177416 - Certain JavaScript calls to Java Plugin cause a hang

Status: RESOLVED INCOMPLETE

(Core Graveyard :: Java: OJI, defect)

Description

[Martin Lawyer]

This problem was reported by HP against Netscape 7.0, but has been reproduced
using the latest nightly build of Mozilla.  

There is a complete test case package with detailed description available here:
ftp://depot.netscape.com/private/martinl/JAGae44329.tar.gz

Note:  The term "Liveconnect" is used inside the detailed description only to
refer to the Java/Javascript communication.  The reporter has stated that the
latest API's are being used.

Also Note:  For Netscape people who might be looking at this problem, the test
case is set up on a server inside the Netscape firewall.  Everything needed to
set up the test case is in the package above.

Summary extracted from package above: 

JPI 1.4 cannot be called from JavaScript in NS 6.x/7.x

This bug describes a class of problem where JPI sends a work request to the
browser through a work socket during a Javascript to Java invocation.

The test case consists of FormSocket.html, Formsocket.java, and kit.html. 
FormSocket.html makes a Javascript to Java call to validateField function in
FormSocket.java.  The validateField function opens a URL connection to
"http://haathi.nscp.aoltw.net/hp/qa/HP_tests/kit.html" to read the text stream
from kit.html.

Before a stream is opened, JPI (javaplugin.jar) sends a request to work socket
to the browser via libjavaplugin_oji.sl to get the cookie info for this URL.

The problem is the request is posted on the work socket, and
libjavaplugin_oji.sl detects the request on work socket, and schedule a task
with the browser event queue so that the request can be read from the work
socket.  The browser is supposed to call the JPI handler to read the work
socket, but the handler is not called.  Therefore, the test case hangs, because
the JPI child process is waiting for the response for the request.

Comment 1

[Martin Lawyer]

This bug was reported by Kit Chatsinchai ( [e-mail] ) at HP.  He did all
of the initial debugging, and is available to help further in the final
resolution of this bug. 

Comment 2

[Phil Schwartau]

Setting component, QA to OJI -

Comment 3

[Dan Weinstein]

A few notes for this bug:
The website setup with the test case is at:
http://haathi.nscp.aoltw.net/hp/qa/HP_tests/test_case/FormSocket.html

click on the "URL" button to cause the hang.

To catch this bug at the hang, set a break point at jsj_method.c:1473, start the
browser, open the test case page, click the "URL" button, once it stops at the
previous break point, set a break point at ptio.c:1815, and continue.  It hangs
on this line, in the function recv.

Comment 4

[Bob Clary [:bc] (inactive)]

-> default assignee for old netscape assigned bugs.

Comment 5

[Benjamin Smedberg]

Mass-closing bugs in the "OJI" component: OJI plugin integration was replaced with npruntime long ago, and these bugs appear to be irrelevant now. If there is in fact a real bug that remains, please file it new in the "Core" product, component "Plug-ins".