Should we remove __webpack_require__.el?

[ahabhgk]

Should we remove __webpack_require__.el and alignment block_promise with webpack?

motivation:

1. __webpack_require__.el is inconsistent with webpack

2. already caused more inconsistent, and we don't how many inconsistent it will cause in the future

   • AsyncDependenciesBlockIdentifier.from

3. lead bigger output, for example:

   // entry: a.js
   import("./shared.js")

   // entry: b.js
   import("./shared.js")

   the __webpack_require__.el map of both a.js and b.js will be:

   __webpack_require__.el = { "./a.js@0:21": ["shared"], "./b.js@0:21": ["shared"] }

   which contain duplicate key

   this is caused by __webpack_require__.el is actually generated at runtime module codegen, don't know which block actually belong to current runtime, so we can only iterate all blocks, and get its chunkGroup.runtime, if the chunkGroup.runtime is the superset of the current runtime, then we include it into the __webpack_require__.el map

   for example, the chunkGroup of block "./a.js@0:21" and block "./b.js@0:21" are pointed to same chunkGroup "shared", and its runtime is ["a", "b"], so it is included into the __webpack_require__.el map of a.js and b.js

4. the reason of why we create __webpack_require__.el is for generate the correct output assets in incremental build, but actually, incremental build only affect the creation of module graph, the incorrect output assets is caused by the cache of codegen didn't invalidate correctly

[hardfist]

if we could solve the incremental build problem, then strong +1 remove the __webpack_require__.el, since it's actually a public api of module, it will be very difficult when it's relied by other runtime implementation

[ahabhgk]

I think we can solve it, will have a try once I finish MF related stuff