refactor(onerror): inline error page template as string constant#5868
refactor(onerror): inline error page template as string constant#5868killagu merged 5 commits intoeggjs:nextfrom
Conversation
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
📝 WalkthroughWalkthroughThe onerror plugin now embeds and exports a built-in HTML error-page template, exposes it as a public subpath, defaults Changes
Sequence Diagram(s)sequenceDiagram
participant App as App Init
participant Config as Config
participant FS as FileSystem
participant Template as Built-in Template
participant Client as HTTP Client
App->>Config: read onerror.templatePath
alt templatePath truthy
App->>FS: read template file
FS-->>App: template content
App->>Client: serve rendered error page (file template)
else templatePath empty
App->>Template: import ONERROR_PAGE_TEMPLATE
Template-->>App: template content
App->>Client: serve rendered error page (built-in)
end
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes Possibly related PRs
Suggested reviewers
Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request inlines the error page template into a new TypeScript file, onerror_page.ts, and updates the onerror plugin to use this template by default when templatePath is not configured. This change simplifies the plugin by removing the need to read an external HTML file from the file system in the default case. Feedback was provided to correct minor string formatting issues within the inlined JavaScript template, specifically regarding line number alignment and the display of method locations.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
🧹 Nitpick comments (1)
plugins/onerror/src/lib/onerror_page.ts (1)
710-722: Consider limiting sensitive information exposure in error pages.The error page displays
appInfo.baseDirandappInfo.configin non-production environments. While this is useful for debugging, be mindful that this could expose sensitive configuration values if not properly filtered upstream.Based on learnings: "Never expose sensitive configuration values in logs or error messages" - the filtering should be handled in
ErrorViewto ensure sensitive config values are redacted before rendering.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@plugins/onerror/src/lib/onerror_page.ts` around lines 710 - 722, The error page currently renders raw appInfo.baseDir and appInfo.config which can leak secrets; update ErrorView to sanitize appInfo before rendering by implementing a redaction step (e.g., sanitizeAppInfo or redactConfig) that removes or masks sensitive keys/values from appInfo.config and optionally obfuscates baseDir, then ensure the template rendering in onerror_page.ts uses the sanitizedAppInfo fields (sanitizedAppInfo.config and sanitizedAppInfo.baseDir) instead of the raw appInfo; keep the redaction centralized in ErrorView so all error rendering paths use the sanitized object.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Nitpick comments:
In `@plugins/onerror/src/lib/onerror_page.ts`:
- Around line 710-722: The error page currently renders raw appInfo.baseDir and
appInfo.config which can leak secrets; update ErrorView to sanitize appInfo
before rendering by implementing a redaction step (e.g., sanitizeAppInfo or
redactConfig) that removes or masks sensitive keys/values from appInfo.config
and optionally obfuscates baseDir, then ensure the template rendering in
onerror_page.ts uses the sanitizedAppInfo fields (sanitizedAppInfo.config and
sanitizedAppInfo.baseDir) instead of the raw appInfo; keep the redaction
centralized in ErrorView so all error rendering paths use the sanitized object.
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: e249aee1-450d-499f-86e3-1f74048ca748
📒 Files selected for processing (4)
plugins/onerror/package.jsonplugins/onerror/src/app.tsplugins/onerror/src/config/config.default.tsplugins/onerror/src/lib/onerror_page.ts
There was a problem hiding this comment.
Pull request overview
Note
Copilot was unable to run its full agentic suite in this review.
Refactors the onerror plugin to inline the error page HTML template as a TypeScript string constant so it can be statically bundled (e.g., by Turbopack), removing the runtime import.meta.dirname + readFileSync dependency.
Changes:
- Adds an inlined
ONERROR_PAGE_TEMPLATEconstant containing the full error page template. - Changes the default
templatePathconfig to''(use built-in template when empty). - Updates the boot logic to use the inlined template unless a custom
templatePathis provided, and exports the new module entry.
Reviewed changes
Copilot reviewed 4 out of 4 changed files in this pull request and generated 4 comments.
| File | Description |
|---|---|
| plugins/onerror/src/lib/onerror_page.ts | Adds the inlined HTML template constant used as the default error page. |
| plugins/onerror/src/config/config.default.ts | Removes default filesystem template path; documents and defaults templatePath to empty string. |
| plugins/onerror/src/app.ts | Switches to inlined template when templatePath is not set; retains file override behavior. |
| plugins/onerror/package.json | Exposes the new lib/onerror_page entry for source and dist builds. |
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## next #5868 +/- ##
==========================================
+ Coverage 85.48% 85.52% +0.03%
==========================================
Files 660 661 +1
Lines 18828 18861 +33
Branches 3646 3658 +12
==========================================
+ Hits 16096 16131 +35
+ Misses 2361 2359 -2
Partials 371 371 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
killagu
force-pushed
the
split/02-onerror-inline-template
branch
from
f99e0e7 to
4254124
Compare
|
已按 review comments 把这轮 onerror 修改收完：默认模板改成按需加载，删除了遗留 mustache 文件，补了 frame 相关空值 guard，同时修了 |
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (4)
plugins/onerror/test/onerror.test.ts (1)
82-112: Consider also covering thedefaultConfigIgnoreListfallback.The new test always supplies
dump.ignore, so thecatchbranch ingetConfigIgnoreList()(which falls back todefaultConfigIgnoreList) is never exercised. A second tiny case constructing anErrorViewwhoseapp.configlacks adumpproperty would lock in the default-list behavior (e.g.password,keys,/secret/i).it('should fall back to default ignore list when dump.ignore is unavailable', () => { const view = new ErrorView( { request: {}, app: { config: { baseDir: '/tmp/app', password: 'p', keys: 'k', token_secret: 's', normal: 'v' } } } as any, new Error('test') as any, '', ); const { config } = view.serializeAppInfo(); assert.match(config, /<Redacted>/); assert(!config.includes("'p'")); assert(!config.includes("'k'")); assert(!config.includes("'s'")); assert.match(config, /normal: 'v'/); });🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@plugins/onerror/test/onerror.test.ts` around lines 82 - 112, The test suite currently only covers the path where app.config.dump.ignore is provided, so add a second test that constructs an ErrorView without a dump property to exercise getConfigIgnoreList()'s fallback to defaultConfigIgnoreList; instantiate ErrorView (same pattern as the existing test) with app.config containing sensitive keys like password/keys/token_secret and a visible key, call serializeAppInfo() and assert the serialized config redacts sensitive values (contains "<Redacted>" and does not include the raw sensitive strings) while still including the visible key—this ensures defaultConfigIgnoreList behavior is validated.plugins/onerror/src/lib/error_view.ts (1)
340-361: Special object types degrade to{}in the rendered error page.The
typeof === 'object'branch thenObject.keys(...)path iterates enumerable own keys only, soDate,RegExp,URL,Map,Set,Buffer, and class instances all get serialized as plain objects (typically empty{}), unlike egg core'sconvertValueinpackages/egg/src/lib/core/utils.tswhich preserves them viatoString()/ primitive passthrough. Since this is the fallback used whendumpConfigToObjectis unavailable, consider mirroring that handling for consistent inspect output:redactConfig(value: unknown, ignoreList: (string | RegExp)[], ...): unknown { if (!value || typeof value !== 'object') { return value; } + if (value instanceof RegExp || value instanceof URL || typeof value === 'symbol') { + return (value as { toString(): string }).toString(); + } + if (Object.getPrototypeOf(value) !== Object.prototype && !Array.isArray(value)) { + // Date, Map, Set, Buffer, class instances → defer to util.inspect later + return value; + } ... }🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@plugins/onerror/src/lib/error_view.ts` around lines 340 - 361, The redactConfig function currently treats any non-null object by iterating Object.keys and thus serializes special types (Date, RegExp, URL, Map, Set, Buffer, class instances) to empty objects; update redactConfig (and where it calls shouldRedactConfigKey) to detect these special types before the Object.keys path and return sensible primitive/string/array representations instead (e.g., Date/RegExp/URL -> toString(), Buffer -> Buffer.toString('utf8') or base64, Map -> Array.from(map.entries()), Set -> Array.from(set), and for other class instances fall back to instance.toString() or a shallow property-redaction that preserves their string form). Keep the circular-detection via seen, still recurse for arrays, and only use the Object.keys branch for plain objects so rendered error pages match egg core's convertValue behavior.plugins/onerror/src/lib/onerror_page.ts (2)
1320-1322: Minor:$method/$filecan benulland will render as the string"null".
getAttributereturnsnullwhen the attribute is absent, so if a frame is ever rendered withoutdata-method/data-file(e.g., a future change to the Mustache template, or a frame produced by a stack-trace parser variant), line 1321 producesinnerHTML = "null"and line 1322 produces"null lineColumn". Today Mustache always emits these, so it's not currently broken — flagging as a small defensive nit while you're touching this code.Suggested defensive coercion
- $('#frame-file').innerHTML = $file; - $('#frame-method').innerHTML = $method + ' ' + $lineColumn; + $('#frame-file').innerHTML = $file || ''; + $('#frame-method').innerHTML = ($method || '') + ' ' + ($lineColumn || '');🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@plugins/onerror/src/lib/onerror_page.ts` around lines 1320 - 1322, Guard against null attribute values when assigning to innerHTML for element IDs '#code-drop', '#frame-file', and '#frame-method' by coercing $file and $method to safe strings (e.g., use an empty string or a default like '(unknown)') before assignment; ensure $lineColumn is appended only if present so that $('#frame-method').innerHTML doesn't become "null lineColumn" — locate where $file, $method, and $lineColumn are read/used and replace direct assignments with a small safe-coercion step for each variable.
2-1345: Maintainability note on the inlined ~1.3k-line template.Inlining the template as a TS string constant achieves the stated goal (statically bundleable, no
import.meta.dirname+readFileSync), but it does fold a sizable chunk of vendored Prism.js source into the TS module — every future Prism upgrade will require re-doubling backslashes by hand. Two low-effort options to keep the ergonomics without giving up bundleability:
- Keep the template in
onerror_page.htmland use a build-time transform (tsup/tscplugin,unplugin-raw, or a tiny prebuild script that emitsonerror_page.tswithJSON.stringify(html)). This gives you HTML syntax highlighting, no manual escaping, and the same runtime behavior.- If you want to stay in TS, consider
String.rawfor the regex-heavy script blocks so the doubled backslashes aren't required — though this conflicts with the surrounding template literal, so it would need to be a concatenated piece.Not blocking — purely a future-maintenance suggestion. Consider whether option (1) fits the wider 19-PR refactor.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@plugins/onerror/src/lib/onerror_page.ts` around lines 2 - 1345, The ONERROR_PAGE_TEMPLATE constant embeds a ~1.3k-line HTML/Prism script which is painful to maintain (escaped backslashes etc.); fix by moving the literal into a raw HTML asset and generating the TS export at build-time: create an onerror_page.html with the full template and add a build step (unplugin-raw, tsup/tsc plugin, or a small prebuild script) that reads that file and emits a TS module exporting ONERROR_PAGE_TEMPLATE (e.g., via JSON.stringify or by writing the literal into the generated file), or alternatively split the large regex-heavy script into a separate String.raw segment and concatenate it into ONERROR_PAGE_TEMPLATE to avoid double-escaping; update references that import/use ONERROR_PAGE_TEMPLATE to consume the generated module.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@plugins/onerror/src/lib/error_view.ts`:
- Around line 340-361: redactConfig currently uses a WeakSet named seen as a
global visited set so shared non-circular objects are labeled "[Circular]";
change the logic to track only the current ancestor chain by adding the object
to seen before recursing into its children and removing it (delete from the
WeakSet) after finishing that subtree so siblings can visit the same object
without being treated as circular; ensure this add/remove pattern is applied for
both object and array branches inside redactConfig and keep using
shouldRedactConfigKey and redactedValue as the redaction checks.
