docs(guides): illustrate installation, settings and save transfer #264
No reviewers
Labels
No labels
area/accessibility
area/audio
area/input
area/install
area/menu
area/renderer
area/saves
area/upscaling
beanstalk: approved @ 1697603
beanstalk: approved @ 2c5a62e
beanstalk: approved @ 4fd82b7
beanstalk: approved @ 64d6d09
beanstalk: approved @ 6f91fae
beanstalk: approved @ 70a78f9
beanstalk: approved @ 8d361dc
beanstalk: approved @ 9d8356e
beanstalk: approved @ abf2227
beanstalk: approved @ d82f202
beanstalk: approved @ e5f1e40
beanstalk: approved @ ee8ab5f
ci: approved
good first issue
gpu/amd
gpu/igpu
gpu/intel
gpu/nvidia
help wanted
kind
bug
kind
chore
kind
docs
kind
feature
kind
performance
kind
question
needs measurements
platform/linux
platform/macos
platform/steam-deck
platform/windows
preview
priority
critical
priority
high
priority
low
regression
status
blocked
status
confirmed
status
duplicate
status
invalid
status
needs-info
status
wontfix
testing wanted
ci: approved
No milestone
No project
No assignees
4 participants
Notifications
Due date
No due date set.
Blocks
#611 fix(website): improve mobile browsing and launcher downloads
MagicBeanDev/Longshot
Reference
MagicBeanDev/Longshot!264
Loading…
Reference in a new issue
No description provided.
Delete branch "bibiburl/Longshot:docs/guide-screenshots"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Type: docs · Risk: low · Tested on: Linux, game-free checks and source/image review · Needs: successful site build and rendered recheck of the latest update
Why
Players need to recognize the controls in the installation, settings and save-transfer guides. These guides share seven maintained UI screenshots, keep the actions and replacement warnings in readable text, and flag screenshots for review when their source changes.
What changed
Seven UI-only examples illustrate the real maintained views with synthetic state. Graphics was captured from Longshot main
b2911d33; six Launcher pictures were captured from nightly-52-b1d9f7a9, source b1d9f7a9. The follow-up retains those PNGs and their original capture provenance.2b1d38c6is integrated.source-review-e9bed614.mdrecords why the four changed menu inputs permit retaining the labelled Graphics example, and adds the newthreaded_rendering.hdependency. All 59 reviewed inputs also match2b1d38c6. The off-screen row changes the scrollbar proportion; no pixel-equivalence claim is made.Measurements
Verdict: the combined previous head and main reproduced four screenshot-source failures and one restart-list test failure. After review and correction, all seven image/source checks, 18 Python screenshot tests and 76 website tests pass. The portable backup path now matches Launcher source. No gameplay or performance improvement is claimed.
Flags and settings
No game/runtime defaults change relative to main. These existing variables are pinned only by the isolated capture helpers:
HOS_DISPLAY_MODES1280x720in the menu fixtureLONGSHOT_LAUNCHER_HOMELONGSHOT_LAUNCHER_OFFLINE1in the Launcher fixtureLONGSHOT_LAUNCHER_REDUCED_MOTION1in the Launcher fixtureThe displayed game-build selection, hardware/tools-ready results and Custom graphics values are illustrative. Launcher and game-build versions are recorded separately. No capture helper was rerun for this follow-up. The optional C++ capture fixture is Linux-only tooling, not a host/runtime interface change.
