When Era Eriksson suggested this to
ant(_at_)notatla(_dot_)demon(_dot_)co(_dot_)uk as a way to
save the exit code of a command:
E> :0
E> * 1^1 ! ? hashcash
E> { SCORE=$= }
E> will actually put the exit code of hashcash in the variable SCORE
E> which you can then peruse as you see fit.
I wrote,
T> I wonder what will happen if the command generates a negative exit code,
T> though. Just in case, perhaps one should do this:
T> :0
T> * 1^1 ! ? hashcash
T> { }
T> SCORE = $=
To which Roderick Schertler responded,
S> By definition no such thing exists. A Unix exit code is an 8 bit
S> unsigned value, it is always between 0 and 255 inclusive.
In any case, the command might legitimately exit with return code 0, so I
still recommend leaving the braces empty and assigning SCORE outside them.
If you're further interested in this, read on.
The thing is that I've seen "exit -1" in scripts from time to time, so I
thought it was legitimate. In my .procmailrc, I tried a few things with
* 1^1 exit -1 ;
and with
* ! 1^1 exit -1 ;
and really couldn't get any definitive results. It turns out that pdksh
(my $SHELL) considers "exit -1" a syntax error (because -1 is not a legal
*option* for exit!) and returns code 1 on those grounds. The sh here
recognizes -1 as a number but as an illegal one for an exit code (as Roderick
said) and thus returns code 2 for the usage error. Neither returns 255.
So I guess that all those scripts I've seen that say things like "exit -1"
give unpredictable results, depending on which shell they run under, or that
there must have been some shell at some time for which "exit -1" meant some-
thing different from "exit -2". I'll have to hope to find such a shell and
try again.
But in any case, for a different reason, my recommendation stands: because
the exit code might be 0, assign SCORE = $= outside the braces, not between
them.