{"entries":[{"lastCommit":{"author":{"email":"larsvandartel73@gmail.com","name":"LarsvanDartel","when":"2026-08-18T22:21:56.000000+02:00"},"committer":{"email":"larsvandartel73@gmail.com","name":"LarsvanDartel","when":"2026-08-18T22:21:56.000000+02:00"},"hash":"fdd27e8bd9d61cea233dba05060daeb4f09dc22d","message":"Fit the title, footer and videos on one slide\n\nThree things, all downstream of adding a slide title last commit.\n\nThe footer ignored the page's horizontal margin and hugged the edge: touying's\n`zero-margin-footer` defaults to true, which pads the footer out to full page\nwidth. Header and footer now sit inside the margin and line up with the body.\n\nVideos ended up alone on untitled pages. touying spills overflowing content onto\nextra pages by default, and the title I added consumed just enough height to tip\nthe video slides over — turning a 12-page deck into 14, two of them mystery\nslides with no heading. A talk wants one authored slide per slide, so the theme\nsets `breakable: false`; overflow now raises a warning, which presentypst\nalready shows in both views, instead of silently inventing a page.\n\nThe title is smaller (1.05em, tighter spacing) and the top margin is trimmed,\nwhich is what actually buys the video slides their room back.\n\nTwo things this surfaced, fixed here too:\n\n- A focus slide inherited the previous slide's title, because freezing the slide\n counter makes it share a logical number. Only genuine subslides inherit now.\n- touying's overflow warning arrived as line noise. Packages raise custom\n warnings by provoking \"unknown font family:\" and backspacing over the prefix,\n which reads fine in a terminal and not at all in a browser; the erased span is\n now dropped.\n\nThe demo deck is back to 12 pages, both videos sit on their titled slides, and\nit compiles with no warnings.\n\nCo-Authored-By: Claude Opus 5 \n","tree":"0ab7ea00108190dbf040b25195bc93662788468a"},"mode":"0100644","name":"compile.rs","size":11872},{"lastCommit":{"author":{"email":"larsvandartel73@gmail.com","name":"LarsvanDartel","when":"2026-08-18T17:07:29.000000+02:00"},"committer":{"email":"larsvandartel73@gmail.com","name":"LarsvanDartel","when":"2026-08-18T17:07:29.000000+02:00"},"hash":"a1cb06bd26fb78feb2081f5014fed568082c4058","message":"Route poster images through the media store, and guard page fractions\n\nTwo fixes found reviewing the compile pipeline:\n\n- A `poster:` is a path relative to the deck, but it was handed to the\n browser verbatim, which would simply 404. It now registers with the media\n store like any other file and gets a real URL.\n- Point-to-fraction conversion is now a guarded helper. A degenerate page\n would previously produce infinity or NaN, which serializes to null and\n leaves the browser positioning an overlay at `NaN%`.\n\nCo-Authored-By: Claude Opus 5 \n","tree":"cc1c8c90cd1ce3abf12f2f80808843dc2b023aa4"},"mode":"0100644","name":"deck.rs","size":4976},{"lastCommit":{"author":{"email":"larsvandartel73@gmail.com","name":"LarsvanDartel","when":"2026-08-18T17:39:30.000000+02:00"},"committer":{"email":"larsvandartel73@gmail.com","name":"LarsvanDartel","when":"2026-08-18T17:39:30.000000+02:00"},"hash":"6a148c6d6889179537da641bc42fd2c168643605","message":"Fix the start prompt swallowing keys, and the YouTube 403\n\nTwo problems reported from real use.\n\nThe display's start overlay only listened for a click. Pressing space — the\nfirst thing anyone does at a lectern — navigated the deck behind an overlay\nthat then never went away, which is exactly how it was described. Any gesture\nnow dismisses it: key, pointer, touch or scroll. Two related changes fall out\nof that: it no longer forces fullscreen, which is a jarring thing to do to\nsomeone's screen uninvited (`f` still asks for it), and dismissing now primes\nthe video elements while the gesture is in hand, which is the actual point of\nthe overlay rather than a side effect of it.\n\nThe YouTube 403 was yt-dlp's default player-client set leading with\n`android_vr`, which YouTube currently refuses. Naming clients is not just a\npreference order — yt-dlp pools formats across every client listed and picks\nthe best overall, so an earlier attempt at `mweb,android,default` still failed:\nthe broken client's format won the selection. The default is now `mweb,android`\nalone, and `--yt-dlp-arg` lets a presenter route around the next breakage\nwithout waiting for a release, since this will certainly rot again. A 403 or a\nmissing format now says so instead of surfacing a bare HTTP code.