issue209: regression: prevent_automatic_form_mode does not work

Priority: bug Status: resolved
msg455 (view) Author: dclemente Date: 2009-10-13.15:15:26
Hi, with latest Conkeror, browser_prevent_automatic_form_focus_mode(true); seems not to work. There are
still pages which move focus to a form field, for instance
  It does: document.f.sercxata.focus();
  Each time I reload it with r, the focus is moved to the first field.

  I tried changing browser_automatic_form_focus_window_duration to 0, 20 and 1000: same problem.
XULrunner on Debian GNU/Linux.

  Via bisection I found the culprit commit:

1e1a96da42d12b107dc7c6c0041a355f9763c7a1 is the first bad commit
commit 1e1a96da42d12b107dc7c6c0041a355f9763c7a1
Author: John Foerch <>
Date:   Sat Aug 29 23:59:47 2009 -0400

    move input-mode framework into buffer.js, eliminating circular dependency
    This commit fixes which was caused by
    a frameset document in which a form existed in one frame, whose target
    was another frame.  When the form was submitted, with focus still on the
    input-box, results would be loaded in this other frame.  That load would
    trigger a location-change, which would in turn trigger conkeror
    erroneously switching to normal-input-mode.  The fix was to only use the
    function content_buffer_update_input_mode_for_focus to determine the
    proper mode, and never calling normal_input_mode directly.
    added create_buffer_early_hook, fixes
    under certain conditions, it is possible for a document title to be
    changed before the buffer constructor has run to completion.  This causes
    tab_bar_update_buffer_title to get run before tab_bar_add_buffer has been
    run, so does not yet exist.
    This commit begins to establish the idea of modular input-mode systems.
    There are further changes that need to be made in order to complete that
    project, and many of them have been marked with explanatory XXX and
    FIXME notes.
msg456 (view) Author: retroj Date: 2009-10-13.17:12:49
thanks for the help narrowing down when the breakage happened.  I found the culprit lines that caused
the problem, and I'll work on a fix today.
msg457 (view) Author: retroj Date: 2009-10-13.17:29:14
pushed a fix.  basically I had just made a careless change in that patch of August 29, but now I have
put back code that I had thought unnecessary at the time, and I added expanatory comments.
msg459 (view) Author: dclemente Date: 2009-10-13.19:21:44
It works; thanks
Date User Action Args
2009-10-13 19:23:36dclementesetstatus: chatting -> resolved
2009-10-13 19:21:44dclementesetstatus: resolved -> chatting
messages: + msg459
2009-10-13 17:29:14retrojsetstatus: chatting -> resolved
messages: + msg457
2009-10-13 17:12:49retrojsetstatus: unread -> chatting
messages: + msg456
2009-10-13 15:15:26dclementecreate