Testing
Tested:
Integrated baseline: main
2b1d38c682. Updated PR:a0a93f11fe. Linux, Node 26.13.1; npm dependencies restored from the unchanged pinned lockfile.scripts/check_guide_screenshots.py --review-sources --launcher-source ...: all seven assets and recorded UI/fixture inputs pass, using an isolated source archive of exact Launcher b1d9f7a9b19fa99b91ec3e8cd24455d4703d6d25. The review covers 59 menu inputs. Without the external checkout the checker correctly reports six Launcher source checks as NOT CHECKED.tests/test_guide_screenshots.py: 14 passed.tests/test_launcher_capture.py: 4 passed. All seven existing PNGs inspected; image bytes, captured-source records, dates and fixture sources remain unchanged.node --test tests/*.test.ts: 76/76 passed on the final combined tree. The restart test first reproduced the missing Threaded rendering expectation; the fix preserves all existing coverage/restart assertions. Guide tests render the canonical Markdown and verify local image links and numbered-step placement.LauncherService.SaveEnv,LauncherHome.PortableSaveBackupsandLauncherLayout.PortableHomeat Launcher b1d9f7a9, and the rc.3 portable backup override at 028ad66d.python3 -I -B tests/run_all.py, outside the sandbox: 499 passed, 12 failed, 69 skipped in 62 seconds. Every failure was rerun on clean main at the integrated revision; all 12 fail there too:test_lift_cleanup_error(missing_lift_storefixture),test_make_game_cache(unexpected compiler subprocess),test_appimage_player_build(prebuilt-renderer path assertion),test_bake_portability(missing zstandard),test_cpm_discovery(stub missingfind),test_crash_dialog(clang treats trailing-x noneas an error),test_env_on(switch-source check),test_linux_preflight(unexpected profile-probe process),test_readback_late(range-log source expectation),test_vk_pgo(profile-note expectation),test_vram_budget_row(fixture missingOptions::preset), andtest_walk_key(missinghos_oot_autoload_inputstub). These unrelated failures remain outside this documentation change. Missing tool/header and platform checks are reported as skips, not passes.npm run buildcurrently fails before page generation with Astro'sMISSING_EXPORT: loadDevToolbarAppsfromastro:toolbar:internal. It reproduces on unchanged maine9bed614using the same pinned dependency tree, including afternpm ci; the website dependency/config files are unchanged by subsequent main2b1d38c6. No successful final build or link-count result is claimed.git diff --checkpasses. No runtime/renderer/third-party differences from the integrated main snapshot, no changed PNGs, no private assets, saves or installed-game settings.Historical evidence, not a fresh result: head
e4627be1built 15 pages and checked 803 links/assets; its full suite was 503 passed, 6 failed, 69 skipped. The original captures were built and inspected at that head. The maintainer review of e4627be1 reports Chromium checks at nine widths in both themes, image links, keyboard/touch navigation and zoom. It also identifies pre-existing home-page overflow and a search-result rendering limitation. Those results are attributed to that reviewer and revision.Not tested: a successful website build or browser rendering of this latest update, Firefox/Safari, real mobile devices, screen readers, Windows execution, a fresh menu/Launcher capture, live Vulkan equivalence, actual save transfer or gameplay. The local Astro error prevents using a fresh site build for browser checks here. Source/image review does not establish pixel equivalence.
Must be tested: Linux and Windows, any GPU, by hand: once the baseline Astro failure is resolved or a working CI build is available, run
npm run buildandnpm run check, expecting a successful build and passing link/asset checks. Open/install/,/settings/and/saves/transfer/at 320, 390, 800 and 1120px in both themes with keyboard and zoom, expecting the installed/portable backup paths to remain readable, numbered steps and captions to stay aligned, full-size image links to work, and no guide overflow. T-R2H9 remains the existing browser-testing request; the maintainer's prior-head results are linked above, not claimed as a result for this update.Screenshots
Unchanged UI-only examples captured 2026-11-14. Launcher nightly-52-b1d9f7a9 on Linux; Graphics from development source
b2911d33. Values are not recommendations or new-install defaults. Links deliberately retain the original image commit.Risks
What could break: future source changes require review of the image, dependency set and adjacent text. Hashes detect recorded-input drift but cannot establish visual usability or discover new dependencies automatically. The nightly and rc.3 differ on Windows saves; the guide distinguishes them. A retained historical image is not a pixel-identical rendering of current main.
Turning it off: revert the documentation/capture/website changes. Original capture provenance stays in Git.
source-review-e9bed614.mdis the current retention decision; the earliercf605c85note concerns a replaced image.Compatibility: no runtime, save-format, options, cache or renderer ABI change relative to integrated main. No game build, launch, runtime experiment, actual save transfer, official release or production deployment was performed for this follow-up.
The Longshot screenshot workflow is included. Required branch protection and the separate Launcher companion check remain maintainer setup; the README does not claim they are enabled.
Issues
See #104 (comment) for the guide-image request and the review threads on this PR. See #584 for the activity-title wording; its underlying game issue is not fixed here.
