On Wed, Oct 17, 2007 at 10:46:46AM +0800, Herbert Xu wrote: > On Tue, Oct 16, 2007 at 10:30:15PM +0200, Oleg Verych wrote: > > > > > 1) The real Korn shell does the same thing; > > > 2) It makes the code bigger :) > > > > Well, i don't know all that korn/bourne/tcsh/bash/whatever hell. > > The POSIX specification is mostly based on the Korn shell and > it is the prototypical POSIX compliant shell. > > However, more importantly this behaviour is consistent with > POSIX. Since one of the primary aims of dash is to be as small > as possible (the one being as fast as possible) while staying > within the realms of POSIX-compliance,
OK. > I'd rather not apply this patch. Unless of course if you can show me a > way to do it while at the same time making dash smaller :) That long historical bugs journey on opengroup's site about `test` makes me sick. I think, this is example of how not to implement things. I like size optimized system programming. But while going that, i also like to have silly-bug free implementations, especially in "ah, 'cmon, arithmetics is hard" part :). BTW i did size optimization patch before this one exactly because i wanted to have overall new-old delta negative. And it is so. > > Also, this hides real bugs, when arithmetical comparsion succeeds, even > > I think this is debatable. Well, before major shell scripting i decided to make all my stuff `set -e` strong. Works for me, but not for shells (some times :). Also this is very good requirement in Debian, When test is used like this: [ "$PARAM" -ge 0 ] It will work in `dash`, but will fail in `bash`. Oops, > Even if dash made this an error it still may escape detection because > the test command is not a special built-in and cannot cause the whole > script to abort. So all this will do is print something to stderr and > cause the test to return false instead of a possible true value. > > BTW, even bash is not completely consistent here: let's not touch `bash`, please! It's very doomed stuff, IMHO. > $ bash -o posix -c '[ "" -lt 3 ]' > bash: line 0: [: : integer expression expected > $ bash -o posix -c '[ " " -lt 3 ]' I know, but see above :) How many broken and slow things its developer brought to the world... This must be left behind and remembered as classic case of "what not to do". ____ -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
