Bump swc #8029

mischnic · 2022-04-29T20:46:12Z

Some changes are needed for feat(es/resolver): Use different syntax context for unresolved refs swc-project/swc#4436 (when we add previously undefined variables)
There were some breaking changes already, but let's wait for preset_env with bugfixes for Safari 15 messes up syntax contexts swc-project/swc#4488
Add test for Symbol.toStringTag #7911
First source mapping is missing swc-project/swc#4608 is blocking the tests

devongovett · 2022-05-01T15:10:58Z

Can the unresolved_mark be used to solve issues like #7911?

mischnic · 2022-05-06T21:18:30Z

Can the unresolved_mark be used to solve issues like #7911?

I just had the exact same thing happen with __filename not being renamed by hoist. In that case the problem was 2ad46b5, but this doesn't solve that issue, maybe we can get swc to add global_mark for that variable (I imagine this is what's happening)

Though I'm not actually sure if that commit was correct or just made things worse...

mischnic · 2022-05-07T08:40:26Z

So it turns out that global_mark is only supposed to be set for user-defined global variables. That explains why #7911 is broken.
Unfortunately the only valid check for whether a variable is top-level is by checking in the visitor whether it's in a function/block/...

devongovett

Is this ready or is the first checkbox in the description still pending?

mischnic · 2022-05-09T14:56:54Z

This whole global_mark/global_syntax situation is still not fully resolved (which is also the problem for #7911). Maybe I can add some hacks to make it work for now

mischnic · 2022-05-09T20:39:27Z

This should now be correct (and fix that one issue), but it introduces one extra collect_decls for everything and additionally when scope hoisting a fold with chain!(hygiene(),resolver(unresolved_mark, global_mark, config.is_type_script)) plus another collect_decls call.

So we'll want to benchmark if this makes a difference. If it's too slow, we could alternatively create a custom visitor that collects all top level variables into a set and then check against that instead of has_mark(global_mark) in hoist.

mischnic · 2022-05-10T20:00:10Z

Last blocker for the tests: swc-project/swc#4608

mischnic · 2022-05-10T20:47:10Z

So we'll want to benchmark if this makes a difference. If it's too slow, we could alternatively create a custom visitor that collects all top level variables

Build ak-editor is equally fast

parcel-benchmark · 2022-05-11T18:36:57Z

Benchmark Results

Kitchen Sink 🚨

Timings

Description	Time	Difference
Cold	FAILED	-0.00ms
Cached	FAILED	-0.00ms

Cold Bundles

No bundles found, this is probably a failed build...

Cached Bundles

No bundles found, this is probably a failed build...

React HackerNews ✅

Timings

Description	Time	Difference
Cold	10.59s	-781.00ms 🚀
Cached	528.00ms	-2.00ms

Cold Bundles

Bundle	Size	Difference	Time	Difference
dist/index.js	486.35kb	+25.00b ⚠️	5.54s	-470.00ms 🚀
dist/PermalinkedComment.46b19af5.js	4.21kb	+0.00b	5.54s	-471.00ms 🚀
dist/UserProfile.f8cd7884.js	1.57kb	+0.00b	5.54s	-470.00ms 🚀
dist/NotFound.960ab92b.js	429.00b	+0.00b	5.54s	-471.00ms 🚀
dist/logo.c5bb83f1.png	246.00b	+0.00b	5.48s	-475.00ms 🚀

Cached Bundles

Bundle	Size	Difference	Time	Difference
dist/index.js	486.35kb	+25.00b ⚠️	5.63s	-196.00ms