Checklist
Everyone
type(scope): description) of the status line's type.main. Integrated main through2b1d38c6by merge; the exact validation baseline is named above.python -I -B tests/run_all.pypasses (any failure that also fails onmainis named above).third_party/untouched (a new folder with aPROVENANCE.mdinstead).When it applies
Assisted-by:/Co-Authored-By:trailers.mainand passing here.tests/that reads changed source updated in this PR, not weakened.Added Saves and manual-import screenshots in
9253f97, plus exact rc.3 button labels and confirmation steps. The release-tag source confirms Windows saves remain per build while Linux saves are shared; the captions make that distinction explicit. All three PNG asset copies pass a byte-for-byte check. Settings fixture, source drift check and full website validation remain in progress.Added the developer maintenance plan: local menu/UI/font changes will trigger screenshot source review; Launcher UI/release authors get an explicit cross-repository review command and companion-docs responsibility. The planned Launcher hook is documented as a follow-up integration, not claimed as installed. Capture provenance remains separate from later human review records.
Implemented the maintenance plan in
6d8a424: an offline PNG/source-fingerprint checker, 12 passing regression tests and a read-only workflow definition for every PR/main push. Changed source identifies each affected screenshot; missing external Launcher source reports NOT CHECKED. The README assigns UI author/reviewer/docs maintainer responsibilities and separates the planned Launcher counterpart and required branch-protection configuration from what this PR actually adds. Settings integration and final validation remain WIP.Maintenance follow-ups require fingerprints for both UI capture source files (13 tests now pass) and specify that the planned Launcher-side review job must check out and verify the exact candidate/final-release commit SHA. Missing Launcher review and branch-protection setup remain explicitly separate from this PR’s local guard.
Added the real-menu UI-only settings fixture and fourth shared guide image. All image/source checks and 13 checker tests pass; the site builds 14 pages and passes 792-link/asset validation plus all 8 website test files. The developer plan now specifies every-PR source review, a separately installed Launcher companion job, exact-release checks, and required branch protection. Those cross-repository setup steps are not claimed as enabled. Original capture provenance is preserved; current menu comparison changes only the version label. WIP remains for browser-rendered review because the browser access host is unavailable.
Two independent agents reviewed screenshot placement: a UI reviewer and a documentation designer. Their documented plan puts the import chooser inside step 3 before confirmation, the welcome capture directly beside Build, and the settings layout beside navigation. It recommends full-size links and prioritizes a synthetic-state import-confirmation capture. The PR description records exact anchors, reader questions, proposed versus existing assets and acceptance criteria. This was a review pass: guide placement and image files are unchanged; rendered mobile/theme review remains outstanding.
Added the three clarity captures in
112bd10: Manage saves navigation, game-file selection, and import confirmation. Each sits beside its action with a full-size link. The real Launcher views render isolated synthetic state; import remains a dry-run preview. Updated the capture provenance and developer-maintained review plan. All 7 image/source checks, 14 guard tests, and the website build/check pass (14 pages, 802 links/assets, 8 Node test files). WIP remains for rendered browser review and maintainer-side required-check setup.WIP: docs(guides): illustrate installation, settings and save transferto docs(guides): illustrate installation, settings and save transferRead-only review of head
112bd103against main514591a3. I read the changed sources, looked at all seven pictures and read the existing check logs. I ran no tests, builds or capture helpers.What's good
website/src/repo-docs/assets.tsusesfileURLToPath, the shared mapping reaches both the asset-copy hook and the link rewriting, and the new test checks a destination with spaces in it.Must change
scripts/screenshots/capture_launcher.pystages onlyApp/,Core/,Directory.Build.propsandDirectory.Packages.props. The Launcher'sDirectory.Build.targetsat v1.0.0-rc.3 is left out, and that file is the opt-out: it declares an emptyAvaloniaStatstarget, with the comment that Avalonia.BuildServices otherwise reports each build to Avalonia's servers.DOTNET_CLI_TELEMETRY_OPTOUT, which the helper does set, is a different switch. So as written the capture build is the one Launcher build that would report. Stage that file too and include it in the fingerprint, or reproduce its effect explicitly.README.mdconflicts. Main has since added AMD FSR 4 and limits the fallback sentence to FSR 3.1; keep that when resolving.Should change
NuGet.Configas well. It restricts restores to nuget.org. Locked package versions do not carry the feed configuration, so the staged restore currently uses whatever sources the machine has.docs/saves.mdlines 7 and 51 show the capture run's temporary folder name, including in the confirmation's backup path. They are throwaway fixture paths, not anyone's saves, but a guide reads better with neutral ones. Recapture with a fixed fixture path rather than repainting the labels.docs/images/guide/README.mddescribes the check that now runs on every pull request. A link from CONTRIBUTING would help: even a change that only touches a recorded renderer header will ask for a screenshot review.Small
docs/saves.mdline 37 changes Frizti to Fritzi. The Launcher's own list at rc.3 says Frizti (Fritchii), so the old spelling matches what a player sees.Readiness
The description says work in progress while the title does not; the description check on this head lists what is still missing (https://git.magicbean.dev/MagicBeanDev/Longshot/actions/runs/2448/jobs/0). The README and reference move may need coordinating with #236.
