We have a serious desktop hang in Plugin/JRE startup under Windows 98.

If you use explorer to visit a plugin enabled HTML file on the local disk
BEFORE you have visited any remote URLs, then explorer hangs. This causes
the entire desktop to totally hang, which is pretty bad.

If you visit a remote URL first and then visit the same local HTML file,
everything is OK. This showed up because I was testing a CDROM based
install where there was also a plugin demo HTML file on the disk. Naturally
the first thing I did after installing and rebooting was to visit the
demo file, which caused my desktop to hang. Woe is me.

It is extremely reproducible.

I tracked down where we are hanging. It is while loading the "net" DLL in
InetAddress initialization code
due to a call on InetAddress.getLocalHost
AppletClassLoader.getPermissions
SecureClassLoader.getProtectionDomain
SecureClassLoader.defineClass
...
AppletClassLoader.loadClass
...
applet startup code

To narrow down why this was happening, I changed the InetAddress code to
load WSOCK32 before loading net:
AccessController.doPrivileged(new LoadLibraryAction("WSOCK32"));
AccessController.doPrivileged(new LoadLibraryAction("net"));

This showed that we were hanging while loading WSOCK32.

I added extra tracing code in linker_md.c to report on the call into the
windows LoadLIbrary procedure. This code shows that we do a LoadLibrary
call on C:\WINDOWS\SYSTEM\WSOCK32.DLL and this call never returns.

This was in 1.2.0 and is also in Cricket RC1.

This looks like it is really a Windows 98 bug. Unfortunately I think we
still have to figure out how to deal with it. :-(


graham.hamilton@Eng 1999-05-27

The system is a Dell Dimension XPS 100 MHZ pentium with 32 Megs.
It's in my office, but I'm happy to loan it to anyone who wants
to look at it!

I did a fresh install of windows 98 from the MSDN cdrom. It was
a "from scratch" install and not an upgrade.

I hit the problem with the initial default configuration. I tried
tweaking various windows settings to try and pin down the bug, but
they didn't seem to make any difference.

DLL versions are:
    ws2_32.dll version 4.10.1998
    wsock32.dll version 4.10.1998


graham.hamilton@Eng 1999-05-27


OK, the good news and the bad news is that so far we can NOT reproduce this on
other (larger) systems. We've tried:
     Gary's 400 MHz desktop
     My 150 MHz 48 Meg laptop
     My 266 MHz 128 Meg desktop (brand new install of windows 98)
and they all run fine without showing the bug.

Just for grins, I deleted Windows 98 from my 100 MHz PC and reinstalled it (again).
This machine is still showing the bug. And I have Jerome as a witness that I'm not
just imaging it. ;-)

So one theory is that I need to sacrifice a chicken and un-curse my PC.
The other theory is that this is a bug that affects some low-end small
memory PCs.

Given our lack of success in reproducing this, I'll downgrade this to P2.

graham.hamilton@Eng 1999-05-27