[REBOL] Problems with 'call
From: tim-johnsons::web::com at: 19-Apr-2006 17:19
I'm picking up this topic again:
there are features to 'call on linux that are downright baffling, and if
I were to use it in an application, rather strange to me.
Let's look at some examples:
>> res: call "ls *.r"
user.r ;; okay
== 0 ;; exit code of 0 is correct
;; now let's introduce a legitimate call that has unsuccessful results
res: call "ls *.cpp"
ls: *.cpp: No such file or directory
== 0 ;; hmmm! now let's do the same from the shell.
[tim-linus tim]$ ls *.cpp
ls: *.cpp: No such file or directory
[tim-linus tim]$ echo $?
1
;; NOTE the differences in the exit code.
Now let's 'buffer' the output
>> buffer: copy ""
== ""
res: call/output "ls *.r" buffer
== 214 ;; say what?
>> buffer
== "user.r^/" ;; good
;; now the following is definitely unsettling, because it disables the
;; programmer's ability to gracefully handle errors:
>> clear buffer
== ""
>> res: call/output "ls *.cpp" buffer
;; Hoy moly, prints to stdout
ls: *.cpp: No such file or directory
== 214 ;; again exit code is inconsistant with shell command
>> buffer
== "" ;; and nothing in the buffer
;; Trying another refinement
res: call/error "ls *.cpp" buffer
== 214
>> buffer
== "ls: *.cpp: No such file or directory^/"
;; now that's a bit better, at least we've captured the error message.
;; but I'm still baffled about the anomalies in the exit codes.
;; BTW: using the /info refinement returns the same exit code as
;; an attribute
Just curious. Does anyone think that this should be treated as a bug?
Otherwise, I could wrap this in a function that could handle some of
these inconsistancies.
tim
--
Tim Johnson <tim-johnsons-web.com>
http://www.alaska-internet-solutions.com