Not checked
Nothing was built or run here: not the capture helpers, not the website build, not a real save transfer, and nothing on Windows. I did not compare the pictures against a live Launcher.
112bd103c154ffa6b5e054ffa6b5e0f035bbd54fAddressed the review in the rebased head
f035bbd(main0815d98):Validation: all 7 image/source checks and 18 Python tests pass; both fixture builds succeed; the final site builds 15 pages and validates 852 links/assets. Main/PR/main/PR website tests give 71/72 versus 74/75, with the same existing Audio-output assertion failing on both. Full game-free run: 492 passed, 1 failed, 62 skipped; the sole HoldToggle fixture failure is reproduced on exact main. No game or real save transfer ran.
Independent follow-up review found no blocking issue in the fixes. The description now matches ready-for-review status and records current evidence instead of the old WIP text. Browser rendering review remains outstanding because the browser host is unavailable; no test request or successful visual inspection is claimed. Ready status is retained as requested. Please re-review the addressed findings.
Posted browser-testing request T-R2H9 for the Install, Settings and Save Transfer guides: desktop and 360–390px layouts, both themes, zoom, keyboard full-size links and search. The system says it is waiting for a maintainer to add the
testing wantedlabel before anyone can claim it. Please add that label to enable testing. The description now links the request; rendered review remains pending, and no successful browser test is claimed.Updated in
d7c62a3: simplified the three player guides, removed the separate full-size screenshot labels, and aligned save-transfer steps, images and captions with the article while preserving native numbering and screenshot links.Exact PR-source validation: website build succeeds (15 pages); 851 local links/assets/anchors pass; all 75 website tests pass. The stale Audio output test now mutates the current restart annotation and still requires the same mismatch error. All seven screenshot integrity/source checks and 14 guard tests pass.
Browser access remains unavailable. The PR description explicitly asks the independent reviewer to check the complete transfer page, responsive layouts, both themes, keyboard links and search. No deployment or merge was performed.
