Re: proposal: "-batch" command line argument

[Pascal J. B. <pj...@in...>]

On 2010-03-03, at 11:17, <Joe...@t-...> wrote:

> Sam Steingold wrote:
>
>> 2. saveinitmem should accept :batch argument can take 3 values:
>> :default  - the current behavior based on other command line args
>> t - argv_batchmode_p is true
>> nil - argv_batchmode_p is false
>
> Wouldn't that break the old "requirement" of always being able to
> get a REPL even when started from an image?
> (Or did I forget that such requirement had been dropped at some time?)
>
> :default is a misnomer. Imagine someone does
> (saveinitmem T/NIL)
> starts that image and therein does
> (saveinitmem :default)
> One would expect :default to preserve the setting (i.e. T/NIL).
> :standard would clearly mean: "set standard behavior".

Perhaps what we want here is a general fd <-> stream mapping facility.

generate-password \
3| clisp
   -d 0:i=cl:*standard-input* \
   --file-descritor 1=cl:*standard-output* \
   -d 2:o=cl:*error-output* \
   -d 3:i=cl-user:*password-input* \
   -d 4:o=cl-user:*my-special-output* \
   -d 5:o=cl:*trace-output* \
   -d 6:io=cl:*remote-repl* \
   < input > output 2> error 4> special 5> trace 6<>/dev/tcp/localhost/ 
23

[-d|--file-descriptor] <fd>[':'['i'|'o'|'io']'='e<symbol>

<symbol> would be read with the standard readtable, in the CL-USER  
package.


The above --batch option would map semantically to a set of --file- 
descriptor options.
This doesn't preclude a REPL, only one which goes to some specific  
file descriptors.

-- 
__Pascal Bourguignon__
http://www.informatimago.com/