Bug 87327 – REGRESSION (r118013-r118031): Loops/Reloads under www.yahoo.com, quits after three tries with error

• Component: JavaScriptCore
• Status: RESOLVED FIXED

[Joe Strzemp]

In order to reproduce the issue, simply visit www.yahoo.com and
the page reloads three times and then quits with:

    Safari can't display this webpage
Safari has repeatedly encountered an error while trying to display
    "http://www.yahoo.com"

This worked prior in nightly r117946, but fails in r118039 and
also fails in the latest (r118179) nightly build.


Side note:
There is an error that appears briefly in the console log on the
first page load, but there doesn't appear to be an obvious way to
capture that output since the console log is cleared when the
page reloads.

It would be nice if there was some way to capture/save that
output, stop the reload process on first error, or to see a
history of prior errors.

[Joe Strzemp]

The nightly r118336 still has the problem, but it may require
drilling down on www.yahoo.com instead of just the home page.
Seems most any story cause the reload problem.

[Alexey Proskuryakov]

*** Bug 87331 has been marked as a duplicate of this bug. ***

[Alexey Proskuryakov]

Further reduced to <http://trac.webkit.org/log/trunk/?rev=118031&stop_rev=118013>. But this appears resolved in latest nightly.

[Joe Strzemp]

As I noted in Comment #1, this is still broken on pages *under* yahoo.com
(even though it's not on problem on the yahoo home page).

Just checked, still broken on: r118394

Hers's an example that (still) shows the problem:

http://news.yahoo.com/blogs/sideshow/another-bear-falls-tree-colorado-being-tranquilized-photos-190859388.html

[Filip Pizlo]

I will look at this soon; right now I'm plodding through a couple of other bugs.

[Filip Pizlo]

Yup, I can repro in r118408

[Filip Pizlo]

Created attachment 143966 [details]
the patch

[Filip Pizlo]

Created attachment 143967 [details]
the patch

Removed unnecessary noise form the patch.

[Filip Pizlo]

Created attachment 143970 [details]
the patch

[Filip Pizlo]

Fascinating, I think I can see why I broke this code in the first place.  I need something more general because in inspector/elements/undo-dom-edits.html, we can have an abstract value corresponding to something other than a cell flowing into the CheckStructure.  In particular, we need to look at both m_structure and m_type.

[Filip Pizlo]

Created attachment 143976 [details]
the patch

[Geoffrey Garen]

Comment on attachment 143976 [details]
the patch

r=me

[Filip Pizlo]

Landed in http://trac.webkit.org/changeset/118476