Closed Bug 95661 Opened 23 years ago Closed 23 years ago

Page with Java hangs / freezes entire OS

Categories

(Core Graveyard :: Java: OJI, defect, P3)

Product:
Core Graveyard
Component:
Java: OJI
Platform:
x86
Windows 98
Type:
defect

Tracking

(Not tracked)

Status:
RESOLVED WORKSFORME
Milestone:
mozilla0.9.8

People

(Reporter: wd, Assigned: joe.chou)

References

(
URL
)

Details

Attachments

(2 files)

Java console output when opening the page www.arkpc.com
4.89 KB, text/plain
Details
original class files + decompiled code (jbuilder 5 project file)
26.02 KB, compressed/zip
Details
BuildID: 2001081504 The URL listed above will hang the entire OS. (WinMe) Reproducible: Always Steps to Reproduce: 1.Go to: http://www.smartmoney.com/eqsnaps/index.cfm?story=ratings&symbol=nvda&fk=7756 2. 3. Actual Results: OS Hangs. Mouse still moves, but nothing else works Ctrl-Alt-Del hit twice gives BSOD Expected Results: No hang, no crash The page has at least 3 java applets on it. Once the one in the top left corner starts to show the contents, the whole OS hangs.
I see it too, on W98SE, JRE 1.2.1. Boiling down a test case currently, and will comment again shortly.
Well, this WORKSFORME with Sun's JDK1.3.1 on WinNT with Tuesday of this week's trunk. I'll ask QA to try it on Win98 or WinME.
Status: NEW → RESOLVED
Closed: 23 years ago
Resolution: --- → WORKSFORME
Yeah, I hadn't quite had tha chance to come back yet. YEah, it doesn't crash on NT, 2k, nor XP. Reopening, it's a JRE/Win9x problem.
Status: RESOLVED → REOPENED
OS: Windows ME → Windows 98
Resolution: WORKSFORME → ---
QA Contact: shrir → pmac
Hi Grey, A want to get our QA to look at it as well, so let's mark it WORKSFORME until they do. I've asked them to look at it and post debug information.
Status: REOPENED → RESOLVED
Closed: 23 years ago → 23 years ago
Resolution: --- → WORKSFORME
But it's _not_ WFM. Your'e using abuild 2 days old, on the wrong OS for the bug.
I can reproduce this bug on Win98 with 0.9.3 build, java 1.3.1
Just reproduced this on my home Win98SE machine with 2001081603, too. Looking further into it, the OS isn't *entirely* frozen. I happened to be playing an MP3 at the time and altough I couldn't interact with the OS at all, the MP3 continued to play through until the end. If I hit ctrl-alt-del, nothing happens except about 10 seconds later the task manager window comes up. At which point I am able to kill Mozilla and the OS returns to normal. The last thing displayed in the taskbar before the hang is: Applet AnalystRec2 started ed, I would think this should be re-opened since it's been reproduced on 4 different Win9x machines so far...
Sorry for the spam, I'm thinking it's possible that the above page is causing a very large / very quick drop in User or GDI resources? Once those resources get below a certain level I've seen Win9x behave in a similar way. If the resources drop by a large amount very rapidly, Win9x may not have time to give you the warning it usually does. (The WinNT line does not have the 64k "resource" limitation that 9X does...) So, in reproducing this bug, the amount of available resources available before attempting to open the page may have a definate effect on it.
Tried today with jdk1.3.1 builds 0.9.2 and 0.9.3 under Win98 and WinME. But mozilla do not hang/freez(in all four cases). Applets rendered and work. OS also isn't freezes. Realyy I don't know reasons for hungup. So leave it WORKSFORME. Please attach additional information.
Hey WD, you're on to something. How could I spuriously consume more resources before launching mozilla?
Status: RESOLVED → REOPENED
Resolution: WORKSFORME → ---
Ed: First, load the Resource Meter. Under Accessories -> System Tools on my PC. It may not be installed by default with Windows. This will give you a meter of your available resources in the system tray. As you load various different applications, you can watch the meter drop. If it drops to red, you're in danger of hanging the OS. Perhaps try reproducing this bug when the meter is in the lower-green or yellow range?
A godo way to suck up resources is load Netscape 4 on a form control intesive page. Slashdot is good if you have moderator access, just open a story with a lot of comments. Another way is to load one of the Medium size images at http://www.javasux.com, but be warned, it'll REALLY bog things down.
http://www.arkpc.com/ is another testcase for this bug.
OK. I see this for a while also. But the OS does not hangs - just press CTRL-ALT-DEL and wait some time - the close programm dialog appears amd you can kill the mozilla. But sometimes the java does not closes and only reboot helps.
My WinMe computer at work does seem to totally hang - I can wait 20 minutes after hitting Ctrl-Alt-Del on that system and nothing happens. An additional ctrl-alt-del reboots the machine. But on my Win98SE home PC I do see the same behavior as Eugene Describes. AFter about 10-15 seconds after the 3-finger salute I do get a task manager where I can kill mozilla.
workforme on windows 98 (branch build: 2001-09-07-05-0.9.4).
I get the hang as well. Even the mouse hangs. 2nd Ctrl+Alt+Del reboots. I have toggle keys installed which does beep, but the keyboard lights don't. OS: Windows 95 B Mozilla: Build ID: 2001091003 JRE: 1.3.0_01
Tests using java <options> sun.applet.AppletViewer <url> No options: Locked up computer as before. -verbose: Locked up but able to kill using Ctrl+Alt+Del. -verbose -Xint: Worked -verbose -Xmixed: Worked. Strange as -Xmixed is supposed to be default!
Mozilla Hangs here too ( 2001091103 win2K ) but OS still responds tried both JRE 1.3.2 and 1.4b2
tested page : http://www.arkpc.com/
http://www.smartmoney.com/eqsnaps/index.cfm?story=ratings&symbol=nvda&fk=7756 works though in Win2k. maybe these are two different bus?
I opened http://www.arkpc.com/ with the console open this time and it gave a unsptoppable list of nullpointer acceptions and security violations, so this is just a bad applet, that consumes all processor power with it's errors. IE doesn't cause this error though unless you set security to maximum. http://www.smartmoney.com/eqsnaps/index.cfm?story=ratings&symbol=nvda&fk=7756 Also gives security errors, but doesn't get in a loop. I've attached the console output for arkpc to this bug I'll decompile the applet on www.arkpc.com and see if I can find out which exact statement cases this. and report back.
I decompiled the classes on www.arkpc.com and looked at the source and then tried to compile it using J2EE 1.3 that came with Jbuilder 5 Enterprise. it gave an error on line 393. Fixed this and ran the class. It has about many uncaught nullpointers scattered all throughout the class while attempting to load images. This is an example of a very badly written piece of JAVA code. I'll attach the decompiled class files, as well as the compiled classes, so that people can have a look.
Severity: critical → normal
Priority: -- → P3
Target Milestone: --- → mozilla1.0
*** Bug 97072 has been marked as a duplicate of this bug. ***
Based on bug 90815 and 95867 platform and os should be changed to all (these bugs are about the same applet as on www.arkpc.com and they confirm this bug on Linux too).
Bug 90815 is LDAP, totally unrelated. Not a dupe. Bug 95867 is Linux only, and doesn't crash on Win32 at all. Even if it did, this isn't a dupe, as this bug locks up the whole OS in many instances. Leaving platform and OS alone.
typo on the first bug (I'll look up the correct number), but the second bug 95867 uses EXACTLY the same applet as this one (www.arkpc.com), and causes EXACTLY the same behaviour it just get's handled different on different OS'es it seems...
After installing the latest jre 1.3.1_01 Early Edition from the following url: http://developer.java.sun.com/developer/earlyAccess/j2sdk131_01a/ this problem goes away for me in Win95, win98 and win2k Can anyone confirm this?
OK, with the JRE at the above URL and 2001091903, WindowsMe no longer hangs at: http://www.smartmoney.com/eqsnaps/index.cfm?story=ratings&symbol=nvda&fk=7756 Which is great!! But, the page itself still does not work quite right. Of the 3 java applets on that page I have never gotten them all to load up. Sometimes none of them get past the "Loading Java Applet..." part. Sometimes the bottom one does load, but so far never the top 2. When I attempt to reload the page I get talkbacks: TB35596023Y TB35596106G TB35596196H
Now that Windows no longer hangs, I am able to check the Java console, and I found this: groupToSetScale=1 load: class AnalystRec2.class not found. java.lang.ClassNotFoundException: java.io.FileNotFoundException: File not found: http://www.smartmoney.com/eqsnaps/ø¸/AnalystRec2/class.class Killing mozilla.exe and going to the exact same URL again a couple of times gives these sort of lines in the console: load: class AnalystRec2.class not found. java.lang.ClassNotFoundException: java.io.FileNotFoundException: File not found: http://www.smartmoney.com/eqsnaps/PäG0äG/smc/AnalystRec2/class.class load: class SnapshotComparison.class not found. java.lang.ClassNotFoundException: java.io.FileNotFoundException: File not found: http://www.smartmoney.com/eqsnaps/°H`HHÀH/SnapshotComparison/class.class load: class AnalystRec2.class not found. java.lang.ClassNotFoundException: java.io.FileNotFoundException: File not found: http://www.smartmoney.com/eqsnaps/ E??E?/smc/AnalystRec2/class.class What's up with the strange characters in the URL for the class??
Ressign to Joe Chou, as I am no longer working officially on OJI.
Assignee: edburns → joe.chou
Status: REOPENED → NEW
Target Milestone: mozilla1.0 → mozilla0.9.6
FWIW, JRE 1.4.0 Beta 2 doesn't hang the OS nor does it cause mozilla to crash in the URL for this bug.
Re-assign to laa.
Assignee: joe.chou → laa
Target Milestone: mozilla0.9.6 → mozilla0.9.8
Works for me with mozilla 0.9.4 & netscape 6.2.1 builds, jre 1.3.1_01 & 1.4.0 Reporter, can you still reproduce the problem?
Yes, I can reproduce the problem with 1.3x JRE 1.4x works fine, though.
Which mozilla build do you use? Can you try 6.2?
I don't have any Win9x machine to test this on anymore, and I have not seen such a hang in a long time. Unless somebody speaks up who can still see a problem here, I suggest this should be marked WFM.
Just checked with win95 and 98 WFM (wihth the JRE 1.40 FINAL)
Ok, marking WFM per WD & Jesse.
Status: NEW → RESOLVED
Closed: 23 years ago → 23 years ago
Resolution: --- → WORKSFORME
Chris Petersen is a new QA contact for oji component. His email is: petersen@netscape.com
Assignee: laa → petersen
fixing small error for pmac@netscape.com (filter with : SPAMMAILSUCKS)
Assignee: petersen → joe.chou
QA Contact: pmac → petersen
Product: Core → Core Graveyard
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: