WebKit Bugzilla
New
Browse
Log In
×
Sign in with GitHub
or
Remember my login
Create Account
·
Forgot Password
Forgotten password account recovery
RESOLVED FIXED
230085
[iOS] Ensure the CA transaction has been synchronously committed before snapshotting it
https://bugs.webkit.org/show_bug.cgi?id=230085
Summary
[iOS] Ensure the CA transaction has been synchronously committed before snaps...
Said Abou-Hallawa
Reported
2021-09-08 23:52:01 PDT
Committing the CA transaction may be an asynchronous operation.
Attachments
Patch
(1.76 KB, patch)
2021-09-09 00:00 PDT
,
Said Abou-Hallawa
no flags
Details
Formatted Diff
Diff
Patch
(6.60 KB, patch)
2021-09-09 23:16 PDT
,
Said Abou-Hallawa
no flags
Details
Formatted Diff
Diff
Patch
(7.00 KB, patch)
2021-09-13 18:33 PDT
,
Said Abou-Hallawa
no flags
Details
Formatted Diff
Diff
Patch
(6.97 KB, patch)
2021-10-18 18:03 PDT
,
Said Abou-Hallawa
thorton
: review+
Details
Formatted Diff
Diff
Patch
(6.95 KB, patch)
2021-10-19 18:37 PDT
,
Said Abou-Hallawa
no flags
Details
Formatted Diff
Diff
Show Obsolete
(3)
View All
Add attachment
proposed patch, testcase, etc.
Said Abou-Hallawa
Comment 1
2021-09-08 23:58:23 PDT
<
rdar://81800118
>
Said Abou-Hallawa
Comment 2
2021-09-09 00:00:51 PDT
Created
attachment 437714
[details]
Patch
Tim Horton
Comment 3
2021-09-09 00:11:08 PDT
Comment on
attachment 437714
[details]
Patch This seems like wild overkill for 99.9% of snapshotting use (which is snapshotting the currently displayed content). Is it really ok to -[CATransaction synchronize] every navigation, for example? (When we take back/forward snapshots)
Tim Horton
Comment 4
2021-09-09 00:19:18 PDT
Comment on
attachment 437714
[details]
Patch View in context:
https://bugs.webkit.org/attachment.cgi?id=437714&action=review
> Source/WebKit/UIProcess/API/ios/WKWebViewIOS.mm:3208 > + [CATransaction synchronize];
At a /minimum/, it seems like you need to not do this if WKSnapshotConfiguration.afterScreenUpdates=NO (which currently isn't plumbed in here but could be, and is one way to avoid it in the swipe snapshot case)
Said Abou-Hallawa
Comment 5
2021-09-09 23:16:50 PDT
Created
attachment 437838
[details]
Patch
Alexey Proskuryakov
Comment 6
2021-09-12 14:07:26 PDT
Is this likely to fix any flaky tests? :)
Wenson Hsieh
Comment 7
2021-09-13 13:22:42 PDT
Comment on
attachment 437838
[details]
Patch View in context:
https://bugs.webkit.org/attachment.cgi?id=437838&action=review
> Source/WebKit/UIProcess/API/Cocoa/WKWebViewPrivate.h:493 > +- (void)_snapshotRectWithConfiguration:(WKSnapshotConfiguration *)snapshotConfiguration rectInViewCoordinates:(CGRect)rectInViewCoordinates intoImageOfWidth:(CGFloat)imageWidth completionHandler:(void(^)(CGImageRef))completionHandler;
Note that Mail on iOS still uses this method, so we'll need to keep the old SPI around for source/binary compatibility. (The new method should include a `WK_API_AVAILABLE(…)`, as well)
Simon Fraser (smfr)
Comment 8
2021-09-13 13:28:56 PDT
Comment on
attachment 437838
[details]
Patch r- because it changes SPI
Said Abou-Hallawa
Comment 9
2021-09-13 18:33:33 PDT
Created
attachment 438096
[details]
Patch
Said Abou-Hallawa
Comment 10
2021-10-18 18:03:48 PDT
Created
attachment 441666
[details]
Patch
Tim Horton
Comment 11
2021-10-19 15:23:03 PDT
Comment on
attachment 441666
[details]
Patch View in context:
https://bugs.webkit.org/attachment.cgi?id=441666&action=review
> Source/WebKit/UIProcess/API/Cocoa/WKWebViewPrivate.h:493 > +- (void)_snapshotRectWithAfterScreenUpdates:(BOOL)afterScreenUpdates rectInViewCoordinates:(CGRect)rectInViewCoordinates intoImageOfWidth:(CGFloat)imageWidth completionHandler:(void(^)(CGImageRef))completionHandler WK_API_AVAILABLE(ios(WK_IOS_TBA));
drop the "with"
Said Abou-Hallawa
Comment 12
2021-10-19 18:37:06 PDT
Created
attachment 441832
[details]
Patch
EWS
Comment 13
2021-10-20 10:30:38 PDT
Committed
r284547
(
243289@main
): <
https://commits.webkit.org/243289@main
> All reviewed patches have been landed. Closing bug and clearing flags on
attachment 441832
[details]
.
Note
You need to
log in
before you can comment on or make changes to this bug.
Top of Page
Format For Printing
XML
Clone This Bug