
Approval waits on an engineering owner's decision, and three fixes are needed. The direction of this round of experimental theme editor changes was approved for the previous head; the branch has since been rebased and gained a commit that deepens a tint until the text asked with it reads, so an engineering owner needs to record the decision for this head by commenting /approve-spec <full head SHA> on this pull request (the author may do so as an engineering owner). The tint fix works for buttons and badges.
Blocking:
- A tint still writes unreadable white text on cards and on a selected nav item:
how "cards tinted blue with white text"writes white text tokens over a 15% blue tint (1.38:1 on the light page) andhow "selected nav items tinted forest green with white text"writes white over a 15% green tint (setthen reports 1.28:1), because the new pass at_how-question.mjs:25720-25726measures only per-variantcolorlines. - The description bullet "… and a light-only theme reads its own --mode as none" claims a fix the base already has (
_check.mjs:4766-4777is unchanged and both one-mode tests pass on the base code), so drop that clause or say the tests now cover it. - The description's "
_how-question.mjsline delta: +441/-58" is stale: the file changes by +567/-58 at this head.
Three tests for claimed fixes still pass on the base code, and the tint release note does not mention that a text color asked with the tint now deepens it.
Claims and files checked
Review mode: full; reason: rebase
Claims (release notes and description):
- A border asked of a bar goes on the bar unless its pieces draw borders. Verified: its test fails on the base code and passes here.
- A too-faint role text color is fixed in its own hue. Unverified: its test passes on the base code too.
- A light-only border reports nothing about dark. Verified: its test fails on the base code.
- One fill on a group is noted once per kind of cover. Unverified: its test passes on the base code too.
- A color with a pill shape keeps the selected item's state. Verified: its test fails on the base code.
- A text color for every variant goes only where and when it reads, measured on the fill the answer gives. Partly false: verified for button-style variants, but cards and a selected nav item get white over a tint unmeasured (blocking).
- A tinted color fills at low strength over what is beneath. Verified for a tint alone; with a text color the tint now deepens, to 95% for white on blue, which the note does not say (non-blocking).
- A rule with its own verb after a part clause applies to the whole theme. Verified: its test fails on the base code.
--onlyon a part styled on its own is taken and said; a group still refuses it. Verified for the first; the group refusal already held.- A menu radio keeps its circle, and a round ask offers the undo. Verified: both tests fail on the base code; the offered undo runs.
- A part Core marks open is drawn open, so
--state expandedreaches a submenu row and a top nav menu. Verified: its test fails on the base code. - States Core marks on a part are settable, a checkbox or radio box's hover is shown from its holder, and a container's state names the inner part. Verified: those tests fail on the base code.
showdraws sample inputs as the theme's text input, a flyout once beside its menu, a press at its moment, a highlighted palette and the drawer in its cell. Verified: those tests fail on the base code, and the overlays picture shows the flyout once.- Pictures show their whole page, no empty navigation drawer cell, an overlay under its opener and slightly cut text whole. Verified: those tests fail on the base code; the navigation family is 2 pictures instead of 4 and the toolbar menu sits under its opener.
- Rows said of a part are its own rows, and a part word before an unknown noun starts a compound subject. Verified: those tests fail on the base code.
- A state's fill is measured against its own variant's text. Verified: those tests fail on the base code.
- A placeholder as faint as the base drew it is no new problem. Unverified: its test passes on the base code too.
- A one-mode theme's text color on an image fill is written for both modes. Verified: two of its three tests fail on the base code.
- The overlays picture shows about five to a row with a panel cropped to its top. Verified against the overlays sheet styles.
- A tall family picture is cut between rows into pictures of about 1,600px. Verified: the head's navigation pictures are 1,301 and 1,176px tall.
- Editor-only, with no Core, record or official-theme change. Verified: every file is in the CLI or a release note.
- A light-only theme reads its own
--modeas none (description bullet). False: the base already does this (blocking). _how-question.mjschanges by +441/-58. False: +567/-58 at this head (blocking).- It adds mood or feel phrase matches. Unverified: no added hunk matches a mood or feel phrase.