AtlasKit Editor ✅

Timings

Description	Time	Difference
Cold	1.91m	-29.00ms
Cached	2.94s	-75.00ms

Cold Bundles

Bundle	Size	Difference	Time	Difference
dist/index.494dca73.js	3.33mb	-751.00b 🚀	37.46s	+215.00ms
dist/index.8d3be486.js	1.27mb	+1.02kb ⚠️	1.44m	-779.00ms
dist/card.8ef3bfa3.js	366.77kb	-30.00b 🚀	1.44m	-774.00ms
dist/ConfigPanelFieldsLoader.6adc99a5.js	288.22kb	-24.00b 🚀	41.28s	+117.00ms
dist/EmojiPickerComponent.347f8762.js	187.30kb	-10.00b 🚀	37.48s	+197.00ms
dist/mobile-upload.a62de28d.js	68.25kb	-31.00b 🚀	37.46s	+214.00ms
dist/esm.c7dc1640.js	61.96kb	+0.00b	47.35s	-39.72s 🚀
dist/ElementBrowser.1c600f9b.js	57.31kb	-21.00b 🚀	42.62s	+74.00ms
dist/dropzone.ca6653ff.js	13.83kb	-11.00b 🚀	47.35s	+527.00ms
dist/pdfRenderer.9d4c17f1.js	12.17kb	+1.00b ⚠️	47.35s	+527.00ms
dist/workerHasher.e50d242f.js	1.72kb	+0.00b	47.35s	-39.72s 🚀
dist/16.1969624f.js	1.08kb	+0.00b	42.62s	-4.20s 🚀
dist/16.069344b7.js	905.00b	+0.00b	42.62s	-4.20s 🚀
dist/simpleHasher.46d6f2e5.js	742.00b	+0.00b	47.35s	-39.72s 🚀
dist/index.html	240.00b	+0.00b	37.48s	-49.63s 🚀

Cached Bundles

Bundle	Size	Difference	Time	Difference
dist/index.32f08f28.js	3.33mb	-751.00b 🚀	37.15s	-458.00ms
dist/index.ad6d475a.js	1.27mb	+1.02kb ⚠️	1.46m	+1.58s
dist/card.8ef3bfa3.js	366.77kb	-30.00b 🚀	1.46m	+1.58s
dist/ConfigPanelFieldsLoader.6adc99a5.js	288.22kb	-24.00b 🚀	40.91s	-616.00ms
dist/EmojiPickerComponent.347f8762.js	187.30kb	-10.00b 🚀	37.19s	-456.00ms
dist/mobile-upload.a62de28d.js	68.25kb	-31.00b 🚀	37.15s	-458.00ms
dist/esm.c7dc1640.js	61.96kb	+0.00b	46.90s	-39.02s 🚀
dist/ElementBrowser.1c600f9b.js	57.31kb	-21.00b 🚀	42.26s	-587.00ms
dist/dropzone.ca6653ff.js	13.83kb	-11.00b 🚀	46.91s	+19.00ms
dist/pdfRenderer.9d4c17f1.js	12.17kb	+1.00b ⚠️	46.90s	+18.00ms
dist/workerHasher.e50d242f.js	1.72kb	+0.00b	46.90s	-39.02s 🚀
dist/16.1969624f.js	1.08kb	+0.00b	42.26s	-4.62s 🚀
dist/index.html	240.00b	+0.00b	40.91s	-45.05s 🚀

Three.js ✅

Timings

Description	Time	Difference
Cold	7.86s	+136.00ms
Cached	345.00ms	+18.00ms ⚠️

Cold Bundles

Bundle	Size	Difference	Time	Difference
dist/Three.js	580.03kb	+13.00b ⚠️	5.59s	+86.00ms

Cached Bundles

Bundle	Size	Difference	Time	Difference
dist/Three.js	580.03kb	+13.00b ⚠️	5.89s	+217.00ms

Click here to view a detailed benchmark overview.

packages/transformers/js/core/src/lib.rs

devongovett · 2022-05-13T02:40:10Z

packages/transformers/js/core/src/lib.rs

+                  hygiene(),
+                  resolver(unresolved_mark, global_mark, false)
+                ));
+                (collect_decls(&module), module)


So if the node replacements pass doesn't run we do this twice in a row? Also, are there more cases where these extra passes can be skipped (e.g. if no preset-env)?

I've made our replacement passes use the "incorrect" global syntax context directly (which is what will be the case after the resolver reruns anyway). So now it's only needed for preset_env.

collect_decls has to always be rerun here because of line 3:

// Flush (JsWord, SyntaxContexts) into unique names and reresolve to // set global_mark for all nodes, even generated ones. // - This changes the syntax context ids and therefore invalidates decls

hmm I guess the real question is why the first one exists before the node replacements pass, or why the node pass runs after preset-env rather than before?

devongovett · 2022-05-17T02:38:15Z

I guess some tests are failing now?

mischnic marked this pull request as ready for review April 30, 2022 09:25

mischnic mentioned this pull request May 3, 2022

Parcel not respecting tsconfig useDefineForClassFields: false with decorators #8036

