When local state_descr is needed?

When local state_descr is needed?

[Marko Myllynen]

Hi,

It's a bit unclear to me when local state_descr should be declared.
Reading the man page makes me think that it certainly doesn't hurt to
do it especially when using the ->string convention but is it in fact
needed then to avoid any potential or concrete side-effect?

For example, does the patch below matter in practice at all or not?
In the file later on there is:

  "(-f --from-code $exargs)"{-f+,--from-code=}'[specify code set of input file]:code set:->from_codeset'
  "(-t --to-code $exargs)"{-t+,--to-code=}'[specify code set for output]:code set:->to_codeset'

---
 Completion/Unix/Command/_iconv | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/Completion/Unix/Command/_iconv b/Completion/Unix/Command/_iconv
index d040be7..fd74e6b 100644
--- a/Completion/Unix/Command/_iconv
+++ b/Completion/Unix/Command/_iconv
@@ -1,6 +1,6 @@
 #compdef iconv
 
-local expl curcontext="$curcontext" state line variant ret=1
+local expl curcontext="$curcontext" state state_descr line variant ret=1
 
 if _pick_variant -r variant libiconv='GNU*libiconv' glibc='(GNU*libc|EGLIBC|Gentoo)' unix --version; then
   local -a args

Thanks,

-- 
Marko Myllynen

Re: When local state_descr is needed?

[Oliver Kiddle]

Marko Myllynen wrote:
>
> It's a bit unclear to me when local state_descr should be declared.
> Reading the man page makes me think that it certainly doesn't hurt to
> do it especially when using the ->string convention but is it in fact
> needed then to avoid any potential or concrete side-effect?

_main_complete declares it local so it mostly isn't needed.

state_descr was added much later so there were many existing functions
that didn't declare it local. Even before, we've often been lax about
declaring opt_args.

I can think of contrived cases where it would be needed. You'd need to
be using it in a state loop where _arguments/_values is also being
called as part of handling a state.

I tend to only declare it in functions that reference it.
I don't think it's really worth patching vast numbers of functions for
it but I wouldn't care either if someone wants to.

Oliver

Re: When local state_descr is needed?

[Marko Myllynen]

Hi,

On 2016-06-09 17:43, Oliver Kiddle wrote:
> Marko Myllynen wrote:
>>
>> It's a bit unclear to me when local state_descr should be declared.
>> Reading the man page makes me think that it certainly doesn't hurt to
>> do it especially when using the ->string convention but is it in fact
>> needed then to avoid any potential or concrete side-effect?
> 
> _main_complete declares it local so it mostly isn't needed.
> 
> state_descr was added much later so there were many existing functions
> that didn't declare it local. Even before, we've often been lax about
> declaring opt_args.
> 
> I tend to only declare it in functions that reference it.
> I don't think it's really worth patching vast numbers of functions [...]

Thanks, all clear now.

Cheers,

-- 
Marko Myllynen