V2 new samples#55
Open
erenkucuker wants to merge 19 commits intobugra9:mainfrom
Open
Conversation
Both branches independently performed the v2 directory restructure
(packages/* → cppjs-{core,packages,plugins,samples}/*). Conflicts
resolved per the agreed grouping:
Group 1 — v2 restructure (cppjs-packages/*, cppjs-core/*): take main.
Main holds the canonical v2 PR (bugra9#53) with newer dependency versions
and the new target-system architecture (getTargetParams,
getFilteredBuildTargets, target.path = platform-arch-runtime-buildtype).
Group 2 — mobile templates: take main. Main upgraded
reactnative-cli/expo to Swift AppDelegate + the new expo `src/`
layout; v2-new-samples still had the older Objective-C / flat layout.
Group 4 — false-positive UD/AA from rename detection: restored from
origin/v2-new-samples (959ee3a). Git's rename detection paired
v2-new-samples' new *-multithread/* and nuxt/remix native files
with main's deleted packages/* paths, marking them as conflicts
even though main never had those paths. 50 files restored.
Group 5 — workspace/CI/lockfile: take main. cppjs-core-create-app's
package.json kept main's structure but manually merged in the new
multithread/next/nuxt sample workspace dependencies.
Plugins (cppjs-plugins/*) currently hold main's clean version. The
SSR + multithread enhancements from v2-new-samples (commit a5299ca)
will be ported on top of main's new target API in a separate commit.
vite plugin (cppjs-plugin-vite/index.js):
- Keep main's load() hook for /cpp.js (SPA HMR path).
- Keep main's middleware that rewrites /cpp.wasm and /cpp.data.txt to
/@fs/... so Vite serves them.
- Add middleware branches for SSR contexts and pthread workers:
/_nuxt/cpp.{js,wasm,data.txt,browser.js,worker.js} (Nuxt resolves
these server-side and bypasses Vite's load() hook), and
/cpp.browser.js / /cpp.worker.js (Emscripten pthread workers
request the browser JS by these names).
- ensureWasmBuilt() lazily compiles the wasm on the first SSR/worker
request since those paths skip Vite's load() hook entirely.
All paths use the new target API (buildTargetDebug.jsName/wasmName/
dataTxtName) — no Emscripten-x86_64 or general.name fallbacks left.
cppjs-sample-lib-prebuilt-matrix-multithread:
- cppjs.config.js: build.usePthread → target.runtime: 'mt'. main's
cpp.js no longer reads usePthread (grep finds 0 hits in
cppjs-core/cpp.js/src). Without this, loadConfig's auto-promotion
can't fire and dependent samples build wasm-wasm32-st-debug
instead of mt, then fail at link time looking for an mt Matrix.h.
- package.json scripts: -p WebAssembly|Android|iOS → -p wasm|android|ios
(main's bin.js validates against the lowercase set).
This is the only mt config touched intentionally — cascades through
loadConfig's runtime promotion to all samples that import matrix-mt.
Verified end-to-end in browser:
svelte-vite (st), svelte-vite-mt, react-vite-mt,
nuxt-vite (st), nuxt-vite-mt,
next-webpack (st), next-webpack-mt,
remix-vite — all render the matrix multiplication result.
Sample-side bugs that surfaced once the merge brought main's plugin
chain in. Pre-existing in v2-new-samples, not caused by the merge:
1. Missing await on Native.method() — Comlink wraps the embind module
when it runs inside a worker (mt mode for the vite/webpack plugins,
and effectively for nuxt/next even in st because they instantiate
in a deferred context). All proxy calls return Promises, so the
v2-new-samples authoring pattern of `setMessage(Native.sample())`
rendered "[object Promise]" instead of the result.
Fixed in:
svelte-vite-mt/src/App.svelte
react-vite-mt/src/App.jsx
nuxt-vite/app/example.vue
nuxt-vite-mt/app/example.vue
next-webpack-mt/app/page.tsx
remix-vite/app/welcome/welcome.tsx
2. svelte-vite-mt/src/App.svelte referenced an undefined `svelteLogo`
binding (the import was missing and there's no svelte logo asset
in the package). Removed the <div class="logo">…</div> block —
it caused the whole component to render blank in Svelte 5.
