build.env-append working with merger in muniