Repository navigation
Port picker language icons, reference spacing fixes, and menus and type for every display - #332
Merged
Merged
Conversation
why: The language selector identifies every port by its language's own mark except these three, whose icon.svg has been the libtmux library tile since each port was added. Design record 0070 sources a mark for each. what: - ocaml: colour-icon.svg from ocaml/ocaml-logo at 1892364, unmodified (Unlicense) - raku: Camelia.svg from Raku/doc-website at 7f0554a, unmodified, with the Artistic License 2.0 text as LICENSE.txt (copyright 2009 Larry Wall; Camelia is his trademark, labelled only when used as a page's main logo) - lisp: Conrad Barski's public-domain Lisp alien, its empty A4 canvas cropped to the drawing; lisplogo_alien.svg keeps the original - provenance.json for each, in the existing schema: pinned source, page, retrieval time, served-file SHA-256, terms, modifications, and the retained source copy for lisp - meta/third-party-notices: rows for the three, and Common Lisp beside Scala as a cropped copy
…icker why: Adoption step 35 (design record 0030) shrank the rows' libtmux marks to 20px and widened the panel to 32rem on every page. On a port home, whose rows are names only, each row became about 240px wide, and 20px marks that differ only in colour do not tell the languages apart. Design record 0068 puts the language's own icon in each row at 26px and widens the panel only for rows that carry a second line. what: - Rows draw /brand/languages/<slug>/icon.svg at 26px in place of the libtmux mark at 20px; the tmux CLI row's logo is 26px too - The trigger's Python offset, dark-theme Rust inversion and forced-colors plate now cover the row icons - The panel keeps the shared 23.75rem picker width unless a row shows an API among several or its differences (data-details), which widens it to 32rem - design-tokens test and check-home-launcher assert the new rows
…rk keeps its gap why: The rules that space a member's kind mark from its name, set it on the name's baseline and fill it on hover select `.api-member-list li > span:first-child`. Rows with legacy anchors emitted those hidden anchor spans first, so on pages such as Perl's LibTmux::Workspace::X::Config the "m" mark touched the name. what: - Emit a row's hidden anchors after its head span - design-tokens test: the head span comes before the anchors
…peats its declaration why: A single-signature method with no prose, extension or separate source rendered a "Methods" heading over an empty entry, as on Perl's LibTmux::Workspace::X::Config::matched_ids. what: - apiMethodHasContent says whether a method entry shows anything beyond the declaration - ApiEntry keeps the entries' anchors and inventory attributes but renders the heading only when one has content; the page outline follows - api-signature-scope test covers both cases
… up row why: With no previous, next or related groups, the page end drew its solid top border and then the up row's dashed border with nothing between them. what: - The up row drops its dashed border and padding when it is the page end's first child - design-tokens test pins the rule
why: Every picker opened as a popover under its trigger on every display, so a landscape phone squeezed the language panel into a 336px strip under the header, and a phone in portrait got an anchored box. Design record 0071 opens menus by display: a bottom sheet on phones in portrait, a side sheet on short landscape screens, the popover elsewhere. what: - lib/breakpoints.ts: BP.large, SHORT and PANEL_QUERIES, beside the existing breakpoints - documentation-navigation.ts: on open, a picker sets data-panel-mode from PANEL_QUERIES and data-panel-side from its trigger; sheets leave placement to CSS, lock page scroll (html[data-sheet-open]) and close on a tap on their backdrop; without matchMedia a picker stays a popover - documentation-picker.css: the bottom sheet (full width, up to 85dvh, grab bar, safe-area padding), the side sheet (full height, min(28rem, 70vw), from the trigger's side), the scrim backdrop, a pinned search, 120ms and 200ms entrances that reduced motion skips, and touch-size rows, triggers and home buttons under pointer: coarse at any width; its four sub-12px labels rise to 0.75rem - SearchModal: full screen below bp-phone and on short landscape screens - Tests: the breakpoint queries, sheet versus popover on open and close, and the workspace picker page now inlines lib/breakpoints.ts before the navigation script, as its helper expects
…pe in rem why: Measured on libtmux.org at ten displays, nothing grew past 1280px: the root stayed 16px and the language panel 512px wide on a 3840px screen, and API pages carried up to 168 visible labels under 12px. Design record 0071 scales the root from 1920px, keeps text at 12px or more, and gives touch screens full-size targets. what: - tokens.css: bp-large (120rem) and bp-short (30rem tall) - global.css: the root is clamp(1rem, 1rem + (100vw - 120rem) / 480, 1.25rem), 16px to 1920px and 20px from 3840px, so every rem size grows - 76 px font sizes and shorthands in 13 files, and the reroll button's px heights, move to rem at the same base size, so they grow too - Text set below 0.75rem rises to it, except glyphs drawn in fixed boxes (the player's step numbers, panel badges); the vendored gp-sphinx stylesheet and the demo pages keep theirs - Under pointer: coarse, the header's chips, scheme button and wordmark keep their drawn size and take a touch-target-tall hit area (::before, since the external-link rule owns ::after)
tony
temporarily deployed
to
docs-preview-cleanup
November 18, 2026 00:33 — with
GitHub Actions
Inactive
tony
added a commit
that referenced
this pull request
Nov 18, 2026
… as happy-dom's why: The sheet-mode test from #332 cast its matchMedia stub to the DOM library's MediaQueryList, which astro check rejects against happy-dom's Window.matchMedia; the error failed main's test job at f772074 and every branch based on it. what: - Cast the stub to ReturnType<typeof win.matchMedia>
tony
added a commit
that referenced
this pull request
Nov 18, 2026
… as happy-dom's why: The sheet-mode test from #332 cast its matchMedia stub to the DOM library's MediaQueryList, which astro check rejects against happy-dom's Window.matchMedia; the error failed main's test job at f772074 and every branch based on it. what: - Cast the stub to ReturnType<typeof win.matchMedia> (cherry picked from commit 740e1e0)
tony
added a commit
that referenced
this pull request
Nov 18, 2026
…its drawn height why: #332 gave every picker trigger a touch-target min-height below the phone breakpoint. In the header that row wraps under the bar at 35rem, so the bar grew past the 97px the responsive-header check allows at 390px, and the docs test job failed on #331. what: - The compact and coarse min-height skips the header's trigger - Under pointer: coarse the header's trigger takes the same hit-area pseudo-element as the header's chips, so its tap is 44px tall while the bar keeps its height
tony
added a commit
that referenced
this pull request
Nov 18, 2026
…its drawn height why: #332 gave every picker trigger a touch-target min-height below the phone breakpoint. In the header that row wraps under the bar at 35rem, so the bar grew past the 97px the responsive-header check allows at 390px, and the docs test job failed on #331. what: - The compact and coarse min-height skips the header's trigger - Under pointer: coarse the header's trigger takes the same hit-area pseudo-element as the header's chips, so its tap is 44px tall while the bar keeps its height
This was referenced Nov 18, 2026
tony
added a commit
that referenced
this pull request
Nov 18, 2026
why: Since #332, pickers open as a bottom sheet on phones in portrait and a side sheet on short landscape screens (design record 0071). Two browser checks still assumed a popover: the picker-filters check wanted 7px from every edge at 320 x 360, where the sheet meets the edges by design, and the header-menu check closed a picker by clicking its trigger, which the sheet covers. what: - check-picker-filters: bounds reads the panel mode and keeps the 7px inset for popovers only - check-header-menu-sections: a sheet closes with Escape; a popover still closes from its trigger
tony
added a commit
that referenced
this pull request
Nov 18, 2026
why: Since #332, pickers open as a bottom sheet on phones in portrait and a side sheet on short landscape screens (design record 0071). Two browser checks still assumed a popover: the picker-filters check wanted 7px from every edge at 320 x 360, where the sheet meets the edges by design, and the header-menu check closed a picker by clicking its trigger, which the sheet covers. what: - check-picker-filters: bounds reads the panel mode and keeps the 7px inset for popovers only - check-header-menu-sections: a sheet closes with Escape; a popover still closes from its trigger
tony
added a commit
that referenced
this pull request
Nov 18, 2026
…ng and ocaml (#331) why: perl, raku, erlang and ocaml added "What is on here" capability directories to their Session, Window and Pane types (libtmux-perl#24, libtmux-raku#19, libtmux-erlang#24, libtmux-ocaml#18). These four publish from the revision of the committed API model, so the directories, and perl's examples without the test-only LIBTMUX_TEST_TMUX, reach libtmux.org only once that data is regenerated at each port's head. what: - Regenerate perl, raku, erlang and ocaml source-bound data at 1f8332193, cce868d81, 48aacd01a and 1611f141a - Map OCaml's options, panes and clients showcases; run the OCaml and Raku exporters from their new paths, falling back to the old ones - Keep the header's language trigger its drawn height, type the matchMedia stub, and let the picker checks meet bottom and side sheets (follow-ups to #332's phone sheets)
This was referenced Nov 18, 2026
This was referenced Nov 18, 2026
This branch was previously deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Shows each language's own icon in the port picker rows again, at the size the row artwork had before, and stops the panel from being 32rem wide on pages whose rows are only names. OCaml, Common Lisp and Raku get their own marks; until now their language icon was the libtmux tile.