- Each commit was gated in an integration branch with the full suite, typechecks and an answer key. Unverified: that branch cannot be read here.
- CI has not run on this head. Verified: no CI workflow ran.
- Any failure found only at this head is fixed here. Verified for the pull request's tests, which all pass.
- A look gate on a sample site found no new blocking finding. Unverified.
Files:
packages/cli/api/theme-edit/_how-question.mjs: white text over a tint on cards and a selected nav item is not measured (blocking).packages/cli/api/theme-edit/_check.mjs: the under-input refusal's holder list is always empty (non-blocking).packages/cli/api/theme-edit/theme-fix-lines-keep-role.test.mjs,theme-state-fill-text-scope.test.mjs: three cases pass on the base code (non-blocking)..changeset/theme-round15-part-02.md: the tint note leaves out the deepening (non-blocking)..changeset/experimental-theme-pictures-agents-open-overlays.md,.changeset/experimental-theme-sheet-cells-b.md: clear.packages/cli/api/theme-edit/_catalog.mjs,_context.mjs,_edit.mjs,_parts.mjs,_preview.mjs,_settings.mjs: clear.packages/cli/clients/cli/commands/experimental-theme-editor-show.doc.mjs: clear (keeps its base step).packages/cli/foundation/render/layers.mjs,page-worker.mjs,picture-chrome.mjs,picture-playwright.mjs,picture-window.mjs: clear.theme-bar-border.test.mjs,theme-edit-only-escape-and-light-only-mode.test.mjs,theme-edit.test.mjs,theme-how-clause-scope.test.mjs,theme-how-draws-modes-clearly-light.test.mjs,theme-how-list-items.test.mjs,theme-how-plain-words.test.mjs,theme-how-tint-and-mode-pairs.test.mjs,theme-how-variant-looks.test.mjs,theme-open-state-drawn.test.mjs,theme-options-written-loosely-words.test.mjs,theme-part-states-marked.test.mjs,theme-pictures-fit-cut-text.test.mjs,theme-pictures-open-layers-samples.test.mjs,theme-pictures-whole-page.test.mjs,theme-refusal-runs-label-swapped.test.mjs,theme-rows-and-product-words.test.mjs,theme-under-input.test.mjs,theme-whole-fill-text.test.mjs(all underpackages/cli/api/theme-edit/),packages/cli/foundation/render/layers.test.mjs: clear; they pass here.
b32fb8a to
af337ac
Compare
There was a problem hiding this comment.
Approval waits on an engineering owner's decision, and one fix is needed. The direction of this round of experimental theme editor changes was approved only for an earlier head, before the commit that deepens a tint until the text asked with it reads; the branch has since been rebased with its own patch unchanged, so an engineering owner needs to record the decision for this head by commenting /approve-spec <full head SHA> on this pull request (the author may do so as an engineering owner). The two description claims flagged last time are fixed.
Blocking:
- A tint still writes unreadable white text on cards and on a selected nav item:
how "cards tinted blue with white text"writes white text tokens over a 15% blue tint (1.38:1 on the light page) andhow "selected nav items tinted forest green with white text"writes white over a 15% green tint (setthen reports 1.28:1), because the pass at_how-question.mjs:25720-25726measures only per-variantcolorlines.
Three tests for claimed fixes still pass on the base code, and the tint release note does not mention that a text color asked with the tint now deepens it.
Claims and files checked
Review mode: carry-forward; reason: none
Claims (release notes and description):
- A border asked of a bar goes on the bar unless its pieces draw borders. Verified: its test fails on the base code and passes on this patch.
- A too-faint role text color is fixed in its own hue. Unverified: its test passes on the base code too.
- A light-only border reports nothing about dark. Verified: its test fails on the base code.
- One fill on a group is noted once per kind of cover. Unverified: its test passes on the base code too.
- A color with a pill shape keeps the selected item's state. Verified: its test fails on the base code.
- A text color for every variant goes only where and when it reads, measured on the fill the answer gives. Partly false: verified for button-style variants, but cards and a selected nav item get white over a tint unmeasured (blocking).