Read-only re-review of
d7c62a39d4against maincf605c85. I read the changes since112bd103, the updated description and replies, the existing check logs and the three replaced pictures. I ran nothing from the pull request.Earlier points
scripts/screenshots/capture_launcher.py:14-21,53-61requires, fingerprints and copiesDirectory.Build.targets. The three source-rendered Launcher records include it. The missing-file tests cover its omission.README.md:407retains AMD FSR 4 and the FSR 3.1 fallback. README is no longer in this PR's diff, and Forgejo reports it mergeable.scripts/screenshots/capture_launcher.py:93-95passes the stagedNuGet.Configexplicitly to both locked restores; it is also required and fingerprinted.capture_launcher.py:25-32,98-99reserves a new directory before capture and refuses an existing one.LauncherCapture.cs:18-19accepts that empty directory. The tests check that existing data survives refusal and that failed capture cleans up its own directory.CONTRIBUTING.md:277-280links the maintenance procedure and explains that implementation-only changes can require review.docs/saves.md:39matches the released chooser.What's good
fileURLToPath, descriptive alternatives and space-containing-path test are retained. Removing the separate labels has not removed the website's full-size links; the test now checks the rendered links.Must change
docs/getting-started.md:17-19now jumps from downloading to pressing Build. The simplification removed both the operating-system instructions and the warning that Linux/SteamOS users must make the AppImage executable. A reader with a non-executable download cannot reach that welcome screen. Restore the short platform links and permission step before Build; the instructions remain inREADME.md:155-161.docs/images/guide/manifest.json:273-274still records the menu files from the0815d98baseline. Current main has changed those files and other recorded headers, including the Intel shader-compilation row and cache/log paths. The green job checked the branch head, not those combined sources; merging unchanged would make the new main-push check fail. Review those changes, then update the review record with a reason for retaining or replacing the picture. Do not just replace the hashes.Should change
Keep a short example-state caption beside
docs/getting-started.md:34anddocs/graphics-and-controls.md:19. The former shows invented hardware/tools-ready results; the latter is a development-menu example with Custom values. Removing the longer captions also removed those qualifications from the player pages. The manifest alone does not tell their readers.Readiness and not checked
T-R2H9 still awaits eligibility and a browser result: #264 (comment). The new page-scoped list and figure styles need that check. The reported 75 website tests and carried whole-suite result are the author's evidence, not runs made here; website and full-suite CI were skipped. Nothing was built, captured or run on Windows here, and no real save transfer or live-game comparison was made. The reference move remains open at #236.
Website preview: https://pr-264-longshot-website.hckr-5e0.workers.dev
This commit (
b905befe78) on its own: https://a952ccab-longshot-website.hckr-5e0.workers.devUpdated in
b905befe78.Addressed source-review item #2 against main
cf605c85, now integrated into the branch. The combined tree first reproduced all nine stale-source failures. I reviewed those nine diffs and the existing Graphics image, then recorded per-file retention reasons indocs/images/guide/source-review-cf605c85.mdandreview.notes. Intel shader compiles is a Windows/Intel-only Debug row; cache/log destinations and pipeline-list error text are not pictured; driver/recorder/sparse-memory headers do not change the fixture drawing path. Original image, capture date, version banner and captured-source provenance remain unchanged. This is a source-and-image review, not a new capture.Also made the author's requested mobile-header cleanup: guide pages retain Starlight's hamburger without the duplicate Menu disclosure; home/splash pages retain their links behind one hamburger. Removed the empty second header row. Generated home/install/settings/transfer HTML has exactly one mobile navigation control per page and retains the main destinations.
Validation on the combined tree: all seven image/source checks pass including exact Launcher rc.3; 14 screenshot tests + 4 capture-policy tests pass; website builds 15 pages, passes 799 link/asset checks and all 75 tests. Full game-free suite: 490 pass, 3 fail, 68 skip; all three failures reproduce on unchanged
cf605c85(missing zstandard, stale module_ids source marker, missing HoldToggle in pad fixture). No runtime source changes.Fresh UI fixture rendering could not start without Vulkan headers. Browser host remains unavailable, so rendered mobile interaction/layout review is still pending; no new screenshot or live-game verification is claimed. Review item #1 is intentionally unchanged at the author's request, not claimed resolved. The PR description now contains this current evidence.
Fixed the reported hamburger-style mismatch in
31a18758. The home disclosure now uses the same Starlight bars/close icons as the guide sidebar. Shared CSS controls size, centering, background/foreground colours, shadow and open state in both themes.Validation: 15-page build, 799 link/asset checks and all 75 website tests pass. Generated home/install/settings/transfer markup has identical menu/close SVGs and one mobile navigation control per page. Rendered-browser verification remains unavailable. No runtime changes; the previous source-review record is untouched.
The preview label needs adding again for this new commit to refresh the PR preview.