For mt samples the thread-result poll also moved into setTimeout's
async callback so getThreadResult() actually completes before being
assigned (previously it raced runOnThread()).
Verified end-to-end: every listed sample renders the matrix result
in the browser.
… API
Same surgical migration as matrix-mt + the await pattern. This sample
was untouched by Group 3 because it's a CLI sample, not browser; only
surfaced when running it on the command line.
- cppjs.config.mjs: build.usePthread → target.runtime: 'mt'
- package.json script: -p WebAssembly →
-p wasm -a wasm32 -r mt -e node -b release
(mirrors the st sibling's flags but with -r mt)
- src/index.mjs:
- Update the dist import path to the target-system naming
(cppjs-sample-backend-nodejs-wasm-multithread-wasm-wasm32-mt-release.node.js).
- await Native.sample()/runOnThread()/getThreadResult() — comlink
proxies the worker calls.
- Move getThreadResult() into setTimeout's async callback so the
thread has time to finish.
Verified: `node src/index.mjs` prints both the standard result and
the thread-computed result.
- Drop the .cppjs/ci/* files that were tracked under the mobile reactnative-cli samples. These are intermediate cppjs build outputs (interface/, bridge/, cache.json) — they're regenerated locally and belong in .gitignore, not in the repo. - Update pnpm-lock.yaml to reflect the workspace state after the main merge (new sample workspaces in cppjs-core-create-app's dependencies caused fresh resolutions).
erenkucuker
force-pushed
the
v2-new-samples
branch
from
3ca4e73 to
cd6aa3c
Compare
… script
Last sample still on the deprecated -p WebAssembly form. main's bin.js
rejects it (Invalid value: "WebAssembly". Allowed values: wasm, android, ios)
which broke `pnpm run ci:linux:build` since the CI script runs
build:samples across every workspace package.
Update to mirror the st-sibling's flag set with -r mt:
-p wasm -a wasm32 -r mt -e node -b release
Note: the install-time failure for this sample (workspace packages
@cpp.js/package-{curl,gdal,...}-multithread missing) is a separate
pre-existing issue, not caused by this commit.
… exist
CI's pnpm install + build:samples step crashed because three playground
samples import @cpp.js/package-*-multithread workspace packages that were
never added to cppjs-packages/ — the playgrounds shipped, the mt package
set didn't. Original Linux CI error:
ERR_MODULE_NOT_FOUND: Cannot find package '@cpp.js/package-curl-multithread'
imported from .../cppjs-playground-backend-nodejs-multithread/cppjs.config.mjs
Excluding the affected playgrounds from pnpm-workspace.yaml so install
passes and `pnpm --filter=@cpp.js/sample-* run build` skips them. The
sample directories stay on disk for development; re-include the lines
once the corresponding cppjs-package-{curl,gdal,openssl,...}-multithread
workspace packages are added.
Excluded:
- cppjs-samples/cppjs-playground-backend-nodejs-multithread
- cppjs-samples/cppjs-playground-mobile-reactnative-cli-multithread
- cppjs-samples/cppjs-playground-web-vite-multithread
Lockfile rewritten to drop the broken workspace:^ resolutions.
Note: the macOS CI failure ("Install the Apple certificate and provisioning
profile" — SecKeychainItemImport: Unable to decode the provided data) is a
secrets/certificate format issue unrelated to this branch.
