Hi John,
I think you are on to something here. I failed to notice that Javier tried
to launch a debuggee process twice in succession. Apparently JDE failed to
get the message to connect to port 1900 on the first attempt, perhaps
because standard I/O was blocked. On the second attemp, the message from
the first attempt finally reaches the JDE and the JDE connects to socket
1900. Meanwhile JDEbug has created another socket 1901 and has sent a
message to connect to that socket to the JDE. Again this second never
message never reaches the JDE and thus JDE never attempts to connect to the
socket and eventually the socket times out, resulting in the "gave up
waiting" message for JDEbug.
I think the problem may be figuring out what is blocking the JDE from
getting the socket message from JDEbug. Another approach may be just to cut
the Gordian knot and abandon the whole idea of having the operating system
assign a port. Rather, the JDE can include a predetermined (but
configurable) port number in the launch process command to JDEbug. For
example, the default port might by 5656. So JDE would pass this port number
ot JDEbug, wait a little while to give JDEbug time to launch the
application and set up a socket, and then attempt to connect to port 5656,
assuming that JDEbug has had time to create a server socket at 5656.
- Paul
At 11:54 AM 4/27/01 -0700, you wrote:
>Wild guess made without looking at the code:
>
>it may be trying to connect twice because it somehow didnt get hooked up
>with the successful java connection to port 1900.
>
>so java has sucessfully engaged port 1900 and is patiently waiting there.
>
>JDEbug may have timed out and attempted to reconnect. but it's discovered
>that there is 'some guy' on port 1900 so it tries to get java to start up
>with 1901.
>
>java sees no reason to honor the request because it's already busy waiting
>on 1900.
>
>thus, two lonely ships pass in the night :-)
>
>
>again, this is just a guess.
>
>On Fri, 27 Apr 2001, Paul Kinnucan wrote:
>
>> At 11:10 AM 4/27/01 -0400, Javier Lopez wrote:
>> >
>> > I tried it on simple class, (JDE->Debug App)
>> > and I get something like
>> > Error: debugger didn't respond to command:
>> > -1 2607 launch 12 -vmexec javaw -classic
>> >
>> > or something like
>> > No response to command 2612. (process = 13; timeout = 30 sec.)
>> >
>> > But the most intersting one is this one.
>> > this JDE->Debug App wait for it to fail and repeat JDE-Debug App
>> >
>> > *** Debugger Output for Process Test(2) ***
>> >
>> > Launched VM Java Debug Interface (Reference Implementation) version 1.3
>> > Java Debug Wire Protocol (Reference Implementation) version 1.0
>> > JVM Debug Interface version 1.0
>> > JVM version 1.3.0_01 (Classic VM, native threads, nojit)
>> > initSIOConnect: starting standard I/O handshake.
>> > initSIOConnect: starting SIO connect thread.
>> > Debugger waiting for Emacs to connect to app SIO port 1901.
>> > vm started...
>> > All threads suspended...
>> > Launch command line:
>> > javaw -classic Test c:/code/Development/src/ config
ConfigVariables.vars
>> > tmpl
>> >
>> > Emacs connected to standard IO port 1900 for process Test.
>> > VM options: '-classic '
>> > Gave up waiting for Emacs to connect to SIO port: 1901
>> >
>>
>>
>>
>> Wow! Putting a timout on the socket accept process in JDE-2.2.7beta11 really
>> paid off. I should have done this ages ago. It appears that for some reason
>> Emacs and Java are off by 1 on the socket number. This is the first solid
>> information I've gotten on this problem in about a year and would
explain the
>> hangup. Now the question is why the difference? I can't look at this right
>now
>> but if anyone else wants to pursue it, here is some background. One of the
>> steps performed by JDEbug when launching an application is to create a
server
>> socket. JDEbug allows Java to choose the port number of the server socket
>> (e.g., 1900). JDEbug then sends this port number back to the JDE. The JDE
>then
>> triees to connect to the socket on the specified port. What appears is that
>> somehow, on some systems, the port number gets bumped by 1 so that
JDEbug is
>> waiting on port 1901 while the JDE is trying to connect to port 1900.
>>
>> Any thoughts as to why this might be happening?
>>
>> - Paul
>>
>> >
>> > Javier
>> >
>> >
>> > -----Original Message-----
>> > From: Jim Goodwin [mailto:[EMAIL PROTECTED]]
>> > Sent: Friday, April 27, 2001 10:55 AM
>> > To: '[EMAIL PROTECTED]'
>> > Subject: JDEBug vs Win2K and/or NT4: Status?
>> >>
>> >> Hi all,
>> >>
>> >> I see in the mail-traffic that Paul K has changed the JDEBug code to
>avoid
>> >> setting priority on the thread when launching the debuggee process, in
>an
>> >> effort to cure the problems with deadlocking at launch.
>> >>
>> >> What has not appeared is any user report of whether this has helped or
>> >> cause the problems to go away. Has anyone tried this yet with Win2K
>> >> and/or NT 4? And if so, what experience have y'all had?
>> >> jim
>> >
>> >
>>
>>