- A tinted color fills at low strength over what is beneath. Verified for a tint alone; with a text color the tint now deepens, to 95% for white on blue, which the note does not say (non-blocking).
- A rule with its own verb after a part clause applies to the whole theme. Verified: its test fails on the base code.
--onlyon a part styled on its own is taken and said; a group still refuses it. Verified for the first; the group refusal already held.- A menu radio keeps its circle, and a round ask offers the undo. Verified: both tests fail on the base code; the offered undo runs.
- A part Core marks open is drawn open, so
--state expandedreaches a submenu row and a top nav menu. Verified: its test fails on the base code. - States Core marks on a part are settable, a checkbox or radio box's hover is shown from its holder, and a container's state names the inner part. Verified: those tests fail on the base code.
showdraws sample inputs as the theme's text input, a flyout once beside its menu, a press at its moment, a highlighted palette and the drawer in its cell. Verified: those tests fail on the base code, and the overlays picture shows the flyout once.- Pictures show their whole page, no empty navigation drawer cell, an overlay under its opener and slightly cut text whole. Verified: those tests fail on the base code; the navigation family is 2 pictures instead of 4 and the toolbar menu sits under its opener.
- Rows said of a part are its own rows, and a part word before an unknown noun starts a compound subject. Verified: those tests fail on the base code.
- A state's fill is measured against its own variant's text. Verified: those tests fail on the base code.
- A placeholder as faint as the base drew it is no new problem. Unverified: its test passes on the base code too.
- A one-mode theme's text color on an image fill is written for both modes. Verified: two of its three tests fail on the base code.
- The overlays picture shows about five to a row with a panel cropped to its top. Verified against the overlays sheet styles.
- A tall family picture is cut between rows into pictures of about 1,600px. Verified: the head's navigation pictures are 1,301 and 1,176px tall.
- Editor-only, with no Core, record or official-theme change. Verified: every file is in the CLI or a release note.
- The phrase-rule file changes by +567/-58. Verified: that is its line count change at this head.
- It adds mood or feel phrase matches. Unverified: no added hunk matches a mood or feel phrase.
- Each commit was gated in an integration branch with the full suite, typechecks and an answer key. Unverified: that branch cannot be read here.
- CI has not run on this head. Verified: no CI workflow ran.
- Any failure found only at this head is fixed here. Verified for the pull request's tests, which pass on this patch.
- A look gate on a sample site found no new blocking finding. Unverified.
Files:
packages/cli/api/theme-edit/_how-question.mjs: white text over a tint on cards and a selected nav item is not measured (blocking).packages/cli/api/theme-edit/_check.mjs: the under-input refusal's holder list is always empty (non-blocking).packages/cli/api/theme-edit/theme-fix-lines-keep-role.test.mjs,theme-state-fill-text-scope.test.mjs: three cases pass on the base code (non-blocking)..changeset/theme-round15-part-02.md: the tint note leaves out the deepening (non-blocking)..changeset/experimental-theme-pictures-agents-open-overlays.md,.changeset/experimental-theme-sheet-cells-b.md: clear.packages/cli/api/theme-edit/_catalog.mjs,_context.mjs,_edit.mjs,_parts.mjs,_preview.mjs,_settings.mjs: clear.packages/cli/clients/cli/commands/experimental-theme-editor-show.doc.mjs: clear (keeps its base step).packages/cli/foundation/render/layers.mjs,page-worker.mjs,picture-chrome.mjs,picture-playwright.mjs,picture-window.mjs: clear.theme-bar-border.test.mjs,theme-edit-only-escape-and-light-only-mode.test.mjs,theme-edit.test.mjs,theme-how-clause-scope.test.mjs,theme-how-draws-modes-clearly-light.test.mjs,theme-how-list-items.test.mjs,theme-how-plain-words.test.mjs,theme-how-tint-and-mode-pairs.test.mjs,theme-how-variant-looks.test.mjs,theme-open-state-drawn.test.mjs,theme-options-written-loosely-words.test.mjs,theme-part-states-marked.test.mjs,theme-pictures-fit-cut-text.test.mjs,theme-pictures-open-layers-samples.test.mjs,theme-pictures-whole-page.test.mjs,theme-refusal-runs-label-swapped.test.mjs,theme-rows-and-product-words.test.mjs,theme-under-input.test.mjs,theme-whole-fill-text.test.mjs(all underpackages/cli/api/theme-edit/),packages/cli/foundation/render/layers.test.mjs: clear; they pass on this patch.
af337ac to
9841b07
Compare
There was a problem hiding this comment.
