147364 - on Back browser loads about:blank (Trunk M1RC3)

Status: RESOLVED WONTFIX

(Core :: DOM: Navigation, defect)

Description

[Adam Hauner]

Repro:
1. go to URL
2. click on any headline in small IFRAME (-> article will be opened) 
3. hit Back button or Alt + <-

Actual:
about:blank page appear.

Expect:
Previous page.

2002052508/trunk/W2K

Comment 1

[Boris 'pi' Piwinger]

I see it in Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.0.0+) Gecko/20020527.

pi

Comment 2

[Radha on family leave (not reading bugmail)]

This is probably the same problem as 148794.

Comment 3

[Radha on family leave (not reading bugmail)]

Just visiting the above site creates 4 entries in session history. This is
because the "about:blank" loaded in the iframes don't get registered in session
history. This is what is going on:

When the iframes are loaded with "about:blank",

1) nsDocShell::SetCurrentURI()  is called once from
CreateAboutBlankContentViewer() which sets mCurrentURI to "about:blank" .
2) The iframe then comes around for regular loading through
nsDocShell::CreateContentViewer() and since mCurrentURI is already set, in step
1)  nsDocShell::OnNewURI() decides that it is a repeat load of an existing
document and changes the loadType of the iframe to LOAD_NORMAL_REPLACE. So no
entries get created in session history for these iframes
3) Then when the onLoad Handler in the page, loads the actual url on the
iframes,  they get added to session history as top level documents. Thus simply
visiting the apge creates 4 entries in session history. (One for the main page
and 3 for the 3 iframes).  This ofcourse causes further trouble for back/forward
buttons when the user does subframe navigation in the site. 

looking to see if step 1) can be eliminated, which would allow proper creation
of the session history entries and therefore any subframe navigations...

Comment 4

[Radha on family leave (not reading bugmail)]

More comments about the patch that causes of misbehavior is in bug 104361.

Comment 5

[jag (Peter Annema)]

Dan, can you take a look at comment 3? It seems we overlooked something when we
fixed setting the URI for those "gotta show something" about:blank documents.

Comment 6

[Radha on family leave (not reading bugmail)]

When loading this page in the latest trunk builds, I see several JS warnings and
I don't see the iframe displayed at all. 

CSS Error (http://www.abicko.cz/css/clanek_plus.css :7.22): Unknown property 'sc
rollbar-arrow-color'.  Declaration dropped.
CSS Error (http://www.abicko.cz/css/clanek_plus.css :8.21): Unknown property 'sc
rollbar-face-color'.  Declaration dropped.
CSS Error (http://www.abicko.cz/css/clanek_plus.css :9.26): Unknown property 'sc
rollbar-highlight-color'.  Declaration dropped.
CSS Error (http://www.abicko.cz/css/clanek_plus.css :10.24): Unknown property 's
crollbar-3dlight-color'.  Declaration dropped.
CSS Error (http://www.abicko.cz/css/clanek_plus.css :11.23): Unknown property 's
crollbar-shadow-color'.  Declaration dropped.
CSS Error (http://www.abicko.cz/css/clanek_plus.css :12.27): Unknown property 's
crollbar-darkshadow-color'.  Declaration dropped.
CSS Error (http://www.abicko.cz/css/clanek_plus.css :13.22): Unknown property 's
crollbar-track-color'.  Declaration dropped.
JavaScript strict warning:
http://www.abicko.cz/pages/uvod.html line 33: assignment to undeclared variable
pxDepth

Comment 7

[jag (Peter Annema)]

Step 1 can be eliminated, patch coming up.

Comment 8

[jag (Peter Annema)]

Attachment: Don't use SetCurrentURI, set mCurrentURI directly

Comment 9

[Radha on family leave (not reading bugmail)]

Jag, your attachement doesn't quite help me  with this bug. The problem is not
in calling SetCurrentURI() or firing onLocationChange. The problem is setting
mCurrentURI to about:blank, which is again going to be set when about:blank
comes for loading asynchronously. However, now per my previous comment, I'm not
able to load the site properly. Let me see. 

If there are situations when about:blank is loaded only synchronously, I would
like  try them out. Let me know if you are aware of any such situations. 

Comment 10

[Radha on family leave (not reading bugmail)]

Docshell recently changed behavior so that page loads to iframes are not
recorded in session history. So, that changes the behavior of this website and I
intend to keep it that way, since the widespread expectation is that, iframe
loads should not go into session history.