---
Nitpick comments:
In `@plugins/onerror/src/lib/error_view.ts`:
- Around line 340-361: The redactConfig function currently treats any non-null
object by iterating Object.keys and thus serializes special types (Date, RegExp,
URL, Map, Set, Buffer, class instances) to empty objects; update redactConfig
(and where it calls shouldRedactConfigKey) to detect these special types before
the Object.keys path and return sensible primitive/string/array representations
instead (e.g., Date/RegExp/URL -> toString(), Buffer -> Buffer.toString('utf8')
or base64, Map -> Array.from(map.entries()), Set -> Array.from(set), and for
other class instances fall back to instance.toString() or a shallow
property-redaction that preserves their string form). Keep the
circular-detection via seen, still recurse for arrays, and only use the
Object.keys branch for plain objects so rendered error pages match egg core's
convertValue behavior.
In `@plugins/onerror/src/lib/onerror_page.ts`:
- Around line 1320-1322: Guard against null attribute values when assigning to
innerHTML for element IDs '#code-drop', '#frame-file', and '#frame-method' by
coercing $file and $method to safe strings (e.g., use an empty string or a
default like '(unknown)') before assignment; ensure $lineColumn is appended only
if present so that $('#frame-method').innerHTML doesn't become "null lineColumn"
— locate where $file, $method, and $lineColumn are read/used and replace direct
assignments with a small safe-coercion step for each variable.
- Around line 2-1345: The ONERROR_PAGE_TEMPLATE constant embeds a ~1.3k-line
HTML/Prism script which is painful to maintain (escaped backslashes etc.); fix
by moving the literal into a raw HTML asset and generating the TS export at
build-time: create an onerror_page.html with the full template and add a build
step (unplugin-raw, tsup/tsc plugin, or a small prebuild script) that reads that
file and emits a TS module exporting ONERROR_PAGE_TEMPLATE (e.g., via
JSON.stringify or by writing the literal into the generated file), or
alternatively split the large regex-heavy script into a separate String.raw
segment and concatenate it into ONERROR_PAGE_TEMPLATE to avoid double-escaping;
update references that import/use ONERROR_PAGE_TEMPLATE to consume the generated
module.
In `@plugins/onerror/test/onerror.test.ts`:
- Around line 82-112: The test suite currently only covers the path where
app.config.dump.ignore is provided, so add a second test that constructs an
ErrorView without a dump property to exercise getConfigIgnoreList()'s fallback
to defaultConfigIgnoreList; instantiate ErrorView (same pattern as the existing
test) with app.config containing sensitive keys like password/keys/token_secret
and a visible key, call serializeAppInfo() and assert the serialized config
redacts sensitive values (contains "<Redacted>" and does not include the raw
sensitive strings) while still including the visible key—this ensures
defaultConfigIgnoreList behavior is validated.
🪄 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
Run ID: 39b00bad-75ec-4cf5-b606-6bf89f1f5fb1
📒 Files selected for processing (4)
plugins/onerror/src/app.tsplugins/onerror/src/lib/error_view.tsplugins/onerror/src/lib/onerror_page.tsplugins/onerror/test/onerror.test.ts
🚧 Files skipped from review as they are similar to previous changes (1)
- plugins/onerror/src/app.ts
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 6 out of 6 changed files in this pull request and generated 1 comment.
Comments suppressed due to low confidence (1)
plugins/onerror/src/lib/onerror_page.ts:2
src/lib/onerror_page.mustache.htmlno longer exists (template is now inlined), butplugins/onerror/tsdown.config.tsstill tries to copy that file intodist/. This will likely break the package build/publish step; update the build config to remove that copy step (or point it at the new source) now that the template lives inonerror_page.ts.