Approval waits on an engineering owner's decision, and two release-note fixes are still needed. The branch was rebased onto the updated base with no change to its own patch, and the whole editor test suite passes on it. No current spec decides what the experimental theme editor reads or writes, and the direction of this round was approved only for an earlier head, so an engineering owner needs to provide or identify the governing spec, or record the decision for this head by commenting /approve-spec <full head SHA> on this pull request; the author may do so as an engineering owner.
Blocking:
- The release note
.changeset/experimental-theme-host-keeps-own-states.mddescribes a Button fix for a problem no release had, because census states arrive in this same pull request and on the base Button has noselectedstate and a list item takes none; delete it, or add one clause to the round note's census line. - The Pagination example at
.changeset/theme-round15-part-02.md:22answers the same on the base, which already namespagination-item; use an example this pull request changes, such as--on list --state selected, which now nameslist-itemwhere the base names no part.
Also worth a look:
- Miss: after you run the
buttons tinted blue with white textanswer,setreports the ghost Cancel button at 1:1 as if it had no fill, and offers a black-text line for it. --on top-nav --state expandedoffers--on button --state expanded, which restyles every expanded Button rather than the top nav.
Claims and files checked
Review mode: full; reason: rebase
Claims (release notes and description):
- A part drawn on another part's element keeps its states, so Button no longer takes the pager cell's selected state while a list item still takes its row's. False as a change: on the base Button has no selected state and a list item takes none (blocking).
- A border asked of a bar goes on the bar unless its pieces draw borders. Verified: its test fails on the base code and passes here.
- A border set for light alone reports nothing about dark. Verified: its test fails on the base code and passes here.
- A colour with a pill shape keeps the selected item's state. Verified: the base writes the pill corners with no state; its test passes here.
- A text colour for every variant goes to each variant only in the modes it reads in. Verified: its test fails on the base code.
- A tint is the colour at low strength; it deepens until a text colour asked with it reads, otherwise the text is kept to the modes it reads in and an ink is named. Verified: buttons deepen to 95%, and the yellow case keeps white to dark mode with an ink named.
- A rule with its own verb after a part's clause applies to the whole theme. Verified: its test fails on the base code.
--onlyon a part styled on its own is taken and noted, and a group still refuses it. Verified for the first; the group refusal already held.- A menu radio keeps its circle, and a round ask offers the undo. Verified: both tests fail on the base code.
- A part Core marks open is drawn open, so
--state expandedreaches a submenu row and a top nav menu. Verified: its test fails on the base code. - A list item's selected state and a checkbox or radio box's hover can be set. Verified: those tests fail on the base code.
- A container's state names the part inside it, shown with the Pagination example. Verified for lists, checkbox lists, menus and the top nav; the Pagination example answers the same on the base (blocking).
showdraws sample text fields as the theme's input, a flyout once beside its menu, a press at its moment, a highlighted palette and the drawer in its cell. Verified: those tests fail on the base code, and the overlays picture shows the flyout once, the highlighted palette and the drawer in its cell.- Pictures show their whole page, no empty drawer cell, an overlay under its opener, and slightly cut text whole. Verified: those tests fail on the base code, and the overlays picture shows the toolbar menu under its opener.
- Rows said of a part are its own rows, and a part word before an unknown noun starts a compound subject. Verified: those tests fail on the base code.
- A state's fill is measured against its own variant's text. Verified: those tests fail on the base code.
- A placeholder as faint as the base drew it is no new problem. Unverified: its test passes on the base code too.
- A one-mode theme's text colour on an image fill is written for both modes. Verified: two of its three tests fail on the base code.
