Bug 220364

Summary:

Should SharedArrayBuffer/WebAssembly.Memory really throw?

Product:

WebKit

Reporter:

Sam Sneddon [:gsnedders] <gsnedders>

Component:

JavaScriptCore

Assignee:

Yusuke Suzuki <ysuzuki>

Status:

RESOLVED FIXED

Severity:

Normal

CC:

clopez, eric.carlson, ews-watchlist, glenn, jer.noble, keith_miller, mark.lam, msaboff, philipj, saam, sergio, simon.fraser, tzagallo, webkit-bug-importer, youennf, ysuzuki

Priority:

Keywords:

InRadar, WPTImpact

Version:

Safari Technology Preview

Hardware:

Unspecified

OS:

Unspecified

See Also:

https://bugs.webkit.org/show_bug.cgi?id=218944
https://bugs.webkit.org/show_bug.cgi?id=219076
https://bugs.webkit.org/show_bug.cgi?id=218693
https://github.com/web-platform-tests/wpt/issues/27065

Attachments:

Description	Flags
Patch	mark.lam: review+, ews-feeder: commit-queue-
Patch	none

Description Sam Sneddon [:gsnedders] 2021-01-06 08:37:54 PST

Currently in ToT (and STP 117+), trying to create a SharedArrayBuffer throws with "Shared WebAssembly.Memory and SharedArrayBuffer are not enabled".

This has caused a variety of WPT tests to regress (e.g., many encoding tests), due to https://github.com/web-platform-tests/wpt/blob/master/common/sab.js silently creating a non-shared WebAssembly.Memory (as would be expected of an unsupported option) previously but now throwing with that TypeError.

Comment 1 Radar WebKit Bug Importer 2021-01-13 08:38:11 PST

<rdar://problem/73152534>

Comment 2 Yusuke Suzuki 2021-01-22 18:13:10 PST

Created attachment 418204 [details]
Patch

Comment 3 Yusuke Suzuki 2021-01-22 18:26:35 PST

https://github.com/web-platform-tests/wpt/commit/6d9af4c25eb93a12891f8bd88bc21510013b5ed5 Now, sab.js is fixed.

Comment 4 Mark Lam 2021-01-22 19:20:26 PST

Comment on attachment 418204 [details]
Patch

r=me if EWS bots are green.

Comment 5 Yusuke Suzuki 2021-01-22 19:41:46 PST

I'll only update sab.js and wasm for now.

Comment 6 Yusuke Suzuki 2021-01-22 19:43:54 PST

Created attachment 418208 [details]
Patch

Comment 7 Yusuke Suzuki 2021-01-22 20:40:24 PST

Debug failure http/tests/loading/basic-auth-load-URL-with-consecutive-slashes.html is flaky and unrelated to wpt / wasm

Comment 8 Yusuke Suzuki 2021-01-22 20:58:54 PST

Committed r271774: <https://trac.webkit.org/changeset/271774>

Comment 9 Simon Fraser (smfr) 2021-01-23 11:12:36 PST

Comment on attachment 418208 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=418208&action=review

> Source/JavaScriptCore/ChangeLog:8
> +        Not accessing "shared" field if Options::useSharedArrayBuffer() is false.

This doesn't tell me the answer to the question in the title.

Comment 10 Mark Lam 2021-01-23 11:37:01 PST

(In reply to Simon Fraser (smfr) from comment #9)
> Comment on attachment 418208 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=418208&action=review
> 
> > Source/JavaScriptCore/ChangeLog:8
> > +        Not accessing "shared" field if Options::useSharedArrayBuffer() is false.
> 
> This doesn't tell me the answer to the question in the title.

The answer is no.  The above comment was trying to say, "we should not access the `shared` field if Options::useSharedArrayBuffer() is false".

Comment 11 Sam Sneddon [:gsnedders] 2021-05-14 06:48:37 PDT

This seemingly never made it into safari-611-branch, and accounts for a fair bit of the regressions from Safari 14.0 -> 14.1 on https://wpt.fyi/results/?q=seq%28status%3Apass%7Cstatus%3Aok%20status%3Apass%7Cstatus%3Aok%20status%3A%21pass%26status%3A%21ok%26status%3A%21unknown%20status%3A%21pass%26status%3A%21ok%26status%3A%21unknown%29%20&run_id=6016518288572416&run_id=5667597628473344&run_id=5748130572140544&run_id=5766281976348672 :(

Comment 12 Sam Sneddon [:gsnedders] 2021-05-14 06:52:06 PDT

Filed https://github.com/web-platform-tests/wpt/pull/28998 to workaround the constructor unexpectedly throwing.