There was a problem hiding this comment.
🧹 Nitpick comments (1)
plugins/onerror/src/lib/onerror_page.ts (1)
1321-1321: Minor: leading space when method is empty.
getAttribute('data-method')returns the rendered Mustache value, which is the empty string whenframe.methodis undefined. In that case the label becomes" 12:34"with a stray leading space. Non-blocking, but trivial to tighten:✏️ Optional polish
- $('#frame-method').innerHTML = $method + ' ' + $lineColumn; + $('#frame-method').innerHTML = $method ? $method + ' ' + $lineColumn : $lineColumn;🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@plugins/onerror/src/lib/onerror_page.ts` at line 1321, The label can get a stray leading space when getAttribute('data-method') returns an empty string; update the assignment to $('#frame-method').innerHTML to only include the separating space when $method is non-empty (e.g. use [$method, $lineColumn].filter(Boolean).join(' ') or conditional expression like $method ? $method + ' ' + $lineColumn : $lineColumn) so the element shows "12:34" instead of " 12:34" when frame.method is undefined.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Nitpick comments:
In `@plugins/onerror/src/lib/onerror_page.ts`:
- Line 1321: The label can get a stray leading space when
getAttribute('data-method') returns an empty string; update the assignment to
$('#frame-method').innerHTML to only include the separating space when $method
is non-empty (e.g. use [$method, $lineColumn].filter(Boolean).join(' ') or
conditional expression like $method ? $method + ' ' + $lineColumn : $lineColumn)
so the element shows "12:34" instead of " 12:34" when frame.method is undefined.
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 62887ebf-2e74-4112-a842-cb64474193d2
📒 Files selected for processing (1)
plugins/onerror/src/lib/onerror_page.ts
There was a problem hiding this comment.
♻️ Duplicate comments (1)
plugins/onerror/src/lib/error_view.ts (1)
340-365:⚠️ Potential issue | 🟡 MinorShared (non-circular) references are still marked as
[Circular].The
seenWeakSet is added to but never removed after a subtree finishes processing, so any object referenced from two distinct keys (e.g. a shared logger / env / options sub-object onapp.config) will be replaced with'[Circular]'on its second occurrence even though there is no real cycle. Track only the current ancestor chain (add before recursing, delete in afinallyafter) so siblings can revisit the same object safely.🐛 Proposed fix
redactConfig( value: unknown, ignoreList: (string | RegExp)[], - seen: WeakSet<object> = new WeakSet<object>(), + ancestors: WeakSet<object> = new WeakSet<object>(), ): unknown { if (!value || typeof value !== 'object') { return value; } - if (seen.has(value)) { + if (ancestors.has(value)) { return '[Circular]'; } - seen.add(value); - - if (Array.isArray(value)) { - return value.map((item) => this.redactConfig(item, ignoreList, seen)); - } - - const result: Record<string, unknown> = {}; - for (const key of Object.keys(value)) { - result[key] = this.shouldRedactConfigKey(key, ignoreList) - ? redactedValue - : this.redactConfig((value as Record<string, unknown>)[key], ignoreList, seen); - } - return result; + ancestors.add(value); + try { + if (Array.isArray(value)) { + return value.map((item) => this.redactConfig(item, ignoreList, ancestors)); + } + const result: Record<string, unknown> = {}; + for (const key of Object.keys(value)) { + result[key] = this.shouldRedactConfigKey(key, ignoreList) + ? redactedValue + : this.redactConfig((value as Record<string, unknown>)[key], ignoreList, ancestors); + } + return result; + } finally { + ancestors.delete(value); + } }🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@plugins/onerror/src/lib/error_view.ts` around lines 340 - 365, redactConfig currently marks shared (non-circular) objects as "[Circular]" because it adds entries to the seen WeakSet and never removes them; change redactConfig so seen only tracks the current ancestor chain: when you detect a non-primitive object/array, check seen.has(value) as now, then add value to seen before recursing into its children and remove it from seen in a finally block after processing (for both Array.isArray branch and object property loop) so shared sibling references are not misclassified as circular; keep references to redactConfig, shouldRedactConfigKey, and redactedValue when implementing this change.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Duplicate comments:
In `@plugins/onerror/src/lib/error_view.ts`:
- Around line 340-365: redactConfig currently marks shared (non-circular)
objects as "[Circular]" because it adds entries to the seen WeakSet and never
removes them; change redactConfig so seen only tracks the current ancestor
chain: when you detect a non-primitive object/array, check seen.has(value) as
now, then add value to seen before recursing into its children and remove it
from seen in a finally block after processing (for both Array.isArray branch and
object property loop) so shared sibling references are not misclassified as
circular; keep references to redactConfig, shouldRedactConfigKey, and
redactedValue when implementing this change.