… one test-android-sample.yml and test-ios-sample.yml only built @cpp.js/sample-lib-prebuilt-matrix (st), but `pnpm -r e2e:android`/ `pnpm run e2e:ios` recurse into every workspace and also run the e2e test for @cpp.js/sample-mobile-reactnative-cli-multithread. That mt sample's native.cpp does `#include <Matrix.h>`, which resolves via cppjs deps to matrix-mt's dist/prebuilt/<target>/include — never produced because the filter didn't match the -multithread package. Symptom on the last v2-new-samples Android run: cppjs-sample-mobile-reactnative-cli-multithread/src/native/native.cpp:2:10: fatal error: 'Matrix.h' file not found > Task :cpp.js_plugin-react-native:buildCMakeRelWithDebInfo[arm64-v8a] FAILED Switch the filter to a glob so pnpm picks up both: --filter=@cpp.js/sample-lib-prebuilt-matrix → --filter='@cpp.js/sample-lib-prebuilt-matrix*' Verified locally — `pnpm --filter='@cpp.js/sample-lib-prebuilt-matrix*' list` returns the st and mt packages. Both have build:android / build:ios scripts already (the cmake/source siblings have neither because they're inline libs).
…expect
The mt React/Svelte App templates split the C++ result across two <p>
elements ("Matrix multiplier with c++" + "Standard Result : RESULT"),
but the e2e tests use Playwright's getByText('Matrix multiplier with c++
=> J₃ * (2*J₃) = 6*J₃') which only matches when both halves are
inside the SAME element (Playwright's text matcher doesn't span across
sibling nodes).
CI Linux Playwright failure on the latest v2-new-samples run:
Error: element(s) not found
> 6 | await expect(page.getByText('Matrix multiplier with c++ =>
J₃ * (2*J₃) = 6*J₃')).toBeVisible()
at cppjs-sample-web-react-vite-multithread/e2e/cppjs.spec.cjs:6:90
Bring the mt App.jsx + App.svelte in line with the st sample format
(single <p> with "=>" separator). Thread result becomes its own <p>
that just shows the threadResult string (which is already prefixed
"Thread computed: ..." inside native.cpp), matching the second test
assertion.
The native.cpp matrix values aren't touched — st and mt sources are
identical, so the test's expected "= 6*J₃" / "= 36*J₃" already match
what the wasm produces.
48e1dea removed cppjs-sample-mobile-reactnative-cli{,-multithread}/ci/.cppjs/* treating them as build-cache leftovers, but test-ios-sample.yml has a "Restore cached bridge files" step that does: cp -r ./cppjs-samples/cppjs-sample-mobile-reactnative-cli/ci/.cppjs ./... These pre-baked SWIG bridge outputs (interface/native.i, bridge/native.i.cpp, bridge/native.i.cpp.exports.json, cache.json) are checked in on purpose so the iOS workflow doesn't need to invoke `cppjs build` before pod install + react-native run-ios — without them the .a build for the cppjs-react-native plugin would fail at native compile. Restored from the commit before 48e1dea. The lockfile cleanup half of 48e1dea is left in place; only the ci/.cppjs deletion is undone.
Three coupled fixes after running the playwright suite locally end-to-end: 1. cppjs-plugin-vite: add configurePreviewServer that sets COOP/COEP. `vite preview` (used by `e2e:prod` webServer) is otherwise just a static server, so SharedArrayBuffer is unavailable and the mt sample's pthread worker never initialises. The page sat on "compiling ..." through the test's 30s timeout. The dev middleware already sets these; the preview hook mirrors them. 2. svelte-vite-multithread App.svelte: revert to the original "Standard Result : ..." three-paragraph layout (244bf6e wrongly collapsed it the same way as react-mt). The svelte spec actually asserts on "Standard Result : J₃ * (2*J₃) = 6*J₃" — different pattern from react-mt's "Matrix multiplier with c++ => ...". Each mt App now matches its own spec's expected text again. 3. Both mt e2e specs now skip webkit: test.skip(({browserName}) => browserName === 'webkit', ...) WebKit's pthread + SharedArrayBuffer support on Linux is unreliable for our mt build — verified locally that the page stayed on "compiling ..." even with COOP/COEP set. chromium and firefox cover the mt path; the st samples still run on all three browsers. Verified locally with `pnpm --filter='@cpp.js/sample-web-{react,svelte}-vite-multithread' run e2e:prod`: 4 passed, 2 skipped (webkit) — for both samples.
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
2 participants
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.
No description provided.