Bug 213653 – [webkitpy] Automatically detect hw architecture for supporting Apple Silicon

• Status: RESOLVED
• Resolution: FIXED
• Component: Tools / Tests

[Jonathan Bedard]

webkitpy needs to support Macs with an arm64 architecture and needs some testing support for Rosetta.

[Radar WebKit Bug Importer]

<rdar://problem/64817656>

[Jonathan Bedard]

Created attachment 402888 [details]
Patch

[Jonathan Bedard]

(In reply to Jonathan Bedard from comment #2)
> Created attachment 402888 [details]
> Patch

This doesn't handle Rosetta yet, but it means that machines running on Apple Silicon with native WebKit will report correctly.

[Aakash Jain]

Comment on attachment 402888 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=402888&action=review

> Tools/Scripts/webkitpy/common/system/platforminfo.py:117
> +                if output == 'arm64e':

This will never be true since output is assigned to arm64 in above line when it's arm64e.

[Jonathan Bedard]

Created attachment 402901 [details]
Patch

[Aakash Jain]

Comment on attachment 402901 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=402901&action=review

> Tools/ChangeLog:3
> +        [webkitpy] Support Apple Silicon (Part 1)

Bug title can be little more descriptive. Like: [webkitpy] Automatically detect hw architecture for supporting Apple Silicon

> Tools/Scripts/webkitpy/common/system/platforminfo.py:111
> +        try:

can we avoid running 'uname -m' command when is_mac() is false and have an early return instead?

[Jonathan Bedard]

(In reply to Aakash Jain from comment #6)
> Comment on attachment 402901 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=402901&action=review
> 
> > Tools/ChangeLog:3
> > +        [webkitpy] Support Apple Silicon (Part 1)
> 
> Bug title can be little more descriptive. Like: [webkitpy] Automatically
> detect hw architecture for supporting Apple Silicon
> 
> > Tools/Scripts/webkitpy/common/system/platforminfo.py:111
> > +        try:
> 
> can we avoid running 'uname -m' command when is_mac() is false and have an
> early return instead?

I don't think we should.

We have multiple watchOS architectures, the same logic should apply. That's actually the whole reason I didn't use the os version of that function in the first place.

[Jonathan Bedard]

Created attachment 402903 [details]
Patch

[Aakash Jain]

Comment on attachment 402901 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=402901&action=review

> Tools/Scripts/webkitpy/common/system/platforminfo_mock.py:43
> +        ).get(self.os_name, 'x86')

should the default be x86 or x86_64?

> Tools/Scripts/webkitpy/port/mac.py:72
> +            return 'arm64'

might be worth adding a comment here for the reason of doing this.

[Aakash Jain]

rs=me

[Jonathan Bedard]

Created attachment 402905 [details]
Patch for landing

[EWS]

Committed r263592: <https://trac.webkit.org/changeset/263592>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 402905 [details].

[Fujii Hironori]

Comment on attachment 402905 [details]
Patch for landing

View in context: https://bugs.webkit.org/attachment.cgi?id=402905&action=review

> Tools/Scripts/webkitpy/port/base.py:-116
> -            self.set_option('architecture', self.DEFAULT_ARCHITECTURE)

Removing this code causes a trouble for Windows ports.
Filed : Bug 213688 – [Win] run-webkit-tests is failing to run DRT and WTR without --architecture x86_64