eric@pyramid.pyramid.com (Eric Bergan) (02/06/90)
In article <45889@lll-winken.LLNL.GOV> brooks@maddog.llnl.gov (Eugene Brooks) writes: >A group here runs a generic environment which has the window >manager "awm" as the last command in the .xsession file (started >with xdm). The f.exit function causes awm to shut down and >ends the session. The problem is that xterms don't seem to notice >this event and keep running in who knows what state. One some >SYSTEM 5 based systems, these xterms run in a seeming compute bound >loop, perhaps handling signals repetitively or something. It is >not clear what is really happening. Check the signal masks being inherited by .xsession, and hence the xterms. We found that our xdm was setting SIGPIPE to ignore, so in some circumstances, xterm was ignoring the SIGPIPE when it did an operation on the socket to the window that had already disappeared during the killclient. It would then loop, always getting the signal, and always ignoring it. If you are using your own version of xdm, look through the source at all the signal handling. Before going off to do .xsession, it should be careful to return the signal mask to the "normal" setting. If you are using a vendor's implementation, you may have to do something more roundabout, such as have .xsession call a program which resets signal handling, then exec's what you really want to happen. I haven't looked yet to see if R4 has this problem. -- eric ...!pyramid!eric