1557611 - async function resolution order

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[johnplaisted]

Attachment: test2.html

User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_14_5) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/74.0.3729.169 Safari/537.36

Steps to reproduce:

The await keyword has a different resolution order than Promises, which seems incorrect (and fwiw Chrome's await matches promises; Safari appears to share FireFox's behavior).

I think I've found two different cases where the order is different from promises:

• In some cases the await order determines when functions continue (rather than the resolve() order, assuming things resolve synchronously)
• In some other cases it seems to be the opposite of the resolve AND await order when the are the same order. Maybe it is because one is an await <literal> and the other is an await promise?

Tested on FireFox 66.

Actual results:

The order of asynchronous code with an async function doesn't match the equivalent promise code.

When running the file:

• async test fails
• promise test passes
• order of await determines order of async callbacks
• order resolve() determines order of promise callbacks

Expected results:

The async function order matches the Promise equivalent.

When running the file:

• async test passes
• promise test passes
• order of resolve() determines order of async callbacks
• order resolve() determines order of promise callbacks

Comment 1

[johnplaisted]

I filed a webkit bug as well

https://bugs.webkit.org/show_bug.cgi?id=198643

Comment 2

[Daniel Bodea [:danibodea]]

Using the test case above, I confirm this issue occurs in Beta v68.0b14 and Release v67.0.4, but DOE NOT occur in Nightly v69.0a1 (2019-07-02) on MacBook Pro with Mac OS 10.14.5.
Setting bug's component as (Core) JavaScript Engine.

Comment 3

[BugBot [:suhaib / :marco]]

Bugbug thinks this bug is a regression, but please revert this change in case of error.

Comment 4

[Ryan VanderMeulen [:RyanVM]]

(In reply to Bodea Daniel [:danibodea] from comment #2)

Using the test case above, I confirm this issue occurs in Beta v68.0b14 and Release v67.0.4, but DOE NOT occur in Nightly v69.0a1 (2019-07-02) on MacBook Pro with Mac OS 10.14.5.
Setting bug's component as (Core) JavaScript Engine.

Can you please bisect what fixed this?

Comment 5

[Daniel Bodea [:danibodea]]

Attachment: find fix incomplete due to insuficient builds.txt

Unfortunately, there were insufficient builds to bisect the fix completely, but these are the logs. Hope it helps.

Comment 6

[Ryan VanderMeulen [:RyanVM]]

Bug 1546145 stands out as a likely candidate in that range. I don't think that's a fix we could safely consider backporting, however, so this may just need to ride out with 69.

Comment 7

[André Bargull [:anba]]

Yes, the difference is caused by the spec change implemented in bug 1495072 (and enabled by default in bug 1546145).

Comment 8

[Ryan VanderMeulen [:RyanVM]]

Thanks, calling this fixed by those two bugs then. But yeah, that's definitely not a change we can realistically consider backporting to stable branches, so 69+ it is.

Comment 9

[johnplaisted]

Thanks!