topic Re: debugger motif error codes? in Operating System - OpenVMS

debugger motif error codes?

Timothy J Stegner — Thu, 26 Feb 2004 15:02:26 GMT

Running the debugger in decwindows mode to examine a TCP/IP based service, I'm getting the following error - where can I locate an explanation for these codes?

XIO: fatal IO error 65535 on X server "_WSA17:"

VMS V722, with MUP applied. 'UCX' V5.1

thanks,
Tym

Re: debugger motif error codes?

labadie_1 — Thu, 26 Feb 2004 15:32:59 GMT

If you really have Ucx (aka Tcpip) V5.1, you should first apply the Eco 5, reboot and retry.

ana/sys
tcpip
tcpip sh bug
tcpip sh crash

may help

:-)

(joking)

Gerard

Re: debugger motif error codes?

Martin P.J. Zinser — Thu, 26 Feb 2004 16:17:45 GMT

While Gerards expalnations are certainly funny (esp if you have a chance to meet the "real" people ;-) I do not think this will help with the specific question posed.

Unfortunatly I do not think the error code was filled with a reasonable value as it translates to hex FFFF, which just looks like to much of a coincidence. (Somebody in the "know" please correct me if I am wrong!)

Do you have some move information on the circumstances the error occurs in, e.g. is it throwm by your application or the debugger itself etc.

Greetings, Martin

Re: debugger motif error codes?

Willem Grooters — Fri, 27 Feb 2004 02:59:01 GMT

Martin:
Correct.
This is appearently no VMS error code. VMS doesn't have odd fatals.
(For the innocent: for VMS, the last 3 bits deterimine severity: 000 (0) = warning, 001 (1) = success, 010 (2) = error, 011 (3) = Informational, 100 (4) = fatal, rest 'reserved for future use'. So if odd, it's ok, otherwise there's something wrong).

Tym:
I guess it's a unix state, but these won't tell you anything unless you know the source. In this case: the IO-stack of X-Windows...
It might be a conflict between the service you're debugging and decWindows (what actually is an implementation of X_Windows). Try to debug this service using the plain VMS debugger, you'll prevent that problem anyway.

Tell us more, like Martin said. That may shine some light.

Re: debugger motif error codes?

Martin Kirby — Fri, 27 Feb 2004 04:48:18 GMT

65535 is EVMSERR an errno value that means the actual error code is in vaxc$errno.

The text you show isn't that used by the default DECwindows I/O error handler, so I guess Debug must have its own X IO error handler.

Try running DECW$EXAMPLES:ICO to the server. That has a better error handler that can give a more meaningful message.

Martin Kirby

Re: debugger motif error codes?

Timothy J Stegner — Fri, 27 Feb 2004 09:51:09 GMT

What we are trying to debug is an ODBC Server, started by a connection to a particular TCP port.

(Note that all the following worked FINE last week, when I had Multinet V54 running, instead.)

I can run the debugger fine interactively, but it's not possible to debug the problem that way.

We set up the TCP port command file to invoke our ODBC server, and further modify this command file to define the debug environment, including the necessary SET DISPLAY/CREATE command.

DECW$EXAMPLES:ICO.EXE, as well as another private DecW application run FINE. Due to the query client, MSQUERY, making two connections per request, I have two error files, each showing a slightly different view of the problem. (See attachment for both file extracts.)

I'm willing to apply ECO5, or even upgrade UCX to V5.4, but I'm having trouble locating them in the ITC patch system?

Re: debugger motif error codes?

Willem Grooters — Mon, 01 Mar 2004 02:21:55 GMT

So that's has been the change - change Multinet with HP's TCPIP...

I would check if there is anything left of Multinet - I think of installed shared images in the first place, but check the process environment of the service as well, there might be references to Multinet routines left (the service will run the login-procedures of the specified user when started).

