fix(web): explain missing WebGPU adapter instead of panicking#3
Open
proggeramlug wants to merge 1 commit into
Open
fix(web): explain missing WebGPU adapter instead of panicking#3proggeramlug wants to merge 1 commit into
proggeramlug wants to merge 1 commit into
Conversation
A user on macOS Sequoia hit a wall of Rust stack traces and a misleading "bloom engine init timed out". The cause: Chrome exposes `navigator.gpu` (on by default since 113) but returns no adapter — graphics acceleration off, a blocklisted GPU, or a VM/remote session. wgpu decides its backend in Instance::new purely on whether `navigator.gpu` is defined, so it commits to WebGPU, finds no adapter, and panics inside the wasm. Probe requestAdapter() before booting the engine and, when it comes back empty, stop with a message that names the likely causes and points at chrome://gpu. The WebGL2 route is plumbed here too (hiding `navigator.gpu` is what routes wgpu to wgpu-core), but it is gated off: Bloom's renderer eagerly builds compute pipelines, which WebGL2 cannot run, so an automatic fallback would only swap one crash for another. Flip WEBGL2_FALLBACK_SUPPORTED once the engine has a downlevel path; `?renderer=gl` forces the attempt meanwhile. Also add `build-web.sh --engine-src`, which builds bloom_web from the ../engine checkout. The default path cannot carry an unpublished engine fix: it compiles from the npm tarball, whose bundled Rust source does not compile (its web crate calls 3D model APIs the shared crate it ships with doesn't expose) — only its prebuilt pkg/ works.
📝 WalkthroughWalkthroughThe build script adds ChangesWeb runtime setup
Estimated code review effort: 3 (Moderate) | ~25 minutes Sequence Diagram(s)sequenceDiagram
participant BrowserURL
participant selectGpuBackend
participant NavigatorGPU
participant WebGL2
participant EngineBoot
participant LoadingUI
BrowserURL->>selectGpuBackend: read renderer=gl override
selectGpuBackend->>NavigatorGPU: requestAdapter()
NavigatorGPU-->>selectGpuBackend: WebGPU adapter or unavailable
selectGpuBackend->>WebGL2: check WebGL2 support when needed
WebGL2-->>selectGpuBackend: backend availability
selectGpuBackend-->>EngineBoot: return webgpu, webgl2, or none
EngineBoot->>LoadingUI: show backend or formatted error
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@build-web.sh`:
- Line 28: Update the --help branch in the build script’s argument handling to
print only the comment header, changing the sed range from lines 2–16 to the
range ending at line 13 so shell commands are excluded.
- Around line 42-47: Ensure the --engine-src path uses the sibling engine
checkout consistently: update the patch_engine_web.js and perry compile steps to
read from the same $BLOOM_WEB/../engine source used to rebuild pkg/, rather than
node_modules/@bloomengine/engine. Alternatively, use node_modules for all three
steps, but do not mix engine checkouts.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: ee488de1-4dd6-40f8-8c87-003184a636d3
📒 Files selected for processing (2)
build-web.shweb/bloom_ffi.js
| --engine-src) engine_src=true ;; | ||
| --serve) serve=true ;; | ||
| -h|--help) sed -n '2,15p' "$0"; exit 0 ;; | ||
| -h|--help) sed -n '2,16p' "$0"; exit 0 ;; |
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🟡 Minor | ⚡ Quick win
--help prints shell code, not just the header.
The comment header ends at line 13 (--serve); line 14 is blank and line 15 is set -euo pipefail. sed -n '2,16p' therefore dumps set -euo pipefail (and trailing blank lines) into the help output. Narrow the range to the comment block.
🩹 Proposed fix
- -h|--help) sed -n '2,16p' "$0"; exit 0 ;;
+ -h|--help) sed -n '2,13p' "$0"; exit 0 ;;📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| -h|--help) sed -n '2,16p' "$0"; exit 0 ;; | |
| -h|--help) sed -n '2,13p' "$0"; exit 0 ;; |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@build-web.sh` at line 28, Update the --help branch in the build script’s
argument handling to print only the comment header, changing the sed range from
lines 2–16 to the range ending at line 13 so shell commands are excluded.
Comment on lines
+42
to
+47
| if $engine_src; then | ||
| BLOOM_WEB="$JUMP_DIR/../engine/native/web" | ||
| else | ||
| BLOOM_WEB="$JUMP_DIR/node_modules/@bloomengine/engine/native/web" | ||
| [ -d "$BLOOM_WEB" ] || BLOOM_WEB="$JUMP_DIR/../engine/native/web" | ||
| fi |
There was a problem hiding this comment.
