[PATCH 0/3] add "call" command

Rasmus Villemoes rasmus.villemoes at prevas.dk
Fri Sep 25 15:51:42 CEST 2020


On 25/09/2020 15.38, Heinrich Schuchardt wrote:
> On 25.09.20 15:09, Wolfgang Denk wrote:
>> Dear Heinrich Schuchardt,
>>
>> In message <4b00225d-d960-4a14-9aec-110ddddf7f30 at gmx.de> you wrote:
>>>
>>> Further we cannot first introduce a command call and then eliminate it
>>> due to backward compatibility. We should decide on the final version
>>> beforehand.
>>
>> Full agreement.  we need a concept of what is needed / wanted first.
>> And then we should look how current vrsions of hush fit into this.
>>
>>> In the Linux world you can override a command using an alias. So I am
>>> not sure if a built in command should take precedence over a variable of
>>> the same name or the other way round.
>>
>> This is simple. The PoLA (Principle of Least Astonishment) applies
>> here.  Behaviour must be the same as in other (to some extent POSIX
>> compatible) shells.  A shell should parse it's input, not adhoculate
>> it.
> 
> For me this could be realized by enhancing the run command to allow:
> 
> run varname1 varname2 ... varnameN --args argv1 argv2 argv3
> 
> Arguments argv1, argv2, ... are passed to the script identified by the
> last variable (varnameN).
> 
> No new command to learn. Just a new option.

Yes, this is really more to be thought of as a "run_with_args" command
than an extension of hush (though the $1 treatment does need to hook
into the hush code, which is both why I made it dependent on HUSH_PARSER
and made it live in cli_hush.c). I'm certainly open to extending the
existing run command instead of creating a new "toplevel" command.
Though I'd probably make it

  run varname -- arg1 arg2 arg3

instead: Just use -- as a separator [that has precedent as "stop doing
X, use the rest as argv", though X is normally "interpret options" and
now it would be "read function names to run"], and only allow a single
"function" to be called. Otherwise, I don't there's any natural answer
to whether all the varnameX or only the last should be called with the
positional arguments. It's pretty simple to do "for x in v1 v2 v3; do
run $x -- arg1 arg2 arg3 ; done if one has a bunch of functions that
should be called in turn, and it's even more simple to do

  run varname1 varname2 varname{N-1}
  run varnameN -- arg1 arg2 arg3

if one has a bunch of parameter-less functions to call before varnameN.

Rasmus

Rasmus


More information about the U-Boot mailing list