Bug 222323 – Fix model test differences between platforms

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

[Sam Weinig]

Fix model test differences between platforms - see https://bugs.webkit.org/show_bug.cgi?id=222114#c7.

[Sam Weinig]

Created attachment 421334 [details]
Patch

[Sam Weinig]

Created attachment 421335 [details]
Patch

[Darin Adler]

Comment on attachment 421335 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=421335&action=review

> LayoutTests/ChangeLog:10
> +        Try to fix these on iOS by removing the doctype.

Wow, why does that matter?

[EWS]

Committed r273372: <https://commits.webkit.org/r273372>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 421335 [details].

[Radar WebKit Bug Importer]

<rdar://problem/74678133>

[Alexey Proskuryakov]

I'm also very interested in an answer to Darin's question.

[Sam Weinig]

Re-opening, because the test is now skipped.

[Sam Weinig]

Created attachment 421413 [details]
Patch

[Sam Weinig]

(In reply to Darin Adler from comment #3)
> Comment on attachment 421335 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=421335&action=review
> 
> > LayoutTests/ChangeLog:10
> > +        Try to fix these on iOS by removing the doctype.
> 
> Wow, why does that matter?

I'm not sure it does. Was putting the patch up to see. (I don't have a usable iOS currently to test it myself).

On slack, there was discussion about why iOS and macOS would have different render trees for this very simple test, and there wasn't a conclusive answer. Maybe something about rounding? I filed https://bugs.webkit.org/show_bug.cgi?id=222362 to track figuring it out.

[Sam Weinig]

Antti, Alan, Simon, did any of you have an explanation of why iOS and macOS are different in the end?

[zalan]

(In reply to Sam Weinig from comment #10)
> Antti, Alan, Simon, did any of you have an explanation of why iOS and macOS
> are different in the end?
I would need to debug this but I don't have a working iOS sim atm.

[zalan]

(In reply to Darin Adler from comment #3)
> Comment on attachment 421335 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=421335&action=review
> 
> > LayoutTests/ChangeLog:10
> > +        Try to fix these on iOS by removing the doctype.
> 
> Wow, why does that matter?
Removing the doctype puts the document in non-standards mode which makes the descent on the line collapse.
It means that the line is only stretched by the replaced box and since the box sits on the baseline and has no descent, the line is going to be as tall as the replaced box is. 
In standards mode the line is also stretched by the root inline box which has the  default font's descent. (Now I just need to figure out if we produce different geometries on iOS because of some odd rounding mismatching or font differences)

[EWS]

Committed r273421: <https://commits.webkit.org/r273421>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 421413 [details].