Closed Bug 57135 Opened 25 years ago Closed 25 years ago

Browser crashes on page http://www.defense-data.com

Categories

(Core Graveyard :: Plug-ins, defect, P3)

Product:
Core Graveyard
Component:
Plug-ins
Platform:
DEC
OpenVMS
Type:
defect

Tracking

(Not tracked)

Status:
RESOLVED FIXED

People

(Reporter: jakobus, Assigned: colin)

References

(
URL
)

Details

(Keywords: crash)

Attachments

(2 files)

output of nm libjavaplugin_oji.so
203.18 KB, text/plain
Details
patch
1.29 KB, patch
Details | Diff | Splinter Review
The URL: http://www.defense-data.com crashes with the report: Document http://www.mozilla.org/ loaded successfully ->>>>>>>>>>>>>> Write Clipboard to memory Document http://www.defense-data.com/ loaded successfully %SYSTEM-F-ACCVIO, access violation, reason mask=00, virtual address=000000000200048E, PC=00000000011E4C84, PS=0000001B %TRACE-F-TRACEBACK, symbolic stack dump follows image module routine line rel PC abs PC DECW$XTLIBSHRR5 0 000000000003EC84 00000000011E4C84 LIBGTKXTBIN GTKXTBIN gtk_xtbin_realize 28959 0000000000000A18 0000000001E1CA18
I can confirm this bug with 2000101608 on Win2k. Without Java the page works fine but with Java my mouse pointer is flickering and after a short while my mouse pointer is away and seconds later Mozilla seems to hang.
over to plugins, could be OJI though
Assignee: asa → av
Severity: normal → critical
Component: Browser-General → Plug-ins
Keywords: crash
QA Contact: doronr → shrir
This works fine on WinNT 4, NS6 release build id 2000101608. Over to George, likely an OJI problem.
Assignee: av → drapeau
URL: http://www.defense-data.com
Status: UNCONFIRMED → NEW
Component: Plug-ins → OJI
Ever confirmed: true
It works for me with commercial build on Solaris with JRE 1.3.0_01. Also works for me on NS6 PR 3 on Windows 2000. How the heck did you get a Java VM on Open VMS? I didn't know that there was an OJI-compliant JVM available for recent Mozilla builds. I also am sorry to say that I'm not familiar enough with VMS to understand the stack trace included with the bug report. It looks as if the browser crashed somewhere in GTK, not in Java or the OJI module. Can you put more information about the crash in here, please? Do you have a JVM associated with your browser? Can you successfully go to http://java.sun.com/ (which also has applets on it) and view the Industry News scrolling applet on that page? Wow, man: this is interesting.
The reporter does not talk about java. I have this problem with the page and installed java. I tried this page without java and I have a crash. I used aa trunk build. I try this page later with a new trunk build (report after the download)
o.k. i have load the page with the a new trunk build. Now I´m on Windows 98. Open the page http://www.defense-data.com After finish loading open one of the "Lead Articles". I get this Error-Message : java.lang.StringIndexOutOfBoundsException: String index out of range: -342 at java.lang.String.substring(Unknown Source) at PopupMenu.loadMenu(PopupMenu.java:307) at PopupMenu.loadData(PopupMenu.java:298) at PopupMenu.init(PopupMenu.java:132) at sun.applet.AppletPanel.run(Unknown Source) at java.lang.Thread.run(Unknown Source) I have no crash on this system but I try this on the win2k system, where mozilla hang.
This URL now crash not my the new Mozilla trunk build at my win2k system. But I get this Java error message at the linked page.
The Java error is not a Mozilla bug, nor does it appear to be a J2SE (Java 2, Standard Edition) bug. It looks like a bug in the applet itself; the applet developer should look at this behavior and see what's going wrong in the applet. Looks like there's no bug here, unless the original bug reporter can produce some more data.
