Changeset 13441 in webkit


Ignore:
Timestamp:
Mar 22, 2006 2:30:39 PM (18 years ago)
Author:
tomernic
Message:

WebCore:

Reviewed by Kevin Decker.

Part of <rdar://problem/4351664> REGRESSION (420+): extra URL in b/f list - navigating back to previous page fails at apple.com/retail/)
This also fixes <rdar://problem/4477821> REGRESSION (10.4.5-TOT): meta tag specifying refresh is being added to history.

  • bridge/mac/FrameMac.h:
  • bridge/mac/FrameMac.mm: Removed redirectionTimerFired(). This was added as attempt to fix <​http://bugzilla.opendarwin.org/show_bug.cgi?id=7058>. The aim was to cause Safari and WebKit to update their loading status after a redirect. Unfortunately, the fix had a bad side effect. Calling -reportClientRedirectCancelled: on a successful redirect causes WebKit to forget that the redirect was supposed to lock history (i.e. reuse the current back/forward entry for the new page). The end result was that intermediate "quick" redirects were creating back/forward entries when they should not have been. See 4351664. That fix was almost correct, in that we do need to notify the frame load delegate when a redirect ends, either because it succeeded or because it was cancelled. However, this is the wrong place to do it. WebCore's redirect notification logic did not need to change to fix 7058. The never-ending spinning indicators problem was actually caused by a bug at the WebKit level.
  • manual-tests/redirectHistory: Added.
  • manual-tests/redirectHistory/redir-1.html: Added.
  • manual-tests/redirectHistory/redir-2.html: Added.
  • manual-tests/redirectHistory/redir-3.html: Added. Manual test case. I couldn't figure out how to create a layout test for this, because it involves navigation through history and it was unclear how/when to tell DumpRenderTree to dump its output.

WebKit:

Reviewed by Kevin Decker.

Better fix for <rdar://problem/4432562> REGRESSION (TOT): Safari's "stop loading" active, "view source" inactive after page load [7058]

  • WebView/WebFrame.m: (-[WebFrame _transitionToCommitted:]): Cancel the client redirect when we commit the provisional load, if we were waiting for a redirect. This is a better fix for 7058 (<rdar://problem/4432562>). The original fix for 7058 changed the timing of the redirect cancel in such a way that WebKit was precluded from ever reusing back/forward list entries for redirects. Clearing the redirect state here actually makes logical sense, as the redirect's target page is being committed at this point.
Location:
trunk
Files:
4 added
5 edited

Legend:

Unmodified
Added
Removed

  • trunk/WebCore/ChangeLog

    r13440 r13441  
     12006-03-22  Tim Omernick  <timo@apple.com>
     2
     3        Reviewed by Kevin Decker.
     4
     5        Part of <rdar://problem/4351664> REGRESSION (420+): extra URL in b/f list - navigating back to previous page fails at apple.com/retail/)
     6        This also fixes <rdar://problem/4477821> REGRESSION (10.4.5-TOT): meta tag specifying refresh is being added to history.
     7
     8        * bridge/mac/FrameMac.h:
     9        * bridge/mac/FrameMac.mm:
     10        Removed redirectionTimerFired().  This was added as attempt to fix <http://bugzilla.opendarwin.org/show_bug.cgi?id=7058>.  The
     11        aim was to cause Safari and WebKit to update their loading status after a redirect.  Unfortunately, the fix had a bad side
     12        effect.  Calling -reportClientRedirectCancelled: on a successful redirect causes WebKit to forget that the redirect was supposed
     13        to lock history (i.e. reuse the current back/forward entry for the new page).  The end result was that intermediate "quick" redirects
     14        were creating back/forward entries when they should not have been.  See 4351664.  That fix was almost correct, in that we do need to
     15        notify the frame load delegate when a redirect ends, either because it succeeded or because it was cancelled.  However, this is the
     16        wrong place to do it.  WebCore's redirect notification logic did not need to change to fix 7058.  The never-ending spinning indicators
     17        problem was actually caused by a bug at the WebKit level.
     18
     19        * manual-tests/redirectHistory: Added.
     20        * manual-tests/redirectHistory/redir-1.html: Added.
     21        * manual-tests/redirectHistory/redir-2.html: Added.
     22        * manual-tests/redirectHistory/redir-3.html: Added.
     23        Manual test case.  I couldn't figure out how to create a layout test for this, because it involves navigation through history and
     24        it was unclear how/when to tell DumpRenderTree to dump its output.
     25
    1262006-03-22  Eric Seidel  <eseidel@apple.com>
    227

  • trunk/WebCore/bridge/mac/FrameMac.h

    r13440 r13441  
    310310    virtual void startRedirectionTimer();
    311311    virtual void stopRedirectionTimer();
    312     virtual void redirectionTimerFired(Timer<Frame>*);
    313312
    314313private:

  • trunk/WebCore/bridge/mac/FrameMac.mm

    r13440 r13441  
    836836    if (wasActive && d->m_scheduledRedirection != historyNavigationScheduled)
    837837        [_bridge reportClientRedirectCancelled:d->m_cancelWithLoadInProgress];
    838 }
    839 
    840 void FrameMac::redirectionTimerFired(Timer<Frame>* timer)
    841 {
    842     // Note, despite its name, we must call "redirect cancelled" even when the
    843     // redirect has completed successfully. Although that may not have been our
    844     // original intent, Safari depends on this behavior at the time of this writing.
    845     // See Radar 4432562 and Bugzilla 7058.
    846 
    847     // Don't report history navigations, just actual redirection.
    848     if (d->m_scheduledRedirection != historyNavigationScheduled)
    849         [_bridge reportClientRedirectCancelled:d->m_cancelWithLoadInProgress];
    850 
    851     Frame::redirectionTimerFired(timer);
    852838}
    853839

  • trunk/WebKit/ChangeLog

    r13431 r13441  
     12006-03-22  Tim Omernick  <timo@apple.com>
     2
     3        Reviewed by Kevin Decker.
     4
     5        Better fix for <rdar://problem/4432562> REGRESSION (TOT): Safari's "stop loading" active, "view source" inactive after page load [7058]
     6
     7        * WebView/WebFrame.m:
     8        (-[WebFrame _transitionToCommitted:]):
     9        Cancel the client redirect when we commit the provisional load, if we were waiting for a redirect.
     10        This is a better fix for 7058 (<rdar://problem/4432562>).  The original fix for 7058 changed the timing of the redirect cancel
     11        in such a way that WebKit was precluded from ever reusing back/forward list entries for redirects.  Clearing the redirect state
     12        here actually makes logical sense, as the redirect's target page is being committed at this point.
     13
    1142006-03-21  Darin Adler  <darin@apple.com>
    215

  • trunk/WebKit/WebView/WebFrame.m

    r13269 r13441  
    779779                        // see -_opened
    780780                    }
     781
     782                    // This dataSource was a client redirect (we checked above, remember?).  Now that it is being
     783                    // committed, there can no longer be a pending redirect.  Call -_clientRedirectCancelled: here
     784                    // so that the frame load delegate is notified that the redirect's status has changed.  The
     785                    // frame load delegate may have saved some state about the redirect in its
     786                    // -webView:willPerformClientRedirectToURL:delay:fireDate:forFrame:.
     787                    [self _clientRedirectCancelled:NO];
    781788                }
    782789                [self _makeDocumentView];
Note: See TracChangeset for help on using the changeset viewer.