362724 - Crash [@ nsIView::Destroy] on print preview

Status: VERIFIED FIXED

(Core :: Layout, defect)

Description

[Martijn Wargers (dead)]

See upcoming testcases.
The unminimised testcase crashes on print preview. The minimised testcase crashes when closing print preview for me.
The unminimised testcase also crashes on reflow branch builds for me.
But the testcase don't crash on the latest 1.8 branch. It seems to be fixed somehow on the 1.8 branch between 2006-08-02 and 2006-08-03:
http://bonsai.mozilla.org/cvsquery.cgi?treeid=default&module=all&branch=MOZILLA_1_8_BRANCH&branchtype=match&dir=&file=&filetype=match&who=&whotype=match&sortby=Date&hours=2&date=explicit&mindate=2006-08-02+04&maxdate=2006-08-03+07&cvsroot=%2Fcvsroot
Maybe fixed somehow with the patch for bug 302911?

Comment 1

[Martijn Wargers (dead)]

Attachment: Minimised testcase

Comment 3

[Bernd]

This doesn't crash for me, can you produce a reflow log so that I can estimate what frames are involved.

Comment 4

[Martijn Wargers (dead)]

Attachment: logfile

This is the log from the moment when I opened up the unminimised testcase in print preview, until I closed print preview (in which I crash).
I hope it's useful.

Comment 5

[Bernd]

Martijn, is it possible to convert your test cases from pixel heights to page-break properties (http://www.w3.org/TR/REC-CSS2/page.html#page-breaks)? This would make them much more reliable.

Comment 6

[Bernd]

I get a couple of asserts with the unminimized testcase

###!!! ASSERTION: Allowed only one anonymous view between frames: 'ancestorView
== view->GetParent()->GetParent()', file d:/moz_src/mozilla/layout/generic/nsCon
tainerFrame.cpp, line 272
###!!! ASSERTION: Creating a circular frame list, this is very bad.: 'this != aN
extSibling', file d:\moz_src\mozilla\layout\generic\nsIFrame.h, line 677

could you please create minimized test cases for the asserts (bonus points if they work with the page-break properties ;-) )

Comment 7

[Martijn Wargers (dead)]

Attachment: Minimised testcase

(In reply to comment #5)
> Martijn, is it possible to convert your test cases from pixel heights to
> page-break properties (http://www.w3.org/TR/REC-CSS2/page.html#page-breaks)?
> This would make them much more reliable.

Ok, thanks, this allowed me to minimise the minimised testcase to this. I hope this one crashes for you when closing print preview, it does for me.

Comment 8

[Martijn Wargers (dead)]

Attachment: testcase2

Testcase that gives the "###!!! ASSERTION: Allowed only one anonymous view between frames: 'ancestorView== view->GetParent()->GetParent()'" assertion.
It uses page-break-after: always; ==> 1st bonus point for me.

Comment 9

[Martijn Wargers (dead)]

Attachment: testcase3

No bonus points for this one.
This gives the "###!!! ASSERTION: Creating a circular frame list, this is very bad.:" on print preview for me.
It isn't as minimised as far as I could, but still quite a bit complicated.

Comment 10

[Bernd]

cool that works now very smooth to get lizard down to its knees

Comment 11

[Bernd]

Attachment: patch

This is basically what you did 3.5 years ago for blocks
https://bugzilla.mozilla.org/show_bug.cgi?id=201767#c19

Comment 12

[Bernd]

The patch fixes the crash, but the assert from testcase 3 is something different

Comment 13

[Bernd]

Testcase 3 should be split into a separate probably public bug as it has no views involved.

Comment 14

[Martijn Wargers (dead)]

(In reply to comment #13)
> Testcase 3 should be split into a separate probably public bug as it has no
> views involved.

Ok, I filed bug 363120 for that.

Comment 15

[Bernd]

fix checked in, we need this probably on every branch that is maintained

Comment 16

[Daniel Veditz [:dveditz]]

Comment on attachment 247852 [details] [diff] [review]
patch

bug 362837

Comment 17

[Bernd]

fixed on branches on dec. 27th 2006

Comment 18

[Martijn Wargers (dead)]

Verified fixed on the latest 1.8.0.x and latest 1.8.1 branch builds, no crash anymore when closing print preview on the testcases.
The latest trunk builds hang on print preview on certain testcases, I think that's what bug 363120 is basically about.