yml@grebyn.com (Yermo M. Lamers) (10/17/90)
[Reposting this article : the first version didn't seem to make it out.]
I am having some problems with SCO CGI running under SCO Unix Sysv 3.2:
1. I have a Xenix 386 2.3 application that makes use of CGI. It has been
running under Xenix 386 for several months using the ega driver. I
moved the executable which was compiled on the Xenix system to my
SCO Unix system (which happens to run Open Desktop). I let the Xenix
executables run over night. By the morning, the process had died
leaving the screen blank. Multiscreens did not work; nor did anything
I type appear on the screen. Using telnet, I logged into the machine
from another system and ran /etc/clean_screen which seemed to fix the
problem. I noticed that the "LCK.." lock files were still present in
/tmp.
This leads me to believe that there is some problem with the SCO
Unix version of CGI.
2. I then tried to compile the same program using the Unix Sysv 3.2
development system. Since SCO did not supply COFF libraries for
CGI I had to use the -x2.3 compiler option.
After some investigation and several attempted runs of my newly
compiled program, I noticed that the process was never returning from the
"open workstation" call. The screen would blank and after about a
minute the disk would start to thrash as if it were swapping a
process out to disk.
My process opens a pipe to another little process of mine. I
noticed that if I call open workstation before I open the pipe,
open workstation returns correctly. However, if I call it after
opening the pipe, it never returns. (The identical code runs
flawlessly under Xenix 386!)
To continue, I commented out the pipe code for now.
3. I managed to get my 'main screen' up which is your basic Welcome
screen which constantly displays the current time and some other
info until the user presses a key. I let this run for about an
hour when I noticed that text was being displayed at seemingly
randomn locations on the screen. It was as if CGI glitched on
one pass. I've this same "alpha text not being drawn where I
asked it to be drawn" problem in CGI 1.0 under Xenix 286. Has
anyone else seen this problem?
My initial gut reaction was to blame my code. However, the identical
source code has been running without incident under Xenix 386 for
several months now. If it runs under Xenix it should run under Unix.
Right?
Any help with these problems would be GREATLY appreciated,
Yermo Lamers
yml@grebyn.com