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 8572f32d-1116-4650-bbd0-944eb32dcf83
📒 Files selected for processing (2)
plugins/onerror/src/lib/error_view.tsplugins/onerror/tsdown.config.ts
Inline the mustache error page template as a string constant in src/lib/onerror_page.ts so plugins can be statically bundled by turbopack. Previously the template was read from disk via `import.meta.dirname + readFileSync`, which breaks when modules are embedded in a single bundled chunk. `templatePath` default is now `''`; if set, the custom template file is still loaded via fs, otherwise the inlined constant is used. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
killagu
force-pushed
the
split/02-onerror-inline-template
branch
from
58e915b to
1e98ce1
Compare
| if (value instanceof Date || value instanceof RegExp || value instanceof URL) { | ||
| return value.toString(); | ||
| } | ||
|
|
||
| if (Buffer.isBuffer(value)) { |
There was a problem hiding this comment.
In the fallback redaction path, Set/Map values (e.g. Egg’s default config.dump.ignore is a Set) will currently fall through to the generic object handling and end up serialized as {} because Object.keys(new Set()) is empty. Consider adding explicit handling for Set/Map (e.g. convert to arrays/entries and recursively redact) before the generic object branch so AppInfo stays informative.
Summary
Inlines the 1336-line error page HTML template into
plugins/onerror/src/lib/onerror_page.tsas a string constant. Removes the runtimereadFileSync+import.meta.dirnamelookup for the default template, sets thetemplatePathconfig default to an empty string, and updatesapp.tsto lazy-load the built-in template only when no custom template path is configured.The fallback app-info serialization path now redacts sensitive config values when
app.dumpConfigToObject()is unavailable. This is an observable security improvement for rendered error output; user-suppliedtemplatePathbehavior remains unchanged.Why
This is batch 1, part of a 19-PR split of #5863 (the egg-bundler PR). #5863 is kept open as a tracking reference. This PR is independent of the other batch-1 PRs.
Turbopack (and any static bundler) cannot follow
import.meta.dirname + readFileSyncto a template file, so the file would be missing in a bundled deployment. Inlining the HTML as a string constant makes the plugin statically bundleable.Test plan
pnpm exec vitest run plugins/onerror/test/onerror.test.ts- 39/39 passedpnpm --filter=@eggjs/onerror typecheckpnpm run buildpnpm exec oxlint --type-aware --type-check plugins/onerror/src/app.ts plugins/onerror/src/lib/error_view.ts plugins/onerror/src/lib/onerror_page.ts plugins/onerror/test/onerror.test.ts plugins/onerror/tsdown.config.ts- 0 errors, 1 existing warning onapp.close()Stack context
Other batch-1 PRs (independent, can land in any order):
feat(utils): add setBundleModuleLoader runtime hookrefactor(development): inline loader trace template as string constantrefactor(watcher): use direct class imports for event sourcesGenerated with Claude Code
Summary by CodeRabbit
New Features
Bug Fixes & Improvements
Tests
Chores