Why
Adoption step 35 (design record 0030,
fe7ff6102) shrank the rows' libtmux marks to 20px and widened the panel to 32rem on every page. On a port home the rows are names only, so each row became about 240px wide, and 20px marks that differ only in colour don't tell languages apart. Design records 0068 and 0070 (libtmux/design, adoption step 45) settle the fix.What changes
PagePortSwitcher.astro: rows draw/brand/languages/<slug>/icon.svgat 26px instead of the libtmux mark at 20px; the tmux CLI row's logo is 26px too. The trigger's Python offset, dark-theme Rust inversion and forced-colors plate now cover the row icons.data-details), which widens it to 32rem.site/public/brand/languages/ocaml: the OCaml camel from ocaml/ocaml-logo, unmodified, Unlicense.site/public/brand/languages/raku: Camelia from Raku/doc-website, unmodified, © 2009 Larry Wall under the Artistic License 2.0 (camelia.txt), with the licence text asLICENSE.txt. Camelia is a trademark; a TM label is needed only when it is a page's main logo.site/public/brand/languages/lisp: Conrad Barski's public-domain Lisp alien, its empty A4 canvas cropped to the drawing (root width, height and viewBox only);lisplogo_alien.svgkeeps the original, as Scala keepsscala-spiral.svg. Common Lisp has no official logo.meta/third-party-notices.md: rows for the three, and Common Lisp beside Scala as a cropped copy.design-tokens.test.tsandcheck-home-launcher.mjsassert the new rows; the launcher check used to assert that menu rows kept libtmux artwork.Reference spacing fixes
Found on Perl's Workspace Manager reference, and general to every port:
ApiMemberList.astro: a member row's hidden legacy anchors came before its head span, so the rules that space the kind mark from the name (.api-member-list li > span:first-child) never matched and the "m" touched the name (LibTmux::Workspace::X::Config). Anchors now follow the head; the gap is 8px again.ApiEntry.astroandapiMethodHasContent: a lone method that repeats its declaration no longer draws an empty "Methods" heading (matched_ids); its anchor and inventory attributes stay, and the outline drops the entry.PageEnd.astro: a page end holding only the up row no longer draws a dashed rule under its solid border.Every display and orientation
Design record 0071 (libtmux/design, adoption step 46). An audit of libtmux.org at ten displays (360 × 740 to 3840 × 2160, portrait and landscape; ten pages each, in Firefox) found no sideways scrolling, but nothing growing past 1280px (16px text and a 512px language panel at 4K), a 336px strip of panel on landscape phones, and 25 controls under 44px on a 1024 × 768 tablet.
global.csssets the root toclamp(1rem, 1rem + (100vw - 120rem) / 480, 1.25rem): 16px to 1920px, 17.3px at 2560, 19.2px at 3440, 20px from 3840. Every rem size grows with it; the 76 px font sizes and the reroll button's px heights move to rem at the same base size.documentation-navigation.tspicks a mode on open fromPANEL_QUERIES: a bottom sheet on phones in portrait (full width, up to 85dvh, grab bar, pinned search), a side sheet on short landscape screens (full height from the trigger's side), the popover elsewhere. Sheets get the scrim backdrop, lock page scroll and close on a backdrop tap; entrances are 120ms (popover) and 200ms (sheets) and skipped under reduced motion.SearchModalis full screen on phones and short screens.pointer: coarse, picker rows, triggers and home buttons reach--touch-targetat any width, and the header's chips, scheme button and wordmark take a touch-target-tall hit area without growing the bar.Verification
LIBTMUX_DOCS_PORT=rs): on the Rust home the panel is 380px with 26px icons, all 23 loaded; onpane.Paneit is 512px with the "Differs" lines; dark theme inverts the Rust row icon like the trigger; no horizontal overflow.vitest run test/design-tokens.test.ts test/branding.test.ts: 91 passed (the branding test checks each icon's SHA-256 against its provenance); with the spacing fixes, design-tokens, api-signature-scope, chrome-page-end and branding: 117 passed.LIBTMUX_DOCS_PORT=perl): the Config member rows have an 8px mark gap;matched_idshas no Methods heading or outline entry, its method anchor remains; a page end with only the up row computes no top border on the row.tmux-usage-raku.outer.test.ts(the Raku source scanner), which this change doesn't touch.pnpm test:mediumis cancelled by its 10s budget on this host (10.8s even at 8 cores) while runningpackages/api-model's suite; the suites pass when run without the deadline.api-prose-fieldsand the Juliareference-page-linksouter case failed under the parallel run and pass alone;tmux-usage-raku.outerfails as before. The workspace picker suite (64) passes once its page inlineslib/breakpoints.ts.