Bug 232134 - DisplayList::Recorder's clipBounds() becomes empty if a flip is applied to the CTM
Summary: DisplayList::Recorder's clipBounds() becomes empty if a flip is applied to th...
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: New Bugs (show other bugs)
Version: WebKit Nightly Build
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: Tim Horton
URL:
Keywords: InRadar
Depends on:
Blocks:
 
Reported: 2021-10-22 02:10 PDT by Tim Horton
Modified: 2021-10-26 12:43 PDT (History)
6 users (show)

See Also:


Attachments
Patch (10.38 KB, patch)
2021-10-22 02:11 PDT, Tim Horton
ews-feeder: commit-queue-
Details | Formatted Diff | Diff
Patch (9.86 KB, patch)
2021-10-22 02:33 PDT, Tim Horton
no flags Details | Formatted Diff | Diff
Patch (7.54 KB, patch)
2021-10-22 16:18 PDT, Tim Horton
no flags Details | Formatted Diff | Diff
Patch (11.09 KB, patch)
2021-10-22 16:34 PDT, Tim Horton
no flags Details | Formatted Diff | Diff
Patch (11.04 KB, patch)
2021-10-26 11:54 PDT, Tim Horton
no flags Details | Formatted Diff | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Tim Horton 2021-10-22 02:10:02 PDT
DisplayList::Recorder's clipBounds() becomes empty if a flip is applied to the CTM
Comment 1 Tim Horton 2021-10-22 02:11:32 PDT
Created attachment 442135 [details]
Patch
Comment 2 Tim Horton 2021-10-22 02:33:03 PDT
Created attachment 442138 [details]
Patch
Comment 3 Tim Horton 2021-10-22 16:18:19 PDT
Created attachment 442215 [details]
Patch
Comment 4 Tim Horton 2021-10-22 16:33:52 PDT
oops, lost the test
Comment 5 Tim Horton 2021-10-22 16:34:10 PDT
Created attachment 442220 [details]
Patch
Comment 6 Tim Horton 2021-10-22 17:06:33 PDT
smfr asked for a layout test instead of an API test, but I've not found a way to get to the raw GraphicsContext::scale() from web content (the CSS 2D transform `scale(-1)` uses setCTM/concatCTM, which is not affected because mapping the rect through the transform maintains its positive size) in a way that it stays long enough to then introspect the clipBounds :)

Anyway, the API test is kind of nice because it lets us check CG vs. DL directly.
Comment 7 Darin Adler 2021-10-23 22:35:58 PDT
Comment on attachment 442220 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=442220&action=review

> Source/WebCore/platform/graphics/displaylists/DisplayListRecorder.cpp:449
> +    if (std::optional<AffineTransform> inverse = currentState().ctm.inverse())

I would’ve used auto here; writing out the type makes this code harder to read, I think
Comment 8 Tim Horton 2021-10-26 11:54:11 PDT
Created attachment 442514 [details]
Patch
Comment 9 EWS 2021-10-26 12:42:50 PDT
Committed r284888 (243565@main): <https://commits.webkit.org/243565@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 442514 [details].
Comment 10 Radar WebKit Bug Importer 2021-10-26 12:43:18 PDT
<rdar://problem/84677208>