Re-review of
31a18758e3against main0caf06184. I read the replies, the source-review record and the later mobile-header changes. I ran the website tests and build, the site check and the screenshot checks on this head, including the pinned Launcher source.Earlier points
docs/images/guide/manifest.json:244-274now records the integratedcf605c85sources. The nine explanations indocs/images/guide/source-review-cf605c85.md:19-27match the diffs: the Intel control is a Windows/Intel-only Debug row; the path and pipeline-list changes are not pictured; the driver, recorder and sparse-memory additions do not run in the menu drawing example. None of these recorded sources changed betweencf605c85and current main0caf06184. The original picture and capture record are preserved, rather than presented as a new capture. The full source check passes here.docs/getting-started.md:13-19still goes straight from downloading to pressing Build. The reply explicitly leaves this unchanged. A Linux or SteamOS reader whose AppImage is not executable still cannot reach that screen. Restore the short permission step and platform-instructions link; the instructions remain inREADME.md:163-169.docs/getting-started.md:32-36anddocs/graphics-and-controls.md:17-21still omit the short explanation that the hardware/tools-ready display is illustrative and the settings picture is a development example, not recommended defaults. The source-review note explains this, but readers of those player pages do not see it.b905befe78bb, before the latest hamburger styling. It is not a browser result for31a18758e3, and no completed T-R2H9 result is posted.What's good
The earlier fixes remain in place:
capture_launcher.py:14-21,53-61preserves the statistics opt-out;93-95supplies the explicit NuGet configuration;25-32,98-99refuses an existing example directory. The neutral pictures are unchanged.CONTRIBUTING.md:277-280still links screenshot maintenance,docs/saves.md:39says Frizti, andREADME.md:445preserves the FSR 4/FSR 3.1 wording.The save-replacement warning, numbered import steps, full-size image links, shared asset mapping and space-containing-path test are retained. The Audio output test still asserts that its input changed and that the restart mismatch is rejected.
New changes
website/src/components/Header.astro:6,23-30renders its own disclosure only on pages without Starlight's sidebar. Starlight's page frame renders its toggle only when that sidebar exists, so the two branches do not duplicate the control. The new shared styles use the same bars/close icons and appearance rules. I found no new source-level blocker there, but opening, closing, focus, narrow layouts and both themes still need the requested browser review.Checks run
028ad66d.These used the pinned dependencies on Linux, under the shared measuring lock. The build retains the existing duplicate-404-route warning. Nothing was deployed or posted. I did not rebuild the screenshot programs, run the whole game-free suite, make a real save transfer, test Windows or perform a rendered-browser check.
The source-review blocker is closed. The missing Linux launch step is why I am keeping the change request open.
Resolved the Linux launch prerequisite in
b792d805: Getting started now tells Linux/SteamOS users to open the AppImage’s Properties and enable Allow executing file as program before opening the launcher and pressing Build. Links to the full Windows and Linux setup instructions are restored. No other source files changed.Validation on this branch: 15 pages built; 799 link/asset/anchor checks passed; all 75 website tests passed; generated install HTML puts the permission step before Build;
git diff --checkpassed. No fresh browser review or full runtime test run for this text-only edit.Separate review finding: a temporary merge of the previous PR head
31a18758with main414f56e2is conflict-free but fails the screenshot-source guard on seven inputs. Main’s new Custom output size row is absent from the screenshot, so its review/capture needs refreshing. The combined website build also fails on that setting, but that parser failure reproduces on main itself (main tests 64/72 pass; combined 66/75 pass, all failures reach the same mismatch). These issues remain open; the Linux fix does not claim to resolve them. This is implementation follow-up/self-review, not independent approval.Refreshed all seven guide screenshots in
e4627be1, integrating mainb2911d33. Graphics now shows Custom output size; all six Launcher views are captured from nightly-52-b1d9f7a9 (b1d9f7a9) using synthetic state and the real maintained UI. Captions identify nightly/example state, and save guidance distinguishes its shared Windows saves from rc.3. Linux executable permission remains restored.The website importer now handles registered text fields and their restart requirement. A neutral-copy rule also covers the item name in issue #584, which blocked the publication check. All source/fixture checks, 18 Python screenshot tests, 76 website tests and 803 link/asset checks pass; all 15 pages build. All seven PNGs were inspected. No game, real save transfer or deployment occurred.
Full game-free suite: 503 pass / 6 fail / 69 skip. All six failures reproduce on main
b2911d33and are named in the updated description. Main advanced to546dc212during this work: its 58 recorded menu inputs still match; those unrelated commits are not integrated. Browser host is unavailable, so rendered responsive/keyboard/theme review remains pending. The PR description and image links now point to these fresh captures.Re-review of
e4627be1eeagainst main82886e89. I read the new replies (20632, 20810), the description, the changes since31a18758e3and the whole branch against its merge base. I ran the PR's own website and screenshot checks on this head, built the site from it and tried it in a real browser at nine widths in both themes. The hosted preview is stale (it predates this head's Linux step, captions and pictures, and shows the older hamburgers), so I used it only as a comparison.Verdict: close to approve-quality. Two mechanical items must change before merge; the guides, captions, pictures and navigation held up in the browser.