George, are you going to mark this INVALID?
I checked the system and installed the latest patches for: OpenVMS V7.2-1 next I installed the JAVA 2 SDK v 1.2.2-1 with the Virtual Machine 1.2.2-Beta2, but the system still crashes. $ java -version java version "1.2.2-beta2" Fast VM (build J2SDK.v.1.2.2:08/23/2000-10:42, native threads, jit_122) $ $ $ show symbol mozilla MOZILLA == "@SYS$COMMON:[MOZILLA]MOZILLA" $ $ $ mozilla Starting mozilla-bin... Setting content window *** Pulling out the charset Loading page specified via openDialog in SetSecurityButton Document http://www.mozilla.org/ loaded successfully ->>>>>>>>>>>>>> Write Clipboard to memory Document http://www.defense-data.com/ loaded successfully X Toolkit Error: Couldn't find per display information %CXXL-F-TERMINATE, terminate() or unexpected() called %TRACE-F-TRACEBACK, symbolic stack dump follows image module routine line rel PC abs PC MOZILLA-BIN 0 00000000000335DC 00000000000435DC MOZILLA-BIN 0 000000000003568C 000000000004568C ----- above condition handler called with exception 03AB8204: %DWT-F-NOMSG, Message number 03AB8204 ----- end of exception message 0 FFFFFFFF80083C3C FFFFFFFF80083C3C DECW$XTLIBSHRR5 0 000000000003E2D8 00000000011F42D8 DECW$XTLIBSHRR5 0 000000000003DEF0 00000000011F3EF0 DECW$XTLIBSHRR5 0 000000000003E108 00000000011F4108 DECW$XTLIBSHRR5 0 000000000003D104 00000000011F3104 DECW$XTLIBSHRR5 0 000000000003ECBC 00000000011F4CBC LIBGTKXTBIN GTKXTBIN gtk_xtbin_realize 28959 0000000000000A18 0000000001E3CA18 LIBGTK GTKSIGNAL gtk_signal_real_emit 13215 0000000000002960 0000000001732A90 LIBGTK GTKSIGNAL gtk_signal_emit 12327 0000000000000B20 0000000001730C50 LIBGTK GTKWIDGET gtk_widget_realize 24920 0000000000003C0C 000000000176EF2C LIBGTK GTKWIDGET gtk_widget_map 24859 0000000000003878 000000000176EB98 LIBGTK GTKSIGNAL gtk_signal_real_emit 13215 0000000000002960 0000000001732A90 LIBGTK GTKSIGNAL gtk_signal_emit 12327 0000000000000B20 0000000001730C50 LIBGKPLUGIN NS4XPLUGININSTANCE SetWindow 79368 0000000000001564 0000000001DBD924 LIBGKPLUGIN NSPLUGINHOSTIMPL InstantiateEmbededPlugin 73415 000000000000B0A8 0000000001DC9798 LIBGKLAYOUT NSOBJECTFRAME InstantiatePlugin 66409 00000000000050D8 0000000002617CA8 LIBGKLAYOUT NSOBJECTFRAME Reflow 66229 00000000000042A8 0000000002616E78 LIBGKLAYOUT NSLINELAYOUT ReflowFrame 41839 0000000000001A30 000000000260EE70 LIBGKLAYOUT NSINLINEFRAME ReflowInlineFrame 36642 00000000000016A0 0000000002609AD0 cut, the full list doesn't fit in this window.
The page: http://java.sun.com/ crashes just as the reported URL.
I can't mark this bug as INVALID just yet. There's enough data here to suggest that it's a real problem. *My* problem? I don't know enough about OpenVMS! (nor do I have a machine running OpenVMS) Here's what I'm thinking, though, based on the last stack trace that Theo posted (thank you, Theo): it's looking to me as if the Java Plug-In being used is a 4.X style plug-in, not an OJI plug-in. Ed, check out the stack trace and see if you agree with me. Continuing, it looks as if what's happening is that the browser realizes it needs a plugin, is finding some sort of plugin that claims it can handle the MIME type for Java, and the browser is trying to hand over a parent window to the plugin instance. If this is the case, it *could* be that the browser is handing over a structure when a simple XID is expected, or vice versa. In other words, maybe this is the same bug we saw last week on Linux, wherein a pavlov fix ended up breaking compatibility with 4.x-style plugins? (see bug 57046 for details) One way to verify this is to create two different builds of Mozilla: one from, say, today, and another from, say October 10 or a bit earlier. One of them should work, the other should crash. If so, then I predict that this is the cause.
Unless you have a libjavaplugin_oji.so in your mozilla/dist/bin/plugins directory, and it has symbols that very closely match those found in this attachment <http://bugzilla.mozilla.org/showattachment.cgi?attach_id=18099>, we can't help you. If neither of these conditions are met, and you still want help, you need to tell us exactly who supplied your java plugin. Marking Invalid. Please reopen if you can give us the required information.
Status: NEW → RESOLVED
Closed: 25 years ago
Resolution: --- → INVALID
I've found in the release notes http://www.openvms.compaq.com/openvms/products/ips/mozilla_relnotes.html#notes Java Support in Mozilla Q: Why is there no Java support in Mozilla for OpenVMS? A: Mozilla implements Java using OJI (Open JVM Integration). OJI is part of Java Version 1.3, but V1.3 is not yet available on all platforms, including OpenVMS. As soon as Java Version 1.3 is available on OpenVMS, it will be supported in Mozilla. Advantages of using OJI are: Multiple Java Virtual Machines (JVMs) can be supported Upgrading the JVM used by Mozilla is very easy With previous browsers, you were bound to the JVM that had been embedded in the browser code. For more information, see http://www.mozilla.org/oji/
I'm reopening this bug and will assign to myself (I'm the OpenVMS Mozilla maintainer). I'll explain why once I've re-assigned.
Status: RESOLVED → REOPENED
Resolution: INVALID → ---
OpenVMS doesn't have Java 1.3_01 or whatever release is required for a working OJI. We will have, but we don't yet. So I don't expect this to work. But I don't expect it to crash either. Problem is, I'm not sure why its crashing and I'm going to need some help in tracking this one down. I'll attach a full trace (I've removed the line numbers because they don't correspond to reality). If I disable Java in the prefs it doesn't crash, but with Java enabled (which is the default), it crashes. Surely there's a more user friendly way of saying "java not supported"!! Colin.
Assignee: drapeau → colin
Status: REOPENED → NEW
Here's the full trace from the crash. Hopefully this will mean something to someone, could like I said, I need a hint to get me started if I'm going to be able to fix this one. image module routine LIBGTKXTBIN GTKXTBIN gtk_xtbin_realize LIBGTK GTKSIGNAL gtk_signal_real_emit LIBGTK GTKSIGNAL gtk_signal_emit LIBGTK GTKWIDGET gtk_widget_realize LIBGTK GTKWIDGET gtk_widget_map LIBGTK GTKSIGNAL gtk_signal_real_emit LIBGTK GTKSIGNAL gtk_signal_emit LIBGKPLUGIN NS4XPLUGININSTANCE SetWindow LIBGKPLUGIN NSPLUGINHOSTIMPL InstantiateEmbededPlugin LIBGKLAYOUT NSOBJECTFRAME InstantiatePlugin LIBGKLAYOUT NSOBJECTFRAME Reflow LIBGKLAYOUT NSLINELAYOUT ReflowFrame LIBGKLAYOUT NSBLOCKFRAME ReflowInlineFrame LIBGKLAYOUT NSBLOCKFRAME DoReflowInlineFrames LIBGKLAYOUT NSBLOCKFRAME DoReflowInlineFramesAuto LIBGKLAYOUT NSBLOCKFRAME ReflowInlineFrames LIBGKLAYOUT NSBLOCKFRAME ReflowLine LIBGKLAYOUT NSBLOCKFRAME ReflowDirtyLines LIBGKLAYOUT NSBLOCKFRAME Reflow LIBGKLAYOUT NSBLOCKREFLOWCONTEXT DoReflowBlock LIBGKLAYOUT NSBLOCKREFLOWCONTEXT ReflowBlock LIBGKLAYOUT NSBLOCKFRAME ReflowBlockFrame LIBGKLAYOUT NSBLOCKFRAME ReflowLine LIBGKLAYOUT NSBLOCKFRAME ReflowDirtyLines LIBGKLAYOUT NSBLOCKFRAME Reflow LIBGKLAYOUT NSCONTAINERFRAME ReflowChild LIBGKLAYOUT NSHTMLFRAME Reflow LIBGKLAYOUT NSBOXTOBLOCKADAPTOR Reflow LIBGKLAYOUT NSBOXTOBLOCKADAPTOR DoLayout LIBGKLAYOUT NSBOX Layout LIBGKLAYOUT NSSCROLLBOXFRAME DoLayout LIBGKLAYOUT NSBOX Layout LIBGKLAYOUT NSCONTAINERBOX LayoutChildAt LIBGKLAYOUT NSGFXSCROLLFRAME Layout LIBGKLAYOUT NSGFXSCROLLFRAME DoLayout LIBGKLAYOUT NSBOX Layout LIBGKLAYOUT NSBOXFRAME Reflow LIBGKLAYOUT NSGFXSCROLLFRAME Reflow LIBGKLAYOUT NSCONTAINERFRAME ReflowChild LIBGKLAYOUT NSVIEWPORTFRAME Reflow LIBGKLAYOUT NSHTMLREFLOWCOMMAND Dispatch LIBGKLAYOUT NSPRESSHELL ProcessReflowCommands LIBGKLAYOUT NSPRESSHELL HandleEvent LIBXPCOM PLEVENT PL_HandleEvent LIBXPCOM PLEVENT PL_ProcessPendingEvents LIBXPCOM NSEVENTQUEUE ProcessPendingEvents LIBWIDGET_GTK NSAPPSHELL our_gdk_io_invoke LIBGLIB GMAIN g_main_dispatch LIBGLIB GMAIN g_main_iterate LIBGLIB GMAIN g_main_run LIBGTK GTKMAIN gtk_main LIBWIDGET_GTK NSAPPSHELL Run MOZILLA-BIN NSAPPRUNNER main1 MOZILLA-BIN NSAPPRUNNER main MOZILLA-BIN NSAPPRUNNER __MAIN MOZILLA-BIN
Status: NEW → ASSIGNED
Ed, I am still seeing this problem with OpenVMS (in a trunk build of about a week ago and my M0.6 build). I have some debug output that may be relevant and it has your name all over it! InstantiateEmbededPlugin for application/x-java-vm Inside nsPluginHostImpl::FindStoppedPluginForURL... dlsym: looking for NSGetFactory in /dka0/work/m06dbg/plugins/libnullplugin.so dlsym: failed to find NSGetFactory in /dka0/work/m06dbg/plugins/libnullplugin.so debug: edburns ns4xPlugin::CreatePlugin debug: edburns ns4xPlugin::CreatePlugin: cleared callbacks dlsym: looking for NP_Shutdown in /dka0/work/m06dbg/plugins/libnullplugin.so dlsym: looking for NP_Initialize in /dka0/work/m06dbg/plugins/libnullplugin.so debug: edburns: ns4xPlugin::CreatePlugin: callbacks->newstream: 28EA7A0 Inside ns4xPluginInstance::Start(void)... Inside ns4xPluginInstance::SetWindow(4D78E68)... About to create new ws_info... About to create new xtbin of 400 X 100 from 5500018... About to show xtbin(4D6C208)... Next is the access violation. Does this mean anything to you? I don't have a Java plugin. I think its trying to use the null plugin since that claims to handle *, right? But I never get the "do you want to download the plugin" window pop up. Any info you can provide would be very much appreciated. Feel free to assign the bug back to me. I'm just looking for input, not a fix! Thanks, Colin.
Assignee: colin → edburns
Status: ASSIGNED → NEW
Adding myself as a CC
I found the problem. Now I just need to figure out how to fix it!
Assignee: edburns → colin
Status: NEW → ASSIGNED
Component: OJI → Plug-ins
The problem is caused by the call to XtRegisterDrawable at http://lxr.mozilla.org/seamonkey/source/widget/src/gtkxtbin/gtkxtbin.c#338 XtRegisterDrawable and XtUnregisterDrawable were new in X11R6, and OpenVMS only has X11R5. We have code in our "porting layer" which we thought was doing the R5 equivalent, but clearly it wasn't. What is the R5 equivalent in this case? Other places in Mozilla where R6 specific code is used uses an #ifdef so that its not compiled on non-R6 platforms. Is that what we should be doing here? Adding blizzard to the cc list in case he has any ideas for R5 equivalents.
Attached patch patch — — Splinter Review
Chris, Here's a patch to make this code work on pre-R6 systems. Please check in. Thanks, Colin.
r=pavlov
sr=blizzard
Looks ok to me, but are we sure that the mozilla supported compiler and preprocessor supports this >= operator on all platforms?
Every C preprocessor I know of handles all the usual C operators in #if. Nit: how about a space after the >=, just as there is one before it? /be
Space added after >= (twice). I'll check in once the tree opens.
Checked in. gtkxtbin.c rev 1.6.
Status: ASSIGNED → RESOLVED
Closed: 25 years ago → 25 years ago
Resolution: --- → FIXED
Theo, can you please verify since I do not have the required OS.
QA Contact: shrir → jakobus
It's ok now using Mozilla 0.7 for OpenVMS which has build ID: 2001010817
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: