[erlang-questions] [Bug] io:setopts and expand_fun option

[Gunilla Arendt]

Ok, I've made a note about this. Thanks for the bug report.

Regards,
Gunilla, Erlang/OTP team

igwan wrote:
> Hello,
> 
> I encountered a problem while playing with io:setopts trying to set the 
> expand_fun option :
> 
> 5> Completion = fun("") -> {yes, "start", ["start", "stop"]};
> 5> (_) -> {no, "", ["start", "stop"]} end.
> #Fun<erl_eval.6.56006484>
> 
> 6> io:setopts([{expand_fun, Completion}]).
> {error,badarg}
> 
> 7> io:setopts([binary]).
> ok
> 
> 8> io:setopts([binary, {expand_fun, Completion}]).
> ok
> 
> The problem is that setopts doesn't accept the 'expand_fun' option 
> alone, it has to be provided along with either 'binary' or 'list' options.
> This is not stated in the documentation.
> The faulty code seems to be located in kernel/group.erl :
> 
> %% setopts
> setopts(Opts0,_Drv, Buf) ->
>     Opts = proplists:substitute_negations([{list,binary}], Opts0),
>     put(expand_fun, proplists:get_value(expand_fun, Opts, get(expand_fun))),
>     case proplists:get_value(binary, Opts) of
>     true ->
>         put(read_mode,binary),
>         {ok,ok,Buf};
>     false ->
>         put(read_mode,list),
>         {ok,ok,Buf};
>     _ ->
>         {error,{error,badarg},Buf}
>     end.
> 
> expand_fun is put in the process dictionary even if we get a {error, 
> badarg} in response. I think it is not the intended behaviour. At least, 
> it doesn't meet my expectation of "least astonishment"  :)
> 
> igwan