On Mon, 28 Aug 2006, Paul de Weerd wrote:
> On Mon, Aug 28, 2006 at 01:33:59AM +0200, Ingo Schwarze wrote:
> | > Please now try `if ! false; then echo true; fi`.
> | > Why does my shell eat muchos CPU & RAM after such a short pipeline?
> |
> | [EMAIL PROTECTED] $ uname -a
> | OpenBSD idefix 3.9 GENERIC#617 i386
> | [EMAIL PROTECTED] $ /bin/ksh
> | $ date
> | Mon Aug 28 01:28:07 CEST 2006
> | $ if ! false; then echo true; fi
> | true
> | $ date
> | Mon Aug 28 01:28:12 CEST 2006
> | $ date; ( if ! false; then echo true; fi ); date
> | Mon Aug 28 01:28:20 CEST 2006
> | true
> | Mon Aug 28 01:28:20 CEST 2006
> | $ time ( if ! false; then echo true; fi )
> | true
> | 0m0.01s real 0m0.00s user 0m0.01s system
> | $ time ( echo `if ! false; then echo true; fi` )
> | true
> | 0m0.04s real 0m0.00s user 0m0.02s system
> | $ exit
> |
> | Sorry, cannot reproduce.
> | Works for me (on OpenBSD 3.9-release).
>
> Now it takes quite a bit of time before it actually reaches this.
> Searching the manpage a bit more, I found another reference to '!'.
>
> $ set +o csh-history
> $ time ( if ! false; then echo true; fi )
> true
> 0m0.00s real 0m0.00s user 0m0.01s system
> $ set -o csh-history
> $ time ( if ! false; then echo true; fi )
> /bin/ksh: internal error: unable to allocate memory
>
> Since I set csh-history in my .kshrc, this option is set in all my
> shells. I don't think this is intended behaviour, but if it is, it
> probably should be documented in the manpage.
I think this is a bug.
I looked a bit at this and it seems a loop in the lexical anaylzer,
introduced by the code starting at line 162 of lex.c
Sadly this code is quite hard to follow. I'll try to dig a bit further
later.
-Otto
