90962 - Crash in Cert Viewer

Status: VERIFIED DUPLICATE of bug 85451

(Core Graveyard :: Security: UI, defect, P1)

Description

[Bob Lord]

Talkback: TB32961255E

I crashed in the Cert Viewer.

Comment 1

[Stephane Saux]

-> ddrinan

Comment 2

[Stephen Walker]

1. Open the Cert Manager
2. Select ABA.ECOM Root CA from the Authorities Tab
3. Click the 'View' button
4. Click the 'Details' Tab
5. Select 'Certificate Signature Value' under Certificate Fields
6. Select some of the 'Field Values' for 'Certificate Signature Value'
7. Select 'Certificate Signature Algorithm' 

Comment 3

[Bob Lord]

Walker's script does indeed crash for me on Win2K.

Comment 4

[David P. Drinan]

Re-assigning to Javi.

Comment 5

[Stephane Saux]

crahs on linux branch 0716200104.0.9.2

Comment 6

[John Unruh]

Changing platform to all.

Comment 7

[Stephane Saux]

linux talkback is 32962260 and shows the same stack trace as on TB32961255E

Comment 8

[Javier Delgadillo]

Attachment: Patch to fix crash.

Comment 9

[jag (Peter Annema)]

This isn't the right fix for this. Basically this is fixing the symptom, not the
problem. Generally this crash is triggered by putting style="overflow:auto" on a
<tree/>, which causes it to be put in a nsIScrollableFrame, thus causing the QI
to nsITreeFrame to fail. Adding null checks here will subtly break code, not fix
it.

The real fix is to just not use style="overflow:auto" on a <tree/> (generally
used to get scrollbars). Instead, use the attribute rows="X" where X is the
number of rows you want visible (min and max) in the tree. Currently this has a
(small) bug though where the initial height will be 0 if you begin with an empty
tree (<tree/>) and fill in its children through JS later (bug 88138).

Comment 10

[David Hyatt]

Yes, why on earth is a tree being given a style of overflow: auto?  That is
nonsensical and meaningless for a tree widget.  I don't mind bulletproofing
crashes of course, but your tree is going to massively malfunction as long as
you've artificially imposed this extra scrolling frame between the tree's parent
and the tree itself.

In fact, we should add assertions in addition to the returning of the failure
codes when the tree frame is null.

Comment 11

[Stephane Saux]

target->2.1

Comment 12

[Javier Delgadillo]

This seems like a dup of 85451

*** This bug has been marked as a duplicate of 85451 ***

Comment 13

[John Unruh]

Verified dupe, and the crash is gone, at least from the branch.

Comment 14

[lecky]

Attachment: bluetooth.zip