Gregory Carter Reporter
	Description • 24 years ago

The browser will crash if compiled on a glibc-2.2 system, and an attempt is made
to install and load the java plugin 1.3.  Any page loaded from this install will
crash if it requires an applet.

To correct, you must insert into the mozilla startup script the following line:

LD_ASSUME_KERNEL=2.2.5
export LD_ASSUME_KERNEL

This will correct and prevent the plugin from crashing.

-gc

	cls Assignee
	Comment 1 • 24 years ago

So I'm confused here.  How does telling the dynamic linker to assume linux
kernel 2.2.5 semantics (that's what I'm assuming it's doing) have anything to do
with glibc 2.2?  I guess I need to read up more on that particular flag.

	Doron Rosenberg (IBM)
	Comment 2 • 24 years ago

*** Bug 79809 has been marked as a duplicate of this bug. ***

	cls Assignee
	Comment 3 • 23 years ago

Blizzard, do you know anything about this?

	Christopher Blizzard (:blizzard)
	Comment 4 • 23 years ago

glibc changed to use Linux 2.4's new unlimited stack with threads.  It used to
be in the old 2.2 world that stacks were limited to 2 meg in size and the JVM
assumes that behaviour.  Using the environment variable in question tells glibc
to limit stacks to 2 meg in size so the JVM will work.

	ruth@innocent.com
	Comment 5 • 23 years ago

This should probably get a release note entry for 0.9.2

It would be nice to put a work-around using this environment variable in the
mozilla startup script for Linux/Unix (it's harmless on other Unices) but I
don't think there's time before 0.9.2 so how about a note (under Java):

Linux: If you use Linux kernel 2.4 and GNU libc 2.2 or if Java applets always
crash the browser, set the environment variable LD_ASSUME_KERNEL to 2.2.5
before starting the browser. e.g.
# export LD_ASSUME_KERNEL=2.2.5
# mozilla

	edburns
	Comment 6 • 23 years ago

Here's a datapoint from a Sun Engineer:

On 15 June 10:11:39, Michael Colburn - Sun Microsystems wrote:
> 
> On a RedHat 7.1 system using JRE 1.3.1 and mozilla 0.9.1 (also Netscape 6.01)
> I ran/loaded a bunch of demos and tests without a problem (for the most 
part). 
> I did have a hang and a crash once, but this is not unique to Linux. I have
> seen these occasional problems on other platforms when using Netscape 6. 
> The applets did seem to take a little longer to load when using mozilla 0.9.1
> compared to Netscape 6.01.
> 
> I was using the JRE 1.3.1 from our site. I do not know what version of 1.3
> is at the Netscape site http://home.netscape.com/plugins/jvm.html.
> 
> Sites, mentioned in the mozilla bugs, also loaded OK.
> http://frenchopen.org/en realtime scoreboard 
> http://www.bbc.co.uk/dinosaur 
> http://www.waitrose.com
> 
> Note: I did not do any of the "fix" steps (i.e. export 
LD_ASSUME_KERNEL=2.2.5) 

	cls Assignee
	Comment 7 • 23 years ago

There is a post in mozilla-unix in which someone claims that problems with the
plugin that are removed by _commenting out_ the LD_ASSUME_KERNEL line.  The
reporter was using the 0.9.1 RedHat_7x rpms.

news://news.mozilla.org:119/3B2BE377.306C4EC6@hotmail.com
news://news.mozilla.org:119/3B2CD209.939AB872@hotmail.com

	Christopher Blizzard (:blizzard)
	Comment 8 • 23 years ago

Interesting.  Ed, is there a way I can tell which plugin version is installed by
something on the filesystem?  Maybe a file that's in 1.3.1 that wasn't in
previous versions?  I've verified here that some sites work that I've never seen
work before with the new plugin*.  It didn't seem to matter whether or not the
variable was set.  My problem is what if people still have the old plugin
installed that I should still set the variable.

I don't know how setting that variable affects statup like in those messages.
* My wife can play games on the net.  She will be very pleased.

	Christopher Blizzard (:blizzard)
	Comment 9 • 23 years ago

Have a look at this file:

http://lxr.mozilla.org/seamonkey/source/build/package/rpm/SOURCES/mozilla.sh

It's got a test for which JVM is installed in plugins.  lxr's taking it's time
updating so the jvm test should show up there.  Unfortunately, it requires
running the jvm to get the version string.  Yuck.

	cls Assignee
	Comment 10 • 23 years ago

This bug doesn't need to be bounced up more milestones.  Marking WONTFIX and a
relnote will need to be added to inform users to upgrade to the JDK 1.3.1 plugin
or add the env variable setting.

	Jon Granrose
	Comment 11 • 23 years ago

verified.

	cls Assignee
	Comment 12 • 23 years ago

*** Bug 99866 has been marked as a duplicate of this bug. ***