Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!triceratops.cis.ohio-state.edu!karl From: karl@triceratops.cis.ohio-state.edu (Karl Kleinpaste) Newsgroups: comp.emacs Subject: Re: compiling emacs Message-ID: Date: 13 Feb 89 13:43:47 GMT References: <1034@sdcc15.ucsd.edu> <631@wpi.WPI.EDU> <385@frksyv.UUCP> Sender: news@tut.cis.ohio-state.edu Distribution: usa Organization: OSU Lines: 35 In-reply-to: frk@frksyv.UUCP's message of 11 Feb 89 10:48:29 GMT frk@frksyv.UUCP (Frank Korzeniewski) writes: >[I said:] >Not quite true; Rel2.2 [on 3b20s] took care of [termio] problems. GNU >Emacs has been compiling routinely on such machines since about 18.36. >No workarounds weirder than usual for SysV were necessary. I beg to differ (actually I'll just differ anyway). I built 18.52 on SYS V/386 and re-worked it to emulate WordStar keys. Control-G was then the delete forward character. On using emacs, it would occasionaly die when I typed multiple control-g's. First, I was refering to 3b20s, which is where Ron was noting having had problems due to termio bugs. I persist that Rel2.2 for the 3b20 (!= Rel2.2 for any other CPU) cleared up these problems, and I'll ask any GNU Emacs users on cbrme.att.com to let me know if they are aware of any problems remaining. It seems that emacs uses the termio stuff to set control-g as generating SIGQUIT and SIGINTR. If you type a second control-g when the process is still paged out, SYS V then kills it. I had to map the DEL key to SIGQUIT and SIGINTR in order to make emacs useable. The problem is not at all a bug in how termio operates, but rather in unreliable SysVRel2 signals in combination with your PC's slow disc. In SysVRel2 and previous (dating all the way back into prehistory), the delivery of a signal to a process causes that process' handler for that signal to be reset to SIG_DFL. Thus, the process is momentarily susceptible to the signal in question. If a 2nd such signal arrives before the process has been allowed to run the handler, which would (one presumes) re-assert the handler definition, the process will die. Hence, the problem is that you can type ^Gs faster than your disc can page in. Now, that speaks volumes, yes?