fix: not match \n when injecting esbuild helpers

[ddosakura]

Description

fixes #8412

Additional context

A plugin like this could be used to workaround it.

But, I found that the key to the bug was not rollupOptions.output.banner, but \n, the following code does not match \n correctly:

const INJECT_HELPERS_UMD_RE =
  /(.*)(\(function\([^)]*?\){.+amd.+function\([^)]*?\){"use strict";)(.*)/

---

What is the purpose of this pull request?

• Bug fix
• New Feature
• Documentation update
• Other

Before submitting the PR, please make sure you do the following

• Read the Contributing Guidelines.
• Read the Pull Request Guidelines and follow the Commit Convention.
• Check that there isn't already a PR that solves the problem the same way to avoid creating a duplicate.
• Provide a description in this PR that addresses what the PR is solving, or reference the issue that it solves (e.g. fixes #123).
• Ideally, include relevant tests that fail without this PR but pass with it.

[patak-dev]

Oh, good catch. This isn't a problem when not using a banner since the result is minified.

[sapphi-red]

The code change looks ok to me.
But the output might be slightly different from the expected output.
The comment will be inside the wrapper function instead of outside.

iife

const MyLib=function(){"use strict";var u=Object.defineProperty;var e=Object.getOwnPropertySymbols;var i=Object.prototype.hasOwnProperty,s=Object.prototype.propertyIsEnumerable;var c=(n,t,o)=>t in n?u(n,t,{enumerable:!0,configurable:!0,writable:!0,value:o}):n[t]=o,r=(n,t)=>{for(var o in t||(t={}))i.call(t,o)&&c(n,o,t[o]);if(e)for(var o of e(t))s.call(t,o)&&c(n,o,t[o]);return n};/*!
MayLib
*/function n(t){console.log(r({},"foo")),document.querySelector(t).textContent="It works"}return n}();

umd

(function(n,e){typeof exports=="object"&&typeof module<"u"?module.exports=e():typeof define=="function"&&define.amd?define(e):(n=typeof globalThis<"u"?globalThis:n||self,n.MyLib=e())})(this,function(){"use strict";var u=Object.defineProperty;var t=Object.getOwnPropertySymbols;var d=Object.prototype.hasOwnProperty,s=Object.prototype.propertyIsEnumerable;var i=(n,e,o)=>e in n?u(n,e,{enumerable:!0,configurable:!0,writable:!0,value:o}):n[e]=o,f=(n,e)=>{for(var o in e||(e={}))d.call(e,o)&&i(n,o,e[o]);if(t)for(var o of t(e))s.call(e,o)&&i(n,o,e[o]);return n};/*!
MayLib
*/function n(e){console.log(f({},"foo")),document.querySelector(e).textContent="It works"}return n});

Also there are still some broken niche case. #8412 (comment)

Since it was producing invalid js before, I think it is still good to merge it.

[patak-dev]

Maybe we could detect the last */ or a line that doesn't start with // in the regex? It would be good to avoid fixing this halfway.

[sapphi-red]

I think it is difficult to cover all usecases by the current way using regex.

without rollupOptions.output.banner + build.minify: false

var MyLibrary = (function (exports) {
  'use strict';

  exports.foo = { ...'foo' }

  return exports;

})({});

with rollupOptions.output.banner + build.minify: false

var someValueInjected = Object.assign({}, { value: 'by rollup.banner' });
var MyLibrary = (function (exports) {
  'use strict';

  exports.foo = { ...'foo' }

  return exports;

})({});

with rollupOptions.output.banner + build.minify: false

var someValueInjected = Object.assign({}, { value: 'by rollup.banner' });
var u=Object.defineProperty;var i=Object.getOwnPropertySymbols;var a=Object.prototype.hasOwnProperty,c=Object.prototype.propertyIsEnumerable;var n=(r,o,f)=>o in r?u(r,o,{enumerable:!0,configurable:!0,writable:!0,value:f}):r[o]=f,t=(r,o)=>{for(var f in o||(o={}))a.call(o,f)&&n(r,f,o[f]);if(i)for(var f of i(o))c.call(o,f)&&n(r,f,o[f]);return r};var MyLibrary=function(r){"use strict";return r.foo=t({},"foo"),r}({});

without rollupOptions.output.banner + build.minify: false

var u=Object.defineProperty;var i=Object.getOwnPropertySymbols;var a=Object.prototype.hasOwnProperty,c=Object.prototype.propertyIsEnumerable;var n=(r,o,f)=>o in r?u(r,o,{enumerable:!0,configurable:!0,writable:!0,value:f}):r[o]=f,t=(r,o)=>{for(var f in o||(o={}))a.call(o,f)&&n(r,f,o[f]);if(i)for(var f of i(o))c.call(o,f)&&n(r,f,o[f]);return r};var MyLibrary=function(r){"use strict";return r.foo=t({},"foo"),r}({});

IMHO, we should override rollupOptions.output.banner to empty string and then inject it in renderChunks hook by a internal plugin which runs after vite:esbuild-transpile.

[patak-dev]

You're right. Maybe we could cache the output.banner in case it is a function, and then use the value in renderChunk to skip that part directly.
Let's merge this one in the meantime as it helps with the unminified case.