- The overlays picture shows about five to a row, with a panel cropped to its top. Verified against the overlays sheet styles.
- A tall family picture is cut between rows into pictures of about 1,600px, a taller row being its own picture. Verified: the picture cutter does this.
- The change is editor-only, with no Core, record or official-theme change. Verified: every file is in the CLI or a release note.
- It is part 2 of a stack based on the first part's branch. Verified: the base branch is that part.
- The phrase-rule file changes by +625/-58. Verified.
- It adds no mood or feel phrase rules. Verified:
feelandfeelsappear only in the list of joining words. - A faint link keeps its hue, and one fill on a group is noted once per cause. Unverified: both tests pass on the base code too.
- Each commit was gated in an integration branch with the full suite, typechecks and an answer key. Unverified: that branch is not in this repository.
- CI has not run on this head. Verified: no CI workflow ran.
- Each head is checked offline before a push, including the node tests the change reaches. Verified for the node tests only: the whole editor and render suite passes here.
- A look gate on a sample site found no new blocking finding. Unverified.
Files:
.changeset/experimental-theme-host-keeps-own-states.md: describes a change the base never needed (blocking)..changeset/theme-round15-part-02.md: the Pagination example answers the same on the base (blocking)..changeset/experimental-theme-pictures-agents-open-overlays.md,.changeset/experimental-theme-sheet-cells-b.md: clear.packages/cli/api/theme-edit/_check.mjs: the container lookup offers a theme-widebuttonline for the top nav, and the under-input refusal's holder list is always empty (non-blocking).packages/cli/api/theme-edit/_how-question.mjs: the tinted ghost text is measured only on its own fill, sosetdisagrees with the answer (non-blocking).packages/cli/api/theme-edit/_catalog.mjs,_context.mjs,_edit.mjs,_parts.mjs,_preview.mjs,_settings.mjs: clear.packages/cli/clients/cli/commands/experimental-theme-editor-show.doc.mjs: clear (keeps its base step).packages/cli/foundation/render/layers.mjs,layers.test.mjs,page-worker.mjs,picture-chrome.mjs,picture-playwright.mjs,picture-window.mjs: clear.theme-fix-lines-keep-role.test.mjs,theme-state-fill-text-scope.test.mjs,theme-part-states-marked.test.mjs(underpackages/cli/api/theme-edit/): cases for claimed fixes pass on the base code, and one comment still says Button has the pager's state (non-blocking).theme-bar-border.test.mjs,theme-edit-only-escape-and-light-only-mode.test.mjs,theme-edit.test.mjs,theme-how-clause-scope.test.mjs,theme-how-draws-modes-clearly-light.test.mjs,theme-how-list-items.test.mjs,theme-how-plain-words.test.mjs,theme-how-tint-and-mode-pairs.test.mjs,theme-how-variant-looks.test.mjs,theme-list-end-corners-scale.test.mjs,theme-open-state-drawn.test.mjs,theme-options-written-loosely-words.test.mjs,theme-pictures-fit-cut-text.test.mjs,theme-pictures-open-layers-samples.test.mjs,theme-pictures-whole-page.test.mjs,theme-refusal-runs-label-swapped.test.mjs,theme-rows-and-product-words.test.mjs,theme-under-input.test.mjs,theme-whole-fill-text.test.mjs(underpackages/cli/api/theme-edit/): clear; they pass here.
8983a94 to
e4738b2
Compare
83c6f48 to
6e75450
Compare
There was a problem hiding this comment.
Approval waits on an engineering owner's decision, and two release-note fixes are still needed. The branch was rebased onto the updated base with no change to its own patch, and the whole editor test suite passes on it. No current spec decides what the experimental theme editor reads or writes, and the direction of this round was approved only for an earlier head, so an engineering owner needs to provide or identify the governing spec, or record the decision for this head by commenting /approve-spec <full head SHA> on this pull request; the author may do so as an engineering owner.