Earlier points
docs/getting-started.md:17-21puts the permission step and the setup links before Build, and the built Install page has them in that order. The two README anchors it links to exist.getting-started.md:33,44,graphics-and-controls.md:23, the transfer guide). In the browser they are 16 px, contrast 7.97:1 or better in both themes, aligned with the picture and never clipped.Must change
main" is not ticked. The text beside it names main546dc212; main is now82886e89. Merge main, update that sentence and tick the box.b2911d33:gpu/vk/vk_api.h,vk_menu.cpp,vk_menu.handvk_recorder.h(manifest entries). The cause is main's new Threaded rendering row, which sits under "GPU and memory" after "GPU device", below the rows in the Graphics picture, so the picture very likely stays accurate (I read the diff; I did not re-render it). The guide workflow also runs on pushes to main (guide-screenshots.yml:7-8), so an unchanged merge turns main red. In the same tree the website tests are 75 of 76:graphics-settings.test.ts:96-97lists the restart rows exactly and now also findsThreaded rendering. The importer, the build (15 pages) and the site check (804 links) pass on the merged tree. Please merge main, review and record the four inputs, and add the row to that list.Should change (small)
nightly-52-b1d9f7a9, while the Install page's download links redirected tov1.0.0-rc.3when I checked (2026-11-14, 22:21 UTC). The Launcher repository has taggedv1.0.0-rc.4on the pictured commitb1d9f7a9, but no release exists yet. The captions are honest andsaves.mdseparates rc.3 from the pictured build (I checked its save-location and shared-save claims against the Launcher source atb1d9f7a9). When rc.4 ships, the captions ingetting-started.md:33andsaves.mdshould say rc.4.docs/images/guide/source-review-cf605c85.md. It says to retainsettings-graphics.png, which this head regenerated, and nothing in the tree refers to it.Header.astro:23-30puts the<details>hamburger before the search and theme controls in the page source, but it is the right-most control on a phone. Tab order at 390 px: skip link, logo, hamburger, search, theme. Guide pages end with the toggle, as the eye does. Moving the<details>after.toolsfixes it.docs/saves.md:22has lost its instruction. "Pick your adventure." now asks for nothing. Main's step said to note the slot and roughly when you last played, since the launcher lists every save with its slots and times.Header.astro:91-94). Links stay readable; it is cosmetic.Not from this PR, but you will meet it
inset: 10% -5% 12%on the hero halo (LongshotHero.astro:76), a file this PR does not touch; the live site shows the same numbers. Your "Must be tested" line expects no overflow, which the home page does not meet, but the guides do. Clipping the hero's horizontal overflow would fix it, here or later.Question for maintainers
launcher-home.pngcontains the game's full name in the Launcher's own text ("turns your own copy of ... into a PC game"), while the site neutralises that name in prose (public-copy.ts). The earlier rc.3 picture had the same text. Is that acceptable inside a screenshot?Browser results
Built from this head, Chromium 155, scripted. Heights 568, 740, 844, 915, 1024, 768, 800, 900 and 1080 for the widths in order. All 15 pages at all nine widths in both themes (270 combinations) were measured; the four pages below were also captured full page. Light and dark gave the same numbers at every width.
Checks run on this head
All from the PR's own workflows: the seven image and source checks with the exact Launcher source
b1d9f7a9pass (and without it, as CI runs it, with six NOT CHECKED notices); 14 + 4 = 18 Python tests pass; website tests 76 of 76; build 15 pages; site check 803 links, assets and anchors;git diff --checkclean. These match your description. Node 26.13.Not tested
Firefox, Safari, real phones, a screen reader, print, Windows, the result list of the search box, the Graphics picture against current main (source read only), a hosted build of this head, the full game-free suite, the screenshot and Launcher capture programs, any game or real save transfer.
Once main is merged, the four inputs are recorded, the restart list is updated and the description check is green, I would approve. The small items above are optional.
