r61484 changed the behaviour of setFocusedNode by calling setInputMethodState before the focused node is changed, something that Qt depended on and needs to depend on.
Wouldn't it have been better if the patch had introduced a method to finish ongoing composition session? or make setInputMethodState take an enum instead?
EFL is affected as well according to demarchi.
This is impacting the EFL implementation as well. In setInputMethodState we need the new focused node, not the old, in order to get the input method hints.
(In reply to comment #1) > Wouldn't it have been better if the patch had introduced a method to finish ongoing composition session? or make setInputMethodState take an enum instead? Sounds better. Why way would you prefer? a new enum or an new method?
Created attachment 61852 [details] A patch to fix this issue.
Attachment 61852 [details] did not build on qt: Build output: http://webkit-commit-queue.appspot.com/results/3370531
Attachment 61852 [details] did not build on gtk: Build output: http://webkit-commit-queue.appspot.com/results/3581076
Attachment 61852 [details] did not build on chromium: Build output: http://webkit-commit-queue.appspot.com/results/3594045
Created attachment 61854 [details] Fix two typos.
Can we make a regression test for this?
Comment on attachment 61854 [details] Fix two typos. Clearing flags on attachment: 61854 Committed r63602: <http://trac.webkit.org/changeset/63602>
All reviewed patches have been landed. Closing bug.