storm@ambush.UUCP (Kim F. Storm) (10/22/87)
In article <806@mrstve.UUCP> rjk@mrstve.UUCP (Richard Kuhns) writes: >I've got 2 problems running GNU emacs on a 3B2/400, System V Rel 3.1, >4 MB RAM, and I'd greatly appreciate some help. >First, substitute-command-keys doesn't work -- the only character it >can substitute is RET (meaning, where I should normally get >`C-X 4 B RET', all that comes out is RET). My guess is that you have compiled GNU with "cc -O " which among other problems causes substitute-command-keys to behave exactly as you describe. Be shure that C_OPTIMIZE_SWITCH is defined to *nothing* in m-att3b.h, and that you don't change it elsewhere. To quote from m-att3b.h: -O has been observed to make correct C code in Emacs not work. So don't try to use it. When I got issue 4 of the 3B2 C compiler, I tried it anyway to see if the problems had gone away, but NO LUCK: I got the problem you describe, and many others of a similar kind. I tracked most of these problems (including yours) to the function 'mapconcat' which did not work, but I never figured out why (except that it seems that nargs == -1, but ...). Another effect of using -O was that the command-history contains a meaningless expression like ((eval-expression (quote eval-expression)) ...) which I don't think I ever tracked down (perhaps it was mapconcat as well?). Any comments from AT&T ? -- Kim F. Storm, storm@ambush.uucp (or ..!mcvax!diku!ambush!storm) AmbraSoft A/S, Rojelskaer 15, DK-2840 Holte, Denmark. tel: +45 2424 111
dougm@ico.UUCP (Doug McCallum) (10/25/87)
In article <535@ambush.UUCP> storm@ambush.UUCP (Kim F. Storm) writes: >In article <806@mrstve.UUCP> rjk@mrstve.UUCP (Richard Kuhns) writes: > ... > >I tracked most of these problems (including yours) to the function >'mapconcat' which did not work, but I never figured out why (except >that it seems that nargs == -1, but ...). > > >Another effect of using -O was that the command-history contains a meaningless >expression like I've seen the same problem on V.3 on a 386. If the file fns.c is compiled with -O, the problem is seen (on 18.40 but not 18.44). What is happening is that the call to Fconcat in mapconcat is getting expanded inline to a direct call to concat. During the expansion, the optimizer is blowing the stack. At least this is what happens on the 386 compiler. This sounds like a generic RCC optimizer bug.