\n\nVerified with WebDriver against a running server: on load the prompt is\n`display:grid`, after Space it is `display:none`. And end to end through the\napp, the demo deck's YouTube clip now fetches (6,439,819 bytes) and serves as\nHTTP 200 video/mp4.\n\nCo-Authored-By: Claude Opus 5 \n","tree":"bee737dd8708163bc69df349225d57de434e0f73"},"mode":"0100644","name":"main.rs","size":5091},{"lastCommit":{"author":{"email":"larsvandartel73@gmail.com","name":"LarsvanDartel","when":"2026-08-18T22:39:28.000000+02:00"},"committer":{"email":"larsvandartel73@gmail.com","name":"LarsvanDartel","when":"2026-08-18T22:39:28.000000+02:00"},"hash":"05e349a4eb9a5095899c3c3298ab42f52099dbec","message":"Let the presenter enlarge a video to fill the screen\n\n`v`, or the console's `enlarge` button, blows the current slide's video up to\nfill the whole screen and puts it back again. The point is that it is driven\nfrom the presenter and applies to the display, so the audience sees the clip\nfull-size without anyone touching that screen.\n\nDeliberately CSS rather than the Fullscreen API: that needs a user gesture on\nthe document making the request, which a cast screen may never receive, so a\nremotely-triggered requestFullscreen() would simply be refused. Overlay geometry\ntherefore moved from inline styles into custom properties — an inline style beats\nany stylesheet rule, and would have pinned the box to its slide position.\n\nWhich video is enlarged lives on the server like every other piece of\npresentation state, and is offered to clients as a toggle, so the two views\ncannot end up disagreeing about it. Leaving the slide or stopping the clip\nrestores it, since neither a full screen of the next slide's video nor a frozen\nfirst frame is something to leave in front of an audience.\n\nOn the second request: a repeated URL was already downloaded once, which I\nconfirmed with a yt-dlp shim that logs its invocations — three uses of one URL,\none invocation, one cache file, one shared /media key. Added a regression test\nso it stays that way. The claim was only true by timing though: `resolve` wrote\nits pending entry *after* spawning the download, so a fast enough download could\nhave its `ready` overwritten and strand the UI on \"Fetching\". The key is now\nclaimed under the same lock that checks it, and a finished download is never\noverwritten.\n\nVerified by WebDriver against the display: a remote toggle takes it from 628x353\ninside the slide to the full 1366x682 viewport, toggles back, and advancing off\nthe slide drops it.\n\nCo-Authored-By: Claude Opus 5 \n","tree":"40bb7801b498d12b03c3f13119e5e4cbcb819df5"},"mode":"0100644","name":"media.rs","size":14231},{"lastCommit":{"author":{"email":"larsvandartel73@gmail.com","name":"LarsvanDartel","when":"2026-08-19T00:02:03.000000+02:00"},"committer":{"email":"larsvandartel73@gmail.com","name":"LarsvanDartel","when":"2026-08-19T00:02:03.000000+02:00"},"hash":"16df81ff03a50a7a566bcd3c13acd7e760316e84","message":"Make seeking work, and carry the playhead to the display\n\nThree of our own handlers were fighting the scrub bar.\n\nDragging it pauses the video, and a pause was reported as \"stop\" — which rewinds\nto the clip's start, undoing the seek the moment it happened. Pausing now says\n\"pause\" and holds the position; only leaving the slide or stopping the clip\nrewinds. Pressing play afterwards resumed from the start for the same reason, so\nplay carries a time too: no time means \"from the top\", which is what the\noverlay-step advance wants, and a time means resume.\n\nThe loop clamp then dragged anything past `end` straight back. It now sits out a\ndeliberate scrub outside the clip's window, and ignores a paused playhead\nentirely, since a paused position is wherever someone put it. The flag for that\nhas to be set on `seeking`, not `seeked`: a seek raises `timeupdate` in between,\nso doing it on `seeked` is too late and the clamp has already fired — which is\nexactly how the first attempt still failed at this.\n\nSeeking also never reached the display at all. The presenter's controls now\nbroadcast play, pause and seek with a position, so the cast screen lands on the\nsame frame rather than playing on obliviously. Echo suppression became\ntime-based, because `seeked` and `pause` arrive well after the call that caused\nthem and a plain flag was already clear by then.\n\nVerified with two real browser windows, presenter and display: a scrub to 5.5s\nsticks and the display follows to 5.5s; a scrub past the 7s `end` holds at 9.5s\ninstead of snapping to 2s; and pausing keeps 9.5s on both.\n\nCo-Authored-By: Claude Opus 5 \n","tree":"506de757fd2067c7f749aa9423b440048a892bcd"},"mode":"0100644","name":"protocol.rs","size":4224},{"lastCommit":{"author":{"email":"larsvandartel73@gmail.com","name":"LarsvanDartel","when":"2026-08-19T00:02:03.000000+02:00"},"committer":{"email":"larsvandartel73@gmail.com","name":"LarsvanDartel","when":"2026-08-19T00:02:03.000000+02:00"},"hash":"16df81ff03a50a7a566bcd3c13acd7e760316e84","message":"Make seeking work, and carry the playhead to the display\n\nThree of our own handlers were fighting the scrub bar.\n\nDragging it pauses the video, and a pause was reported as \"stop\" — which rewinds\nto the clip's start, undoing the seek the moment it happened. Pausing now says\n\"pause\" and holds the position; only leaving the slide or stopping the clip\nrewinds. Pressing play afterwards resumed from the start for the same reason, so\nplay carries a time too: no time means \"from the top\", which is what the\noverlay-step advance wants, and a time means resume.\n\nThe loop clamp then dragged anything past `end` straight back. It now sits out a\ndeliberate scrub outside the clip's window, and ignores a paused playhead\nentirely, since a paused position is wherever someone put it. The flag for that\nhas to be set on `seeking`, not `seeked`: a seek raises `timeupdate` in between,\nso doing it on `seeked` is too late and the clamp has already fired — which is\nexactly how the first attempt still failed at this.\n\nSeeking also never reached the display at all. The presenter's controls now\nbroadcast play, pause and seek with a position, so the cast screen lands on the\nsame frame rather than playing on obliviously. Echo suppression became\ntime-based, because `seeked` and `pause` arrive well after the call that caused\nthem and a plain flag was already clear by then.\n\nVerified with two real browser windows, presenter and display: a scrub to 5.5s\nsticks and the display follows to 5.5s; a scrub past the 7s `end` holds at 9.5s\ninstead of snapping to 2s; and pausing keeps 9.5s on both.\n\nCo-Authored-By: Claude Opus 5 \n","tree":"506de757fd2067c7f749aa9423b440048a892bcd"},"mode":"0100644","name":"server.rs","size":8160},{"lastCommit":{"author":{"email":"larsvandartel73@gmail.com","name":"LarsvanDartel","when":"2026-08-19T00:02:03.000000+02:00"},"committer":{"email":"larsvandartel73@gmail.com","name":"LarsvanDartel","when":"2026-08-19T00:02:03.000000+02:00"},"hash":"16df81ff03a50a7a566bcd3c13acd7e760316e84","message":"Make seeking work, and carry the playhead to the display\n\nThree of our own handlers were fighting the scrub bar.\n\nDragging it pauses the video, and a pause was reported as \"stop\" — which rewinds\nto the clip's start, undoing the seek the moment it happened. Pausing now says\n\"pause\" and holds the position; only leaving the slide or stopping the clip\nrewinds. Pressing play afterwards resumed from the start for the same reason, so\nplay carries a time too: no time means \"from the top\", which is what the\noverlay-step advance wants, and a time means resume.\n\nThe loop clamp then dragged anything past `end` straight back. It now sits out a\ndeliberate scrub outside the clip's window, and ignores a paused playhead\nentirely, since a paused position is wherever someone put it. The flag for that\nhas to be set on `seeking`, not `seeked`: a seek raises `timeupdate` in between,\nso doing it on `seeked` is too late and the clamp has already fired — which is\nexactly how the first attempt still failed at this.\n\nSeeking also never reached the display at all. The presenter's controls now\nbroadcast play, pause and seek with a position, so the cast screen lands on the\nsame frame rather than playing on obliviously. Echo suppression became\ntime-based, because `seeked` and `pause` arrive well after the call that caused\nthem and a plain flag was already clear by then.\n\nVerified with two real browser windows, presenter and display: a scrub to 5.5s\nsticks and the display follows to 5.5s; a scrub past the 7s `end` holds at 9.5s\ninstead of snapping to 2s; and pausing keeps 9.5s on both.\n\nCo-Authored-By: Claude Opus 5 \n","tree":"506de757fd2067c7f749aa9423b440048a892bcd"},"mode":"0100644","name":"state.rs","size":18733},{"lastCommit":{"author":{"email":"larsvandartel73@gmail.com","name":"LarsvanDartel","when":"2026-08-18T17:07:29.000000+02:00"},"committer":{"email":"larsvandartel73@gmail.com","name":"LarsvanDartel","when":"2026-08-18T17:07:29.000000+02:00"},"hash":"a1cb06bd26fb78feb2081f5014fed568082c4058","message":"Route poster images through the media store, and guard page fractions\n\nTwo fixes found reviewing the compile pipeline:\n\n- A `poster:` is a path relative to the deck, but it was handed to the\n browser verbatim, which would simply 404. It now registers with the media\n store like any other file and gets a real URL.\n- Point-to-fraction conversion is now a guarded helper. A degenerate page\n would previously produce infinity or NaN, which serializes to null and\n leaves the browser positioning an overlay at `NaN%`.\n\nCo-Authored-By: Claude Opus 5 \n","tree":"cc1c8c90cd1ce3abf12f2f80808843dc2b023aa4"},"mode":"0100644","name":"watch.rs","size":4020},{"lastCommit":{"author":{"email":"larsvandartel73@gmail.com","name":"LarsvanDartel","when":"2026-08-18T17:14:35.000000+02:00"},"committer":{"email":"larsvandartel73@gmail.com","name":"LarsvanDartel","when":"2026-08-18T17:14:35.000000+02:00"},"hash":"c00350795795445b3c63f6743aa6148a7dc37968","message":"Actually honour TYPST_PACKAGE_PATH, so the packaged build works offline\n\nThe flake set TYPST_PACKAGE_PATH at the vendored package set and the README\npromised hermetic operation, but nothing read it: TYPST_PACKAGE_PATH is a\ntypst-cli convention, and typst-kit's SystemPackages::new only looks in the\nstandard system directories. A packaged presentypst therefore still tried to\ndownload touying at first compile.\n\nPackage storage is now assembled explicitly from those two environment\nvariables, falling back to the system locations.\n\nAlso syncs the presenter view: driving a video with its native controls now\nroutes through the server, so the display follows instead of the two screens\ndrifting apart, and subslide buttons are numbered rather than repeating the\ninherited slide title.\n\nVerified in a network namespace with an empty cache: 12 slides compiled and\nserved with no network at all.\n\nCo-Authored-By: Claude Opus 5 \n","tree":"680f92e5503d60a6078f0e53fe5e802eef1203d8"},"mode":"0100644","name":"world.rs","size":6172}],"lastCommit":{"author":{"email":"larsvandartel73@gmail.com","name":"LarsvanDartel","when":"2026-08-19T00:02:03.000000+02:00"},"committer":{"email":"larsvandartel73@gmail.com","name":"LarsvanDartel","when":"2026-08-19T00:02:03.000000+02:00"},"hash":"16df81ff03a50a7a566bcd3c13acd7e760316e84","message":"Make seeking work, and carry the playhead to the display\n\nThree of our own handlers were fighting the scrub bar.\n\nDragging it pauses the video, and a pause was reported as \"stop\" — which rewinds\nto the clip's start, undoing the seek the moment it happened. Pausing now says\n\"pause\" and holds the position; only leaving the slide or stopping the clip\nrewinds. Pressing play afterwards resumed from the start for the same reason, so\nplay carries a time too: no time means \"from the top\", which is what the\noverlay-step advance wants, and a time means resume.\n\nThe loop clamp then dragged anything past `end` straight back. It now sits out a\ndeliberate scrub outside the clip's window, and ignores a paused playhead\nentirely, since a paused position is wherever someone put it. The flag for that\nhas to be set on `seeking`, not `seeked`: a seek raises `timeupdate` in between,\nso doing it on `seeked` is too late and the clamp has already fired — which is\nexactly how the first attempt still failed at this.\n\nSeeking also never reached the display at all. The presenter's controls now\nbroadcast play, pause and seek with a position, so the cast screen lands on the\nsame frame rather than playing on obliviously. Echo suppression became\ntime-based, because `seeked` and `pause` arrive well after the call that caused\nthem and a plain flag was already clear by then.\n\nVerified with two real browser windows, presenter and display: a scrub to 5.5s\nsticks and the display follows to 5.5s; a scrub past the 7s `end` holds at 9.5s\ninstead of snapping to 2s; and pausing keeps 9.5s on both.\n\nCo-Authored-By: Claude Opus 5 \n","tree":"506de757fd2067c7f749aa9423b440048a892bcd"}}