REGRESSION: [ Catalina WK2 ] http/wpt/webauthn/public-key-credential-create-success-u2f.https.html is failing

[Jiewen Tan]

REGRESSION: [ Catalina WK2 ] http/wpt/webauthn/public-key-credential-create-success-u2f.https.html is failing.

[Jiewen Tan]

<rdar://problem/51524958>

[Jiewen Tan]

Created attachment 378420 [details]
Patch

[Alex Christensen]

Comment on attachment 378420 [details]
Patch

This shouldn't change any behavior, but it does introduce an unnecessary Victor copy.

[Jiewen Tan]

(In reply to Alex Christensen from comment #3)
> Comment on attachment 378420 [details]
> Patch
> 
> This shouldn't change any behavior, but it does introduce an unnecessary
> Victor copy.

Where?

[Alex Christensen]

Comment on attachment 378420 [details]
Patch

If this does fix anything, we should definitely fix Vector to prevent other such bugs.  I don't think it will fix anything though, but you're welcome to try.  Please revert if it doesn't fix.

[Jiewen Tan]

Comment on attachment 378420 [details]
Patch

Thanks Alex for r+ it. I will definitely keep the bug open and keep an eye on the bots.

[WebKit Commit Bot]

Comment on attachment 378420 [details]
Patch

Clearing flags on attachment: 378420

Committed r249737: <https://trac.webkit.org/changeset/249737>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Jiewen Tan]

Reopened for continuous monitoring bots.

[Jiewen Tan]

(In reply to Alex Christensen from comment #5)
> Comment on attachment 378420 [details]
> Patch
> 
> If this does fix anything, we should definitely fix Vector to prevent other
> such bugs.  I don't think it will fix anything though, but you're welcome to
> try.  Please revert if it doesn't fix.

Believe it or not. It did fix the bots:
https://results.safari.apple.com/?suite=layout-tests&test=http%2Fwpt%2Fwebauthn%2Fpublic-key-credential-create-success-u2f.https.html

I will try to fix the vector issue as well then.

[Chris Dumez]

Comment on attachment 378420 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=378420&action=review

> Source/WebCore/Modules/webauthn/fido/U2fResponseConverter.cpp:102
> +    memset(aaguid.data(), 0, aaguidLength);

Hard to believe this would change anything indeed.

[Jiewen Tan]

(In reply to Chris Dumez from comment #11)
> Comment on attachment 378420 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=378420&action=review
> 
> > Source/WebCore/Modules/webauthn/fido/U2fResponseConverter.cpp:102
> > +    memset(aaguid.data(), 0, aaguidLength);
> 
> Hard to believe this would change anything indeed.

Still not able to reproduce the issue in my other Mac that has less memory. Will upload a patch to confirm the issue.

[Jiewen Tan]

Created attachment 379675 [details]
Try to confirm the cause.

[WebKit Commit Bot]

Comment on attachment 379675 [details]
Try to confirm the cause.

Clearing flags on attachment: 379675

Committed r250397: <https://trac.webkit.org/changeset/250397>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Jiewen Tan]

Reopened for continuous monitoring bots.

[Jiewen Tan]

r250397 confirmed the issue.

[Jiewen Tan]

I guess I need to ask bot watcher for a machine to reproduce the bug before I make any changes to Vector. From code inspection, it is really hard to even make a speculative fix.

[Jiewen Tan]

Reverted r250397 for reason:

caused failure on iOS 13

Committed r250455: <https://trac.webkit.org/changeset/250455>

[Alex Christensen]

The real fix for this is in https://bugs.webkit.org/show_bug.cgi?id=201902

[Jiewen Tan]

(In reply to Alex Christensen from comment #20)
> The real fix for this is in https://bugs.webkit.org/show_bug.cgi?id=201902

It looks super cool.

[Jiewen Tan]

*** This bug has been marked as a duplicate of bug 201902 ***