Chunks are now always in ChunkGroups
ChunkGroups have parents and children, Chunks no longer have this
AsyncDependenciesBlocks point to a ChunkGroup instead of a list of chunks
Entrypoint is now a ChunkGroup (extends from it)
move stuff graph modification methods into GraphHelpers
Change chunk id recording to source-based
Chunks have parents, siblings and children in Stats
rename AutomaticCommonsChunksPlugin to SplitChunksPlugin
add request property to origin (ChunkGroup)
remove moveToParent from AggressiveMergingPlugin
fix issues with AggressiveSplittingPlugin
remove CommonsChunkPlugin
add optimization.runtimeChunk (RuntimeChunkPlugin)
fix a filename bug for initial chunks
remove extract-text-webpack-plugin from test suite
add DebugHash to help debugging hash issues
fix issues with entry modules in different chunk as runtime
redo optimization.splitChunks options
Compilation.entrypoints is now a Map
Compilation.chunkGroups has been added
remove some unittests
update examples
update tests with CommonsChunkPlugin
Allow lazy, lazy-once and eager mode for import()
Fix a bug which caused context chunks to be merged when using chunk name
Allow to work in env where there is no Promise.resolve
* refactor(eslint): Added no-unsed-vars, env specific .eslintrc's for buildin & hot
* This change allows for buildin to now be part of the lint test by addition of an extra .eslintrc file which extends existing but sets es6 to false
* Turned on no-unused-vars for eslint, and then removed all warnings from turning that feature on
* Was getting notified of errors in "quote" property in eslintrc, so changed to correct value from error to 2.
* beautify
* swap invalid eslint config val
The way that the hot loader is included in a bundle breaks in Java 8's
Nashorn engine, because of a missing semicolon. This change introduces
a placeholder value that can be correctly substituted when building
a bundle, but is otherwise inert.
Note that I don't particularly like this change, so if there's a better
fix, I'd like to hear about it.
If the length option on a hash was being used for a file path template
(e.g. '[hash:8].js'), the file would be outputted with the correct name,
but the reference inside require.ensure would not work, because the length
truncation would operate on more than just the hash, and would truncate
some of the characters added to pad the hash value.
Fix is to implement the `hashWithLength` function where this is a problem.
This doesn't feel like a proper solution, as it adds some redundancies
in a few places, but it's only in the chunk loading blocks that this matters,
so it's small in scope.
Generalized identifier cleansing to replace invalid (and unicode)
characters with underscores to ensure a valid function name.
Applied function in other places functions were generated based on
libarary name.
Tobias Koppers
Florent Cailhol
Tim Hambourger
Petter Häggholm
Antoine Caron
Tim Sebastian
Anatoliy Yastreb
SendilKumar N
Sean Larkin
Rory Hunter
Spencer Elliott
Avi Haiat
Jason Anderson
Kurt Harriger