46629 - Crash while entering text into a form's textarea

Status: VERIFIED FIXED

(Core :: DOM: Editor, defect, P3)

Description

[Bob Kanefsky]

Build id: 2000072408
Reproducible:  No

While I was entering comments on bug 20092 as Bugzilla requested, MacsBug 
suddenly appeared.  I had pasted in some text and was (I think) typing.  The 
stack trace suggests it was wrapping some text around in response to a keystroke, 
probably the Delete key, but I don't remember the last key I hit, or where the 
insertion point was, and I forgot to flip back to see the screen before exiting.  
I did, however, write down the "how" and "sc" output:

PowerPC unampped memory exception at 61BA61B8

nsHTMLTextAreaElement::HandleDOMEEvent
nsGenericElement::HandleDOMEEvent
nsEventListenerManager::HandleEvent
nsTextEditorKeyListener::KeyPress
nsHTMLEditorLog::DeleteSelection
nsHTMLEditor::DeleteSelection
nsEditor::EndPlaceholderTransaction
nsEditor::EndUpdateViewBatch
PressShell::EndReflowBatching

Comment 1

[Bob Kanefsky]

I accidentally reproduced exactly the same crash on the same machine (a G3).  I 
was unable to reconstruct it on another model of Mac (8600) with the same Mozilla 
and MacOS, despite getting the text I was inputting to line up identically.  This 
time I know that I hit the DELETE key with text highlighted.

I can actually take a photographic screen-shot of the screen in the state where 
it crashed, showing the exact text flow and selection, if I bring my digital 
camera to work.  Would that help?

Meanwhile, I've typed in the exact four lines of text that were visible on the 
screen (one blank line) and reconstructed a first line that makes them flow exact 
the same way.  I'll post it as a text attachment, with hard newlines representing 
where it wrapped.  I'll describe the state I can't show in the text file:  The 
second line, beginning with "another", was at the top of the visible textarea, 
with a few pixels scrolled off.  The selection extended from the just to the 
right of "report." to the end; i.e., it consisted of "<space><space>(I composed 
it elsewhere <wrap>because I didn't trust it not to cras".

I know it seems to appropriate too be true that it crashed when it did, but it 
was a total coincidence -- or possibly my subconscious chose the best time to hit 
the wrong key and crash it.  As Dave Barry always says, "I'm not making this up".  
As the Arrogant Worms say in "A Japanese Love Song":

   She was the one I'd kiss and hug -- 'til she was crushed like a bug.
   I guess it's hard to see a monster above you.
   She said she'd save her heart for me, but now it's there for all to see.
   And her last words to me were "I AAAAAAAAAAAAAA"

Comment 2

[Bob Kanefsky]

Attachment: Text reproduction of the exact text flow at the time of the crash

Comment 3

[rubydoo123]

asking Sujay to help with a concise, reproducible testcase

Comment 4

[Eli Goldberg]

Kanef, what are you doing in our bug tracking system? ;)

Comment 5

[Eli Goldberg]

CC:ing simon in case the stack trace looks familiar. (I don't see any bugs with 
this in a bugzilla query.)

Comment 6

[Jeffrey Baker]

Dupe of Bug 46626?  Crashing in the textarea when hittin backspace...

Comment 7

[rubydoo123]

we tried this out on mac, and win98 and can't repro this with today's build -- 
yesterday's build was bad for the editor. I am closing this out as fixed do to 
the backing out of the checkin for 46209

Comment 8

[rubydoo123]

*** Bug 46626 has been marked as a duplicate of this bug. ***

Comment 9

[Bob Kanefsky]

Before marking as Verified, be sure to check whether the changes you backed out 

were in fact part of build 2000072408, which is not "yesterday's build".



(What am I doing in your bug tracking system, Eli?  Testing support for Cascading 

Bug Reports, apparently.  This was three deep.)

Comment 10

[Bob Kanefsky]

Attachment: Complete stack trace of second crash

Comment 11

[Bob Kanefsky]

I can't reproduce the crash with the same build on the same machine and same
boot session, using the text I attached (reconstructed from the second crash)
and scroll position and selection I described.  Playing around with it does
reveal some minor bugs involving delete, Undo, and insertion point appearing
below the box, but I'm not going to formally report those at this time.  Maybe
someone else will do that.

Obviously, the fact that I can't even reproduce it myself means (1) it will be
hard to fix and (2) you can't be sure you fixed it.  Again, you should note
whether the backed-out changes were present in the July 24 build.

I did take a photographic screen shot, and will post it, probably tomorrow.

Comment 12

[Bob Kanefsky]

Attachment: Snapshot (and I do mean snapshot) of screen when crash occurred

Comment 13

[rubydoo123]

*** Bug 46626 has been marked as a duplicate of this bug. ***

Comment 14

[leger]

Adding crash keyword for fix and/or verification escalation

Comment 15

[sujay]

marking VERIFIED-FXIED. Bob, if you can reproduce, feel free to reopen this bug.