Casper.Dik at Sun.COM wrote:
> So what exactly does this do:
>
> if [[ "${cmd}" != ~(Elr)(alias|unalias|command) ]] && ! alias "${cmd}"
> >/dev/null 2>&1 ; then
> builtin "${cmd}"
> fi
>
>
> Clearly this needs to be fixed.
Ok, since I've read enough of this to know, I'll answer:
that tests whether the command was invoked with the exact name 'alias',
'unalias', or 'command', or if the command is aliased; if not, it executes the
'builtin' command on it. I also see grammatical errors in the comment
'explaining' what the above line does, which could be fixed to make that some
clearer.
The fact that ~(Elr), a modifier to the pattern-match, does a homedir expansion
is apparently the bug here. That causes the erroneous nameservice lookup.
> Also, it looks at my whole $PATH; it shouldn't do that.
Not sure why that's happening; probably because of "command" below. It seems
to
me that "command" should have -p there; not sure why it doesn't.
> And since there's
> a plan to make more executables ksh93 built-in, we need to make sure that
> doesn't happen until we have fixed all bugs in "sleep".
As Roland noted, it gets even more complex, because sleep is now a binary so
that pgrep sleep works. making one wonder again at the point of this.
I guess the argument is that the paradigm shifts to "userland is a set of
programs" to "userland is at least half a shell that does everything simple
more
or less on its own".
