fix: use version of htmlrewriter which does not make use of asyncify, which looks to have a potential memory leak under high load #2721
+44
−6
Conversation
This file contains 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
📊 Package size report 10%↑
Unchanged files
🤖 This report was automatically generated by pkg-size-action |
pieh
force-pushed
the
jake/htmlrewriter
branch
from
4a359c6 to
b76dfb1
Compare
pieh
commented
Dec 17, 2024
Comment on lines
+100
to
+108
| import { decode as _base64Decode } from './edge-runtime/vendor/deno.land/[email protected]/encoding/base64.ts'; | ||
| import { init as htmlRewriterInit } from './edge-runtime/vendor/deno.land/x/[email protected]/src/index.ts' | ||
| import {handleMiddleware} from './edge-runtime/middleware.ts'; | ||
| import handler from './server/${name}.js'; | ||
|
|
||
| await htmlRewriterInit({ module_or_path: _base64Decode(${JSON.stringify( | ||
| htmlRewriterWasm.toString('base64'), | ||
| )}).buffer }); | ||
|
|
There was a problem hiding this comment.
bundling wasm file as-is doesn't seem to work, so instead we inline it using same method we do for wasm modules used by users in middleware
opennextjs-netlify/src/build/functions/edge.ts
Lines 125 to 133 in 6b56128
and pass module to init function of htmlrewriter (argumentless init doesn't work after vendoring)
There was a problem hiding this comment.
I'm gonna have to just trust you on this one 😄
There was a problem hiding this comment.
https://gist.github.com/pieh/b205a685e518ef62afaab392e61e064f this is standalone script using this method that can be tested.
You can also test that passing crap to init first arg like:
await htmlRewriterInit({
module_or_path: 'wat',
})
would result in errors:
error: Uncaught (in promise) TypeError: Invalid URL: 'wat'
module_or_path = fetch(module_or_path);
^
at getSerialization (ext:deno_url/00_url.js:98:11)
at new URL (ext:deno_url/00_url.js:405:27)
at new Request (ext:deno_fetch/23_request.js:329:25)
at ext:deno_fetch/26_fetch.js:320:27
at new Promise (<anonymous>)
at fetch (ext:deno_fetch/26_fetch.js:316:18)
at __wbg_init (https://deno.land/x/[email protected]/pkg/htmlrewriter.js:1207:22)
at file:///Users/misiek/dev/pgs-next-runtime/edge-runtime/test.ts:11:7
so the argument has effect.
Commenting out init call also result in errors:
error: Uncaught (in promise) TypeError: Cannot read properties of undefined (reading 'htmlrewriter_new')
const ret = wasm.htmlrewriter_new(
^
at new HTMLRewriter (https://deno.land/x/[email protected]/pkg/htmlrewriter.js:781:22)
at Object.start (https://deno.land/x/[email protected]/src/index.ts:53:20)
at Module.invokeCallbackFunction (ext:deno_webidl/00_webidl.js:981:16)
at new TransformStream (ext:deno_web/06_streams.js:6214:16)
at HTMLRewriter.transform (https://deno.land/x/[email protected]/src/index.ts:48:29)
at rewriter (file:///Users/misiek/dev/pgs-next-runtime/edge-runtime/test.ts:39:6)
at file:///Users/misiek/dev/pgs-next-runtime/edge-runtime/test.ts:43:27
serhalp
previously approved these changes
Dec 17, 2024
Comment on lines
+101
to
+103
| // htmlrewriter contains wasm files and those don't currently work great with vendoring | ||
| // see https://github.com/denoland/deno/issues/14123 | ||
| // to workaround this we copy the wasm files manually |
There was a problem hiding this comment.
😢 I left a comment on that issue, as it seems like it should be reopened since Deno landed WASM support
There was a problem hiding this comment.
Deno landed it, but in 2.1 - https://deno.com/blog/v2.1#first-class-wasm-support , so support is not guaranteed :( otherwise we could migrate to imports in https://github.com/netlify/htmlrewriter/blob/edb477af3d08359c8f25edd8d301df69ffcc8e4b/pkg/htmlrewriter.js#L1184 to make use of it
Comment on lines
+100
to
+108
| import { decode as _base64Decode } from './edge-runtime/vendor/deno.land/[email protected]/encoding/base64.ts'; | ||
| import { init as htmlRewriterInit } from './edge-runtime/vendor/deno.land/x/[email protected]/src/index.ts' | ||
| import {handleMiddleware} from './edge-runtime/middleware.ts'; | ||
| import handler from './server/${name}.js'; | ||
|
|
||
| await htmlRewriterInit({ module_or_path: _base64Decode(${JSON.stringify( | ||
| htmlRewriterWasm.toString('base64'), | ||
| )}).buffer }); | ||
|
|
There was a problem hiding this comment.
I'm gonna have to just trust you on this one 😄
… which looks to have a potential memory leak under high load we noticed the memory issue with Netlify's CSP plugin which used the same htmlrewriter library. We've built a new htmlrewriter library which uses the latest version of lol-html and removes the ability to use async-handlers, which is what required asyncify to be included.
Co-authored-by: Philippe Serhal <[email protected]>
pieh
force-pushed
the
jake/htmlrewriter
branch
from
af28413 to
fdb11ed
Compare
youvalv
approved these changes
Dec 18, 2024
Comment on lines
+105
to
+107
| await htmlRewriterInit({ module_or_path: _base64Decode(${JSON.stringify( | ||
| htmlRewriterWasm.toString('base64'), | ||
| )}).buffer }); |
There was a problem hiding this comment.
we could use a uint8array instead of a base64 string here, it would save on memory usage by 33%
I did this in a similar package I made specifically for CSP support - I could do the same for the htmlrewriter package if we want:
await Deno.writeTextFile(
"./pkg/embedded-wasm.ts",
`export const wasmBinary = Uint8Array.from(${
JSON.stringify(
Array.from(
await Deno.readFile(
`./pkg/${wasmFile}`,
),
),
)
});`,
);https://github.com/netlify/csp_nonce_html_transformer/blob/main/scripts/build.ts#L13C3-L25
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Description
This was based on #2707 (new PR, primarily due to problems with running tests against forks).
Need to use bit of hacks and indirections due to:
deno vendornot playing well with wasm files ( Support vendoring modules which read "static" files, like.wasmdenoland/deno#14123), so I added manual pull for this file afterdeno vendorin our build scriptfetchattempt againstfile://url which fails:initfunction, but Netlify bundling also seems to skip.wasmfiles so I inline wasm module using same method we currently inline wasm modules used by user in middlewareDocumentation
Tests
Any test that make use of middleware already tests this implicitly
Relevant links (GitHub issues, etc.) or a picture of cute animal
Fixes https://linear.app/netlify/issue/FRB-1523/nextjs-runtime-uses-the-same-html-rewriter-as-the-csp-plugin-that