Pictures (from this head's build, Chromium, 390 px wide)
@mooki — the two required follow-ups from your latest review are addressed in
a0a93f11fe. Could you re-check this revision?2b1d38c6and updated the description/checklist to name the tested baseline. The local description checker passes.docs/images/guide/source-review-e9bed614.md, and addedthreaded_rendering.hto the dependency record. All 59 recorded menu inputs match that integrated baseline. The existing image retains its original capture provenance; the note explicitly acknowledges the changed scrollbar proportion and makes no pixel-equivalence claim.Validation: all seven image/source checks, 18 Python screenshot/capture tests and 76 website tests pass. Full game-free suite: 499 passed, 12 failed, 69 skipped; all 12 failures reproduce on clean main at the integrated baseline and are named in the description. The site build currently fails with Astro's
loadDevToolbarAppsmissing export, also reproduced on unchanged main with the same pinned dependencies. No fresh browser result or successful final site build is claimed.Your optional UI/copy suggestions are not claimed resolved by this update. The older source-review note remains as history, with the new note explicitly superseding its retention decision. Please re-check the required source-review/restart-list/description fixes and the backup-path correction; a rendered check of the updated table remains pending when a working site build is available.
bibiburl referenced this pull request2026-11-15 01:55:10 +00:00
Reviewed
a0a93f11fe, including the guide text, capture/check helpers, website integration, all seven PNGs and the previous review threads. The prior required fixes are present: Linux AppImage permission precedes Build, example captions remain visible, Threaded rendering is in the exact restart list, and the source-review note preserves the original capture provenance.One merge blocker remains against current main
d55e3bf74c: an isolated conflict-free merge fails the new screenshot guard on four tracked headers. Details are inline. This is new drift since the author's integrated baseline, not a failure of the earlier review work. Please review/record these changes before merge, then rerun the guard on the combined tree.Independent Windows verification (Python 3.14.7, Node 24.22.0):
Browser inspection of this exact build, in the in-app Chromium browser: Install, Settings and Transfer at 320/390/800/1120 CSS pixels, each in light and dark themes (24 combinations). All seven guide images load and stay within their content width, with no document horizontal overflow. The updated backup table scrolls internally on narrow widths, is keyboard focusable, and ArrowRight exposes the backup column. At 1120px both path columns are visible together. The emulator chooser/confirmation remain under steps 3/4, and replacement warnings and example captions remain text. I inspected the mobile guide menu opening by Enter and closing by Escape, retaining focus on Menu. Source images were also inspected at full size.
The previous small copy/navigation suggestions remain optional and are not repeated as blockers. This review does not include a fresh UI capture, live Launcher/save transfer, Linux execution, Firefox/Safari, real phones, screen readers, a new zoom sweep, the complete game-free suite, or browser testing of the temporary merged tree. No game was launched and no deployment was performed.
@ -0,0 +903,4 @@"gpu/vk/oot_cheat_items.h": "8500b50a9a779967d889741fe80e24dae1031ba7127aaed449207f9b8d43648c","gpu/vk/pipeline_share.h": "e14e2cf56f4164ae00f4b7e2b7a647b9d75131ccb69b349c22d5ab75ee6f8487","gpu/vk/threaded_rendering.h": "82c8f5a5b7cc463f980f6517814150698eb1394114d029116d44ee7b63108f8c","gpu/vk/upscaler_rows.h": "c4f3a8c7f80e03ede367aad2c71fc74bec0e4337d98d7a440d3ce314de34a1a8",[P2] Refresh the retained image review for the current merge target. An isolated merge of this head with main
d55e3bf7is conflict-free, butscripts/check_guide_screenshots.py --review-sourcesfails on upscaler_rows.h, vk_framegen.h, vk_fsr3fi.h and vk_temporal.h. Those tracked headers changed after the integrated2b1d38c6baseline (PsychoV35 admission/default policy and captured-present HDR metadata/signatures). Since this PR installs the guard on main pushes, merging unchanged makes that check red. Review the four diffs against the retained image/fixture and record the retention decision and reviewed hashes, preserving captured_source and PNG provenance; recapture only if that review finds a visible mismatch. Then rerun the guard on the combined tree.