Closed

mischnic added 2 commits May 6, 2022 21:07

Bump swc

b1bc4dc

Bump swc

7c92208

mischnic force-pushed the bump-swc branch from bc38b26 to 7c92208 Compare May 6, 2022 19:08

mischnic added 3 commits May 6, 2022 21:53

Update GlobalReplacer to rewrite syntax context

9b541f6

Node_replacer

6fb8e21

Fixup

87927c7

mischnic force-pushed the bump-swc branch from e74a989 to 2ad46b5 Compare May 6, 2022 21:15

devongovett approved these changes May 8, 2022

View reviewed changes

mischnic added 3 commits May 9, 2022 17:11

Bump swc helpers

134cf23

TODO comment

e801248

Bump crates

50c833a

mischnic force-pushed the bump-swc branch from 2ad46b5 to 50c833a Compare May 9, 2022 15:17

mischnic added 3 commits May 9, 2022 17:17

Merge branch 'v2' into bump-swc

65b122f

Hack in hoist to "detect" top level bindings

f96db99

Bump

63b441f

mischnic force-pushed the bump-swc branch from 9827967 to f3b832f Compare May 9, 2022 20:38

Reresolve to set global_mark for generated vars

91dafd3

mischnic force-pushed the bump-swc branch from f3b832f to 91dafd3 Compare May 10, 2022 07:20

mischnic added 2 commits May 10, 2022 18:20

Fixup

7b33986

Bump

b1edad8

mischnic added 2 commits May 10, 2022 22:06

Flowify sourcemaps.js

b24752f

Add test for renaming swc inserted vars

c5cfd7e

mischnic added 2 commits May 11, 2022 18:32

Bump

7e55659

Merge branch 'v2' into bump-swc

54f4ea8

mischnic force-pushed the bump-swc branch from fa11fa7 to 54f4ea8 Compare May 11, 2022 16:33

Handle dummy locs gracefully

73a4b70

Fixup test assertion

6d0b635

mischnic requested a review from devongovett May 11, 2022 20:46

devongovett reviewed May 13, 2022

View reviewed changes

packages/transformers/js/core/src/lib.rs Outdated Show resolved Hide resolved

devongovett reviewed May 13, 2022

View reviewed changes

mischnic added 5 commits May 13, 2022 18:00

Recollect decls only if preset-env ran

c322aa9

Bump

15fe219

Some performance hacks

f429db5

Merge branch 'v2' into bump-swc

07da075

Bump swc

9e6730c

devongovett approved these changes May 17, 2022

View reviewed changes

mischnic added 3 commits May 17, 2022 10:53

Update quote type in tests

45c305d

Run node replacer before preset-env

647ffa2

Merge branch 'v2' into bump-swc

61281ac

devongovett approved these changes May 17, 2022

View reviewed changes

devongovett merged commit 4e40429 into v2 May 17, 2022

devongovett deleted the bump-swc branch May 17, 2022 14:02

mischnic mentioned this pull request Feb 3, 2024

Remove decl_collector pass and use SWC's unresolved_mark instead #9520

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Bump swc #8029

Bump swc #8029

mischnic commented Apr 29, 2022 •

edited

devongovett commented May 1, 2022

mischnic commented May 6, 2022 •

edited

mischnic commented May 7, 2022

devongovett left a comment

mischnic commented May 9, 2022

mischnic commented May 9, 2022

mischnic commented May 10, 2022

mischnic commented May 10, 2022

parcel-benchmark commented May 11, 2022 •

edited

Timings

Cold Bundles

Cached Bundles

Timings

Cold Bundles

Cached Bundles

Timings

Cold Bundles

Cached Bundles

Timings

Cold Bundles

Cached Bundles

devongovett May 13, 2022

mischnic May 13, 2022

devongovett May 17, 2022

devongovett commented May 17, 2022

Bump swc #8029

Bump swc #8029

Conversation

mischnic commented Apr 29, 2022 • edited

devongovett commented May 1, 2022

mischnic commented May 6, 2022 • edited

mischnic commented May 7, 2022

devongovett left a comment

Choose a reason for hiding this comment

mischnic commented May 9, 2022

mischnic commented May 9, 2022

mischnic commented May 10, 2022

mischnic commented May 10, 2022

parcel-benchmark commented May 11, 2022 • edited

Benchmark Results

Timings

Cold Bundles

Cached Bundles

Timings

Cold Bundles

Cached Bundles

Timings

Cold Bundles

Cached Bundles

Timings

Cold Bundles

Cached Bundles

devongovett May 13, 2022

Choose a reason for hiding this comment

mischnic May 13, 2022

Choose a reason for hiding this comment

devongovett May 17, 2022

Choose a reason for hiding this comment

devongovett commented May 17, 2022

mischnic commented Apr 29, 2022 •

edited

mischnic commented May 6, 2022 •

edited

parcel-benchmark commented May 11, 2022 •

edited