build UMD bundle and polyfill language features #4366

boneskull · 2020-07-06T20:14:35Z

remove old ESlint rule to disallow Object.assign
consume Rollup config's output prop in karma-rollup-plugin
simplify rollup.config.js output prop
fix invalid script tag src attrib in lib/browser/template.html (& reformat)

boneskull · 2020-07-06T20:18:42Z

@Munter I found that when using umd output while window.mocha.setup did not exist, window.mocha.mocha.setup did exist.

using object spread syntax module.exports = {...global, ...mocha} causes Rollup (?) to generate a ES module then complains about having default and named exports. I have no idea why.

I think even the global.mocha = mocha and global.Mocha = Mocha lines are probably unnecessary, since the UMD bundle will stuff module.exports into global.

boneskull · 2020-07-06T20:23:00Z

update: the global.mocha and global.Mocha stuff is necessary. I guess I really don't get it. but whatever.

coveralls · 2020-07-06T20:29:01Z

Coverage remained the same at 93.932% when pulling 2cf235b on boneskull/rollup-umd into 02bdb6b on master.

boneskull · 2020-07-06T20:29:51Z

looks like I jumped the gun a little there. it worked for a test file anyway (as created by mocha init)

boneskull · 2020-07-06T21:48:32Z

alright, tweaked the exports. it's now

mocha.Mocha = Mocha;
mocha.mocha = mocha;
module.exports = Object.assign(mocha, global);

boneskull · 2020-07-06T22:18:13Z

AMD bundle seems to "work" per manual test, but the failing test isn't of much value.

boneskull · 2020-07-06T23:22:20Z

added a proper test for AMD; closes #2426

boneskull · 2020-07-06T23:23:02Z

karma.conf.js

@@ -134,19 +135,38 @@ module.exports = config => {
        {pattern: 'test/browser-specific/esm.spec.mjs', type: 'module'}
      ];
      break;
+    case 'requirejs':
+      cfg.plugins.push('karma-requirejs');
+      cfg.frameworks.unshift('requirejs');


why does order matter? I have no idea

boneskull · 2020-07-06T23:23:52Z

karma.conf.js

-    require.resolve('./test/browser-specific/setup')
-  );
+  if (MOCHA_TEST !== 'requirejs') {
+    cfg.files.unshift(


unexpected is broken when used with AMD. something about mismatched defines

boneskull · 2020-07-07T17:34:19Z

@Munter something is not pulling in the target browser list here

boneskull · 2020-07-07T19:49:08Z

wow, it works!

boneskull · 2020-07-07T20:19:53Z

test/unit/utils.spec.js

@@ -496,11 +496,11 @@ describe('lib/utils', function() {
    if (typeof global.Symbol === 'function') {
      it('should handle Symbol', function() {
        var symbol = Symbol('value');
-        expect(stringify(symbol), 'to be', 'Symbol(value)');
+        expect(stringify(symbol), 'to match', /^Symbol\(value\)/);


core-js adds weird characters to the Symbol string representation in IE11. 🤷‍♂️

boneskull · 2020-07-07T20:22:04Z

karma.conf.js

@@ -17,7 +17,7 @@ const browserPlatformPairs = {
 };

 module.exports = config => {
-  let bundleDirpath;
+  let bundleDirpath = path.join(baseBundleDirpath, hostname);


previously, this would have been added to the cfg object before it was defined. now, we only add it if creating the directory succeeds.

maybe we should ignore a mkdir failure and just use the tmp dir... I don't want to run the build again 😜

boneskull · 2020-07-07T20:22:40Z

lib/browser/template.html

+    <script>
+      mocha.setup('bdd');
+    </script>
+    <script src="tests.spec.js"></script>


this was wrong.

we should run linkcheck on this thing.

- removed dead code in `lib/cli/run-helpers.js` - be sure not to send certain options to the worker processes - fix root hook bug in worker - send global fixtures to `Runner` constructor - removed debugs - **add `Process.unhandledRejection` listeners** which might be bad - fix some tests while I'm fixing tests

boneskull · 2020-07-13T19:46:40Z

.travis.yml

@@ -49,7 +49,7 @@ jobs:
      node_js: '10'
      name: 'Node.js v10'

-    - script: npm start test.bundle test.browser
+    - script: npm start test.browser


test.bundle becomes test.browser.requirejs

- removed dead code in `lib/cli/run-helpers.js` - be sure not to send certain options to the worker processes - fix root hook bug in worker - send global fixtures to `Runner` constructor - removed debugs - **add `Process.unhandledRejection` listeners** which might be bad - fix some tests while I'm fixing tests

- remove old ESlint rule to disallow `Object.assign` - consume Rollup config's `output` prop in `karma-rollup-plugin` - simplify `rollup.config.js` `output` prop - fix invalid `script` tag `src` attrib in `lib/browser/template.html` (& reformat)

- fix broken bundle output path

- remove the old `bundle/amd.spec.js` test - _actually_ fix bundle output file path - fix a typo in karma rollup plugin

Signed-off-by: Christopher Hiller <boneskull@boneskull.com>

- `unexpected`, `unexpected-eventemitter`, and `unexpected-sinon` are now external (not bundled) and loaded via their globals - modified `test/browser-specific/setup.js` to use `require()` since we can do that now - bundle now has an inline source map since I couldn't figure out any other way to load it - removed duplication of work in the karma plugin; no call to `bundle.generate()` is needed. furthermore the `code` prop did not include the inline source maps. so we just write the file and read the result. could probably avoid reading the file by manually stitching the `code` and `map` props together, but I'm unsure how - loads `mocha.js.map`

juergba · 2020-07-24T09:14:48Z

@boneskull I can't review this PR explained by my complete incompetence in this matter, sorry.
This rollup thing is a wild fusion of bundling, transpiling and karma/testing without clear distinction. I don't understand neither why we need an additional module definition UMD. What definition have we used before with browserify?

boneskull · 2020-07-28T21:06:44Z

our browserify bundle is not quite UMD, but it 1. defines globals and 2. enables bundling of the bundle. it does not supply a AMD-compatible bundle, but it also does not inhibit AMD bundles.

The rationale for this PR, then, is:

the iife output of Rollup does not seem to allow number 2, which is going to break consumers; we need the umd output style
after Rollup and babel in build pipeline #4293 landed, we're able to use some modern syntax everywhere, but not all modern syntax everywhere. the changes to our babel config in this PR make it possible to use, e.g., async/await anywhere in our sources and tests (because Node.js v10, our minimum-supported version of Node.js, allows this). We could not use the native Promise object, either--but now we can, because of the polyfilling core-js does.
the umd output style of Rollup does provide an AMD bundle, which made the previous test case (test.bundle) fail. I had to write some new tests to assert compatibility (and learn some RequireJS...ugh)

with this PR--for whatever reason--the resulting bundle size is about 20Kb less.

boneskull · 2020-07-28T21:07:14Z

that said, this has been sitting for like 3 weeks, and I need to move forward with it to get global fixtures in.

- adds a generic "plugin loader" which handles root hooks, global setup/teardown, future - renamed `validatePlugin` to `validateLegacyPlugin`. reporters/interfaces should eventually be refactored to use this plugin system - should maybe decouple plugins from the `Mocha` prototype - new events emitted by `Runner` for global setup/teardown - need to add unit tests, need to support parallel mode upgrade eslint-config-semistandard more implementation, tests added more tests, implementation. needs #4366 for browser compat - removed dead code in `lib/cli/run-helpers.js` - be sure not to send certain options to the worker processes - fix root hook bug in worker - send global fixtures to `Runner` constructor - removed debugs - **add `Process.unhandledRejection` listeners** which might be bad - fix some tests while I'm fixing tests reorganize plugin fixtures add "to contain once" mocha result output assertion (to ensure some output is not duplicated) add runGlobalSetup, runGlobalTeardown to Mocha revert changes to Runner `Mocha` will handle global fixtures no events will be emitted remove test/integration/options/require.spec.js which is now in test/integration/plugins/ revert change to ParallelBufferedRunner which invoked global fixtures pull helper code out of test/integration/options/watch.spec.js into integration test helpers - because it's nice to be able to test watching elsewhere - also renamed `runMochaWatch` to `runMochaWatchJSON` and provide `runMochaWatch` for plain output move tests from test/integration/options/require.spec.js into test/integration/plugins/root-hooks.spec.js Mocha#run now runs global fixtures; add watch mode support - add `enableGlobalSetup()` and `enableGlobalTeardown()` methods on `Mocha` - update `globalSetup()` and `globalTeardown()` methods - watch mode runs global teardown upon `SIGINT`. is that right? add command prop to runMochaJSON return value if a fixture run with `runMochaJSON()` fails, the return value now contains a `command` prop (like the result of `runMocha()`) for easier debugging remove unused garbage from RUnner fix bug where mocha would never exit with a non-zero code - I think this is understandable, because `done()` is weirdly _not_ an error-first callback - remove try/catch guards around fixtures and let `process` uncaught/unhandled listeners deal w/ it - also fix `unit/mocha.spec.js` test update package-lock.json move plugin.spec.js into node tests for now remove unused 'watchify', add 'touch' Signed-off-by: Christopher Hiller <boneskull@boneskull.com> remove noisy debugs in Runner fix flaky test in test/unit/mocha.spec.js rename assertion types - `JSONRunResult` -> `JSONResult` - `RawRunResult` -> `SummarizedResult` - `RawResult` remains the same - tweak identification functions integration test helper improvements - better organization - renamed stuff for consistency - `runMochaWatch` -> `runMochaWatchAsync` - `splitRegExp` -> `SPLIT_DOT_REPORTER_REGEXP` - rename parameters, add and fix docstrings - modernize some of it - add various typedefs to align with assertion types - remove `invokeSubMocha` as it is no longer needed now that `mocha` will not fork a process if no `node` options are provided - add special case for _forking_ mocha on win32 in watch mode, as that's the only way we can cleanup cleanly (parallel runs will need this treatment too, but I think it demands a more generalized solution) - fix potential bug in `resolveFixturePath()` - in `runMochaWatchAsync`, add `sleepMs` option; defaults to 2s - extract `createTempDir()` from various test files and expose on helper - prefer `rimraf` over `fs-extra.remove` since the former will retry when dir is locked on win32 - expose `touchFile()` on helper; do not use wonky handrolled touching algorithms and use [touch](https://npm.im/touch) instead - update tests to use new helpers and renamed helpers watch improvements and refactors - print "waiting" msg to stderr (_should_ be ok) - add "cleaning up" message upon ctrl-c - remove needless `afterRun` option from various places - only run teardown fixtures if they exist - add win32 fix for testing - add `Mocha#hasGlobalSetupFixtures()` and `Mocha#hasGlobalTeardownFixtures()` more tests Signed-off-by: Christopher Hiller <boneskull@boneskull.com> upgrade rewiremock Signed-off-by: Christopher Hiller <boneskull@boneskull.com> rename lib/plugin => lib/plugin-loader - rename `createInvalidPluginError` => `createInvalidLegacyPluginError` and soft-deprecate - add `createInvalidPluginDefinitionError` and `createInvalidPluginImplementationError` w/ constants; use them - move some typedefs into `lib/mocha.js` as they are referenced via public aforementioned error factories - remove TS docstrings - better coverage - move `plugin-loader` test back into `test/unit` (removed rewiremock from it)

boneskull added area: browser browser-specific type: chore generally involving deps, tooling, configuration, etc. semver-patch implementation requires increase of "patch" version number; "bug fixes" labels Jul 6, 2020

boneskull requested a review from Munter July 6, 2020 20:14

boneskull self-assigned this Jul 6, 2020

boneskull commented Jul 6, 2020

View reviewed changes

boneskull force-pushed the boneskull/rollup-umd branch from 81339e5 to 063689a Compare July 7, 2020 20:11

boneskull commented Jul 7, 2020

View reviewed changes

boneskull changed the title ~~build a UMD bundle instead of IIFE~~ build UMD bundle and polyfill language features Jul 13, 2020

boneskull commented Jul 13, 2020

View reviewed changes

boneskull added 4 commits July 17, 2020 14:46

build a UMD bundle instead of IIFE

d6694f4

- remove old ESlint rule to disallow `Object.assign` - consume Rollup config's `output` prop in `karma-rollup-plugin` - simplify `rollup.config.js` `output` prop - fix invalid `script` tag `src` attrib in `lib/browser/template.html` (& reformat)

fix broken tests

f112e19

- fix broken bundle output path

add a proper RequireJS integration test

8db4996

- remove the old `bundle/amd.spec.js` test - _actually_ fix bundle output file path - fix a typo in karma rollup plugin

fix travis script

cabb0ab

Signed-off-by: Christopher Hiller <boneskull@boneskull.com>

boneskull added 8 commits July 17, 2020 14:46

use core-js polyfills

b9aebeb

remove regenerator-runtime; copy/paste config from stackoverflow

8a07559

Signed-off-by: Christopher Hiller <boneskull@boneskull.com>

update Symbol test for IE11

556cbc1

Signed-off-by: Christopher Hiller <boneskull@boneskull.com>

revert change to Symbol-in-object test

6486f05

Signed-off-by: Christopher Hiller <boneskull@boneskull.com>

tweak assertion AGAIN for Symbol

fe7227e

Signed-off-by: Christopher Hiller <boneskull@boneskull.com>

merge fixes

9cd9e23

try to make karma.conf.js less indimidating

66369b4

Signed-off-by: Christopher Hiller <boneskull@boneskull.com>

boneskull force-pushed the boneskull/rollup-umd branch from 0cce38e to 2cf235b Compare July 17, 2020 21:46

boneskull requested review from juergba, outsideris, craigtaub and nicojs July 22, 2020 20:14

boneskull merged commit ad03d29 into master Jul 28, 2020

boneskull deleted the boneskull/rollup-umd branch July 28, 2020 21:11

boneskull added this to the next milestone Jul 29, 2020

outsideris mentioned this pull request Aug 8, 2020

"Mocha" not defined #4395

Closed

2 tasks

juergba mentioned this pull request Feb 12, 2021

add integration tests for AMD #2426

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

build UMD bundle and polyfill language features #4366

build UMD bundle and polyfill language features #4366

boneskull commented Jul 6, 2020

boneskull commented Jul 6, 2020

boneskull commented Jul 6, 2020

coveralls commented Jul 6, 2020 •

edited

boneskull commented Jul 6, 2020

boneskull commented Jul 6, 2020

boneskull commented Jul 6, 2020

boneskull commented Jul 6, 2020

boneskull Jul 6, 2020

boneskull Jul 6, 2020 •

edited

boneskull commented Jul 7, 2020

boneskull commented Jul 7, 2020

boneskull Jul 7, 2020

boneskull Jul 7, 2020

boneskull Jul 7, 2020

boneskull Jul 13, 2020

juergba commented Jul 24, 2020

boneskull commented Jul 28, 2020

boneskull commented Jul 28, 2020

build UMD bundle and polyfill language features #4366

build UMD bundle and polyfill language features #4366

Conversation

boneskull commented Jul 6, 2020

boneskull commented Jul 6, 2020

boneskull commented Jul 6, 2020

coveralls commented Jul 6, 2020 • edited

boneskull commented Jul 6, 2020

boneskull commented Jul 6, 2020

boneskull commented Jul 6, 2020

boneskull commented Jul 6, 2020

boneskull Jul 6, 2020

Choose a reason for hiding this comment

boneskull Jul 6, 2020 • edited

Choose a reason for hiding this comment

boneskull commented Jul 7, 2020

boneskull commented Jul 7, 2020

boneskull Jul 7, 2020

Choose a reason for hiding this comment

boneskull Jul 7, 2020

Choose a reason for hiding this comment

boneskull Jul 7, 2020

Choose a reason for hiding this comment

boneskull Jul 13, 2020

Choose a reason for hiding this comment

juergba commented Jul 24, 2020

boneskull commented Jul 28, 2020

boneskull commented Jul 28, 2020

coveralls commented Jul 6, 2020 •

edited

boneskull Jul 6, 2020 •

edited