GNOME Bugzilla – Bug 306238
Screen reader stops reading from the second character of second line in evolution compose
Last modified: 2005-07-06 10:36:05 UTC
1.Start Gnopernicus with Screen reader and Speech. 2.Start evolution with accessibility enabled and compose a mail. 3. type in some words in the first line. Then start a new line and continue typing. Observation and Bug : 1.Screen reader reads all key press events in first line. 2.Second line, after the first character, Screen reader stops reading characters typed.
Created attachment 47130 [details] [review] Here is the patch for the bug.
adding patch keyword
Mengjie: some review comments on the patch - - in html_engine_delete, (e->cursor->offset-len) would become negative when we: - select some text from the beginning of a line to a certain extent and delete it. - keep the cursor at the beginning of a line and delete single characters. So, do we really intend/need to pass this negative value to the signal handler? - in the changelog entry, you have also included insert_empty_paragraph. we should remove that. - insert_object_for_undo changes are fine. Thanks.
the fix is wrong as we signal cursor position and not cursor offset. the position is absolute position in the whole document. (the offset is offset in the cursor object). you can use html_cursor_jump_to_position_no_spell to get the offset (remember to use tmp cursor so that you don't modify engine's cursor)
Created attachment 48081 [details] [review] this is the src part
Created attachment 48082 [details] [review] this is the a11y part
Dear all This is the new patch against this bug. As to your proposal, I signal the cursor position and get the offset of the text inserted or deleted in the a11y part. Because gnopernicus need to check the text deleted before it is really deleted, I emit the text_changed::delete signal before the deleing really happens.
Hi Mengjie, we cannot emit the signal before deletion. It is called object_deleted after all and also evo use that signal and depends on it. Why do you need it signaled before?
Hi, Rodo Firstly, actually I defined the object_deleted signal and I am the only person who use this signal. Secondly, I need it signaled before deletion because I need to access the deleted text in a11y part. I'v check the source code of gtkentry. It does emit corresponding signal before deleting. In addtion, gedit has the same behavior.
Rodo: Can we commit the patch. Thanks.
Hi, ksh I'v talked with rodo. He had agreed my signaling object_delete before really deleting and I'v changed the name of the signal to object_delete as his proposal. He also said if you think the other part of the patch is ok, I could commit it. Here is the updated patch.
Created attachment 48709 [details] [review] this is the src part.
Created attachment 48710 [details] [review] this is the a11y part.
Please commit to Head. Thanks for your work.
Created attachment 48711 [details] [review] this is the a11y part.
Patch committed. Thanks for the review.