To locate the ECO (I understand your trouble):

point to http://www5.itrc.hp.com/service/patch/search.do and search for tcpip. you'll find your patch-description at URL:
http://www5.itrc.hp.com/service/patch/patchDetail.do?BC=patch.breadcrumb.main|patch.breadcrumb.search|&patchid=VMS731_SYS-V0500&context=openvms:alpha:7.3-1

Re: debugger motif error codes?

Martin Kirby — Mon, 01 Mar 2004 12:19:29 GMT

Tym,

In the log there is:

++++++++

XIO: fatal IO error 65535 on X server "_WSA29:"
after 10607 requests (10605 known processed) with 0 events remaining.

....

%XLIB-F-IOERROR, xlib io error
-SYSTEM-F-CONNECFAIL, connect to network object timed-out or failed
----- end of exception message

++++++++

10605 requests processed means it connected to the display server and did a fair bit of work before it terminated. So it is not a simple case of can't connect or security etc.

However, "CONNECFAIL" isn't the usual message when a server connection is terminated, that is LINKDISCON. Not sure what would cause CONNECFAIL.

Does your DBG$INIT file do anything, such as run your application, that might be interfering with the link? Does EXCEED have a log file that might show why it thinks the connection was closed?

Martin Kirby

Re: debugger motif error codes?

Timothy J Stegner — Mon, 01 Mar 2004 16:52:07 GMT

Got ECO 5 applied. Still being a problem.
DEBUG.INI file+does+the+following:
SET LOG SDSK:[xxxxx]debugger.log
SET OUTPUT LOG
SET STEP LINE
set lang C
!SET MODE KEYPAD
!SET MODE SCREEN
set source sdsk:[yyyyy], zz_cmsref:
set image srv
set module mgrmain
... !bunch+of+set+module/set+source+cmds
go

Exceed+trace+file+is+attached.

From+the+client+end, the+steps+are+as+follows:
1. Minimize+all+other+windows+on+the+PC.
2. Ensure+that+Exceed+is+running
3. Start MSQuery
4. Click icon for new query
5. Choose appropriate data source.
6. Debugger window comes up, subwindows, but no focus to that window, nor does any of the debugger displays paint, nor does the prompt come up.
7. After a while, I kill off the MSQuery window by clicking CLOSE in the icon at the bottom of the screen, then agreeing to end the task.
8. Debugger wakes up at that point, and I can EXIT out normally.

I do have Exceed set to give the newly raised window focus, but it doesn't get it.

Running out of ideas-I'm probably gonna call support tomorrow (3/2/04)

thanks for all the help, though.

-Tym

Re: debugger motif error codes?

Martin P.J. Zinser — Mon, 01 Mar 2004 21:50:40 GMT

Hello Tym,

did you recompile your application after moving from Multinet to UCX? If yes, did that work
without problems (I seem to remember there are a number of Multinet routines that are not supported with UCX).

Also what do you use as the transport for X? I suppose TCP/IP. If yes, can you test if you get the same problem directly at the console using the local transport?

Greetings, Martin

Re: debugger motif error codes?

Martin Kirby — Tue, 02 Mar 2004 04:35:39 GMT

Tym,

I had guessed that the debugger windows never showed, although I see you don't actually say that.

If you remove the "go" from the debugger init file does the debugger start-up okay?

If it does then does it crash when you do the "go"?

If it does then I think you need to debug your LIB$INITIALIZE code and see what causes the debugger to crash. If the "go" is causing the crash you might be able to get some information from your process by getting a process dump from it.

Martin

Re: debugger motif error codes?

Timothy J Stegner — Tue, 02 Mar 2004 13:45:13 GMT

Martin, et al:

The application hasn't been specifically recompiled under UCX, mostly because we've never had to do that before - we've been shipping the same image for UCX, Multinet, TCPWARE, etc.

Tried removing the GO from DEBUG.INI - no difference in the results.

On the other hand, I went back to the original programs (non-debug), and find that our fixed version of the program now works under UCX - as a result of applying ECO5.

Need to check with management, but the entire debugger issue may have just become moot.

-Tym