Re: [PATCH 2/2] test-lib.sh: fix color support when tput needs ~/.terminfo

From: Richard Hansen <rhansen@bbn.com>
Date: Jun 17, 2015, 19:55 UTC
Message-ID: <5581D099.7090200@bbn.com>
In-Reply-To: <20150617194315.GE25304@peff.net>

On 2015-06-17 15:43, Jeff King wrote:

Show 22 quoted lines

> On Wed, Jun 17, 2015 at 03:06:26PM -0400, Richard Hansen wrote:
>> +	say_color_error=$(tput bold; tput setaf 1) # bold red
>> +	say_color_skip=$(tput setaf 4) # blue
>> +	say_color_warn=$(tput setaf 3) # brown/yellow
>> +	say_color_pass=$(tput setaf 2) # green
>> +	say_color_info=$(tput setaf 6) # cyan
>> +	say_color_sgr0=$(tput sgr0)
>> [...]
>> +		error|skip|warn|pass|info)
>> +			eval "say_color_color=\$say_color_$1";;
>>  		*)
>>  			test -n "$quiet" && return;;
> 
> I think you could dispense with this case statement entirely and do:
> 
>   eval "say_color_color=\$say_color_$1"
>   if test -z "$say_color_color"; then
>           test -n "$quiet" && return
>   fi
> 
> I guess that is making the assumption that all colors have non-zero
> sizes, but that seems reasonable.

We could test if the variable is set first (test -n "${foo+set}"), at the cost of a bit more complexity.

> I do not mind it so much as you have
> it, but it does mean adding a new field needs to update two spots.

I also don't like the duplicate list of color types, and I considered
doing something similar to what you suggested, but I decided against it.
 I'm a bit worried about bizarre syntax errors or code execution if
say_color() is used improperly.  ('eval' with uncontrolled variables
makes me nervous.)

Thanks for reviewing, Richard