Blocking:
- The release note
.changeset/experimental-theme-host-keeps-own-states.mddescribes a Button fix for a problem no release had, because census states arrive in this same pull request and on the base Button has noselectedstate and a list item takes none; delete it, or add one clause to the round note's census line. - The Pagination example at
.changeset/theme-round15-part-02.md:22answers the same on the base, which already namespagination-item; use an example this pull request changes, such as--on list --state selected, which now nameslist-itemwhere the base names no part.
Also worth a look:
- Miss: after you run the
buttons tinted blue with white textanswer,setreports the ghost Cancel button at 1:1 as if it had no fill, and offers a black-text line for it. - The container lookup offers shared parts beside the container's own row, such as
--on button --state expandedfor the top nav, and each of those lines restyles every Button or Item rather than the container.
Claims and files checked
Review mode: full; reason: rebase
Claims (release notes and description):
- A part drawn on another part's element keeps its states, so Button no longer takes the pager cell's selected state while a list item still takes its row's. False as a change: on the base Button has no selected state and a list item takes none (blocking).
- A border asked of a bar goes on the bar unless its pieces draw borders. Verified: its test fails on the base code and passes here.
- A border set for light alone reports nothing about dark. Verified: its test fails on the base code and passes here.
- A colour with a pill shape keeps the selected item's state. Verified: the base writes the pill corners with no state; its test passes here.
- A text colour for every variant goes to each variant only in the modes it reads in. Verified: its test fails on the base code.
- A tint is the colour at low strength; it deepens until a text colour asked with it reads, otherwise the text is kept to the modes it reads in and an ink is named. Verified: buttons and cards deepen to 95%, and the forest-green nav item keeps white to dark mode with an ink named.
- A rule with its own verb after a part's clause applies to the whole theme. Verified: its test fails on the base code.
--onlyon a part styled on its own is taken and noted, and a group still refuses it. Verified for the first; the group refusal already held.- A menu radio keeps its circle, and a round ask offers the undo. Verified: both tests fail on the base code.
- A part Core marks open is drawn open, so
--state expandedreaches a submenu row and a top nav menu. Verified: its test fails on the base code. - A list item's selected state and a checkbox or radio box's hover can be set. Verified: those tests fail on the base code.
- A container's state names the part inside it, shown with the Pagination example. Verified for lists, checkbox lists, menus and the top nav; the Pagination example answers the same on the base (blocking).
showdraws sample text fields as the theme's input, a flyout once beside its menu, a press at its moment, a highlighted palette and the drawer in its cell. Verified: those tests fail on the base code, and the pictures show the field, the flyout and the palette.- Pictures show their whole page, no empty drawer cell, an overlay under its opener, and slightly cut text whole. Verified: those tests fail on the base code, and the pictures show the toolbar menu under its opener and no empty drawer cell.
- Rows said of a part are its own rows, and a part word before an unknown noun starts a compound subject. Verified: those tests fail on the base code.
- A state's fill is measured against its own variant's text. Verified: those tests fail on the base code.
- A placeholder as faint as the base drew it is no new problem. Unverified: its test passes on the base code too.
- A one-mode theme's text colour on an image fill is written for both modes. Verified: two of its three tests fail on the base code.
- The overlays picture shows about five to a row, with a panel cropped to its top. Verified against the overlays sheet styles.
- A tall family picture is cut between rows into pictures of about 1,600px, a taller row being its own picture. Verified: the picture cutter does this.
- The change is editor-only, with no Core, record or official-theme change. Verified: every file is in the CLI or a release note.
- It is part 2 of a stack based on the first part's branch. Verified: the base branch is that part.
- The phrase-rule file changes by +625/-58. Verified.
- It adds no mood or feel phrase rules. Verified:
feelandfeelsappear only in the list of joining words. - A faint link keeps its hue, and one fill on a group is noted once per cause. Unverified: both tests pass on the base code too.
- Each commit was gated in an integration branch with the full suite, typechecks and an answer key. Unverified: that branch is not in this repository.
- CI has not run on this head. Verified: no CI workflow ran.
- Each head is checked offline before a push, including the node tests the change reaches. Verified for the node tests only: the whole editor and render suite passes here.
- A look gate on a sample site found no new blocking finding. Unverified.