🗄️ Data Integrity & Integration | 🟠 Major | 🏗️ Heavy lift
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Inspect the patch target and how perry/cp consume the engine on the --engine-src path.
fd -a 'patch_engine_web.js' --exec sed -n '1,40p' {}
fd -a 'build-web.sh' --exec sed -n '60,80p' {}Repository: Bloom-Engine/jump
Length of output: 2586
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo '--- build-web.sh outline ---'
ast-grep outline build-web.sh --view expanded || true
echo '--- build-web.sh relevant slice ---'
nl -ba build-web.sh | sed -n '1,120p'
echo '--- search for engine-src and patch_engine_web usage ---'
rg -n --no-heading --line-number --context 2 'engine-src|patch_engine_web|perry compile|BLOOM_WEB' .Repository: Bloom-Engine/jump
Length of output: 287
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo '--- engine package.json web target ---'
fd -a 'package.json' node_modules/@bloomengine/engine 2>/dev/null | head -n 5 | xargs -r -I{} sh -c 'echo "FILE: {}"; nl -ba "{}" | sed -n "1,220p"'
echo '--- sibling engine checkout hint ---'
fd -a 'package.json' ../engine 2>/dev/null | head -n 5 | xargs -r -I{} sh -c 'echo "FILE: {}"; nl -ba "{}" | sed -n "1,220p"'Repository: Bloom-Engine/jump
Length of output: 194
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo '--- locate build-web.sh in repo ---'
git ls-files | rg '^build-web\.sh$|/build-web\.sh$'
echo '--- inspect node_modules engine manifest if present in repo ---'
fd -a 'package.json' 'node_modules/@bloomengine/engine' 2>/dev/null | head -n 1 | xargs -r -I{} sh -c 'echo "FILE: {}"; nl -ba "{}" | sed -n "1,160p"'Repository: Bloom-Engine/jump
Length of output: 268
🏁 Script executed:
#!/bin/bash
set -euo pipefail
printf '%s\n' '--- build-web.sh lines 1-140 ---'
sed -n '1,140p' build-web.sh | cat -n
printf '%s\n' '--- patch_engine_web.js ---'
sed -n '1,120p' store/tools/patch_engine_web.js | cat -nRepository: Bloom-Engine/jump
Length of output: 6665
--engine-src should keep the engine build on one checkout. This path rebuilds pkg/ from ../engine, but patch_engine_web.js and perry compile still read node_modules/@bloomengine/engine. That leaves dist/web/pkg and the game WASM built against different engine ABIs. Point the patch/compile steps at the sibling checkout too, or keep all three steps on node_modules.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@build-web.sh` around lines 42 - 47, Ensure the --engine-src path uses the
sibling engine checkout consistently: update the patch_engine_web.js and perry
compile steps to read from the same $BLOOM_WEB/../engine source used to rebuild
pkg/, rather than node_modules/@bloomengine/engine. Alternatively, use
node_modules for all three steps, but do not mix engine checkouts.
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
1 participant
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.
What
A user on macOS Sequoia reported the web build dying at boot. Console showed
No WebGPU/WebGL adapter found, a wall of Rust stack traces, a wasmunreachabletrap, and a misleadingbloom engine init timed out.It is not a Chrome feature flag. WebGPU is on by default in Chrome since 113. Having
navigator.gpuis not the same as having an adapter —requestAdapter()still returns null when graphics acceleration is off (chrome://settings/system), the GPU is blocklisted, or the browser is in a VM / remote session. wgpu picks its backend once, inInstance::new, purely on whethernavigator.gpuis defined (wgpu-29 api/instance.rs:74), so it commits to WebGPU, finds no adapter, and panics inside the wasm.This probes
requestAdapter()before booting the engine and, when it comes back empty, stops with a message that names the likely causes and points atchrome://gpu.Before → after on an adapter-less browser:
panicked at src/lib.rs:92+ wasmunreachable+init timed outThe WebGL2 fallback is plumbed but gated off
Hiding
navigator.gpufrom the wasm is what routes wgpu to wgpu-core → WebGL2, and that route now works as far as the renderer. It cannot render:Renderer::new_impleagerly builds 14 compute pipelines, storage buffers and a 64KB joint uniform, and WebGL2 (GLES 3.0) has no compute shaders or storage buffers at all. An automatic fallback would trade one crash for another, so it's behindWEBGL2_FALLBACK_SUPPORTED = false, with?renderer=glto force the attempt. Flip it when the engine gains a downlevel / 2D-only init path.Engine-side prerequisites (display handle, GL device limits, swapchain usage): Bloom-Engine/engine#87.
build-web.sh --engine-srcNeeded to ship any unpublished engine fix to the web. The default path builds
bloom_webfromnode_modules/@bloomengine/engine, and the published 0.4.16 tarball's Rust source does not compile — its web crate calls 3D model APIs (EngineState.models,Renderer::cache_model_if_static,staging::take_model) that thesharedcrate bundled in the same tarball doesn't expose. Only its prebuiltpkg/works, which is why--skip-bloombuilds succeed and full rebuilds fail. Worth fixing in the published package separately.Verification
Drove the deployed https://bloomengine.dev/jump/ in a real headless Chrome, both paths:
renderer: webgpu, engine initializes, game boots and plays into level 1 (screenshotted).requestAdapterstubbed to return null — the reporter's exact failure): no panics, no wasm trap, no bogus timeout; the explanation renders.Summary by CodeRabbit
New Features
renderer=glURL option.Bug Fixes
Documentation