Files:
.changeset/experimental-theme-host-keeps-own-states.md: describes a change the base never needed (blocking)..changeset/theme-round15-part-02.md: the Pagination example answers the same on the base (blocking)..changeset/experimental-theme-pictures-agents-open-overlays.md,.changeset/experimental-theme-sheet-cells-b.md: clear.packages/cli/api/theme-edit/_check.mjs: the container lookup offers theme-widebuttonanditemlines, and the under-input refusal's holder list is always empty (non-blocking).packages/cli/api/theme-edit/_how-question.mjs: the tinted ghost text is measured only on its own fill, sosetdisagrees with the answer (non-blocking).packages/cli/api/theme-edit/_catalog.mjs,_context.mjs,_edit.mjs,_parts.mjs,_preview.mjs,_settings.mjs: clear.packages/cli/clients/cli/commands/experimental-theme-editor-show.doc.mjs: clear (keeps its base step).packages/cli/foundation/render/layers.mjs,layers.test.mjs,page-worker.mjs,picture-chrome.mjs,picture-playwright.mjs,picture-window.mjs: clear.theme-fix-lines-keep-role.test.mjs,theme-state-fill-text-scope.test.mjs,theme-part-states-marked.test.mjs(underpackages/cli/api/theme-edit/): cases for claimed fixes pass on the base code, and one comment still says Button has the pager's state (non-blocking).theme-bar-border.test.mjs,theme-edit-only-escape-and-light-only-mode.test.mjs,theme-edit.test.mjs,theme-how-clause-scope.test.mjs,theme-how-draws-modes-clearly-light.test.mjs,theme-how-list-items.test.mjs,theme-how-plain-words.test.mjs,theme-how-tint-and-mode-pairs.test.mjs,theme-how-variant-looks.test.mjs,theme-list-end-corners-scale.test.mjs,theme-open-state-drawn.test.mjs,theme-options-written-loosely-words.test.mjs,theme-pictures-fit-cut-text.test.mjs,theme-pictures-open-layers-samples.test.mjs,theme-pictures-whole-page.test.mjs,theme-refusal-runs-label-swapped.test.mjs,theme-rows-and-product-words.test.mjs,theme-under-input.test.mjs,theme-whole-fill-text.test.mjs(underpackages/cli/api/theme-edit/): clear; they pass here.
…of the commit that makes it pass
… the part's to set - A runtime state the census finds a part's own element marked with is the part's state, though its doc lists it on another target the element also carries: a list item rides an Item, so a list item (a checkbox list's row too) takes --state selected, written as the data attribute Core marks. - A part drawn under an invisible input (a checkbox's or radio's box, a switch) takes hover, active and the focus states from the part holding the input: the build writes them as that part's state around it, :where(.astryx-<holder>:hover *), instead of refusing them. Only a part whose input no themed part holds is still refused. - A state asked of a container that it does not have names the part drawn inside it that has it, its own pieces first, then the parts the census finds inside it: --on pagination --state current points to --on button --state current, the current page's button. - The census reads aria-expanded="true" as a condition, and takes it from the opened overlays for a part the page draws too, so a part drawn open records its open state for --state expanded. - Tests: new theme-part-states-marked; theme-under-input now expects the held states written.
…expanded reaches a submenu row and a top nav menu - The overlay walk keeps a control the page itself draws (a top nav's menu trigger, outside every overlay) as drawn once a press opens it: the part with that control open, without its overlays. - The census reads aria-expanded="true" as a condition even on a part marked with a runtime state of its own: a submenu's row is highlighted while its flyout is open, and open too. - Together with the census taking the open state from the drawn overlays for a part the page draws too, a dropdown menu's submenu row and a top nav's menu trigger take --state expanded, built as :is([aria-expanded="true"]). - Tests: new theme-open-state-drawn; layers.test keeps a control on the page that a press opens as drawn open.
… keeps its circle apart, and a light-only theme reads its own --mode as none
… light-only border reports nothing about dark, one fill on a group says each kind of cover once, a state fill is measured against the text that state draws in its own variant, and the placeholder as faint as the base drew it is no new problem
…ected item's pill keeps its state
… theme's input; a flyout drawn once beside its menu; a pressed control at the moment of its press; a highlighted command palette; a drawer's panel held in its cell
… whole; no closed drawer cell on any page; an overlay drawn after its part sits under its opener; sub-pixel cut text gets room
…ces draw the borders - borderParts hands a part's border to the pieces inside it only where those pieces draw a border of their own (a border style or width that is not none or zero), as a date and time field's two boxes do. A top nav's items and a sidebar's links are rounded but draw no border, so the bar keeps it. - The ask reader moves a border to inner pieces through borderParts too (corners still through cornerParts), so "top nav with a 1px bottom border" writes border-bottom on top-nav and "sidebar with a 1px right border" border-right on side-nav, never one under or beside each item. - New theme-bar-border.
…word that names what a product is about names no part - Rows said of a part that holds no table rows are that part's own rows: the parts inside it that ride an Item, the element Core draws every list and menu row with (found from the census, where the row carries Item's own attributes). "menus with 32px rows" sizes the menu items, never every table row; rows alone, or a table's, stay the table's. - A part word followed by at most two words that name no part and then a product noun (site, app, view, page, screen, portal, dashboard, platform, store, shop, layout) names what the product is about, never that part: "podcast and radio station site: …" reads no radios, and the rest of the ask is the whole theme's. - New theme-rows-and-product-words.
… written for both modes
…re it reads, in each mode, on the variant's own fill or the fill asked with it; a tint is the color at a low strength; a rule with its own verb after a part's clause is the theme's
…ows as nothing begins a compound subject
- Replaces the list of product nouns: a part word right before a word the editor knows as no property, state, part, value or joining word, with no asking word after it before the clause's next stop, is one half of a compound subject ("radio station", "button factory", "card game"), never that part. The clause keeps the compound, which the unknown-subject guard names, writing nothing for it.
- One knownWord reader holds what the editor knows as a word; the unknown-slot guard and the clause reader both use it. A word an asking word follows ("links turn red", "the card zorp border") stays the ask's verb or a word between, as before.
- "radio buttons bold", "button text" and "card border" read as before; a table test covers them with the compounds.
Its assertions read the drawn open states page (a submenu's flyout drawn beside its menu), which the render stand-in does not draw.
…its fill's mode; a tint tints only the color it modifies
…ill word, so a navy header is its own ask
… deepens along its hue until the text reaches 4.5:1, else the text keeps to the modes it reads in
…e tint fills, scoped to the variants the fill reaches; a sharper ask on a menu radio offers no rounding line; a part with one unknown word keeps the part
… cell (pagination-item, selected) on main's Core; Button's current state no longer appears in Pagination Main's Core (Pagination theme targets) gives the pager's cells their own target, pagination-item, whose current page reflects selected. The expectation follows: the answer is right on main's Core. Expectation only, no reader change.
… switch takes focus-within from its field - The census never records a disabled rule under an element drawn open (aria-expanded="true"): what opens is enabled, so Core's own pairs decide what several states at once paint, and a button is never painted disabled and expanded at once (theme-states-core-paints-aria). Its disabled rules are recorded where Core draws it closed. - theme-list-end-corners-scale: a switch's track takes focus-within from its field, which holds the input (main's Core: the field matches :focus-within while its input has focus), as this part made a radio's box do; the refusal for a part nothing inside of takes focus is checked on a progress bar.
…states to itself (moved down from P09); the selected-page example names the item it points to
e4738b2 to
a7998c1
Compare
6e75450 to
244b54a
Compare
Summary: experimental theme: states Core marks, open parts drawn open, and picture inputs and flyouts. Editor-only (experimental theme editor in
packages/cli); no Core, spec record or official-theme change.Part 2 of 25 of round 15, landing as a stack: based on
theme-r15-p01; merge in order (the next part retargets onto main after this one merges).What changes
Phrase rules
_how-question.mjsline delta: +625/-58.Test plan
CI=1, the five cli typechecks,check-cli-structure,check:changesets, and the held-outhowanswer key (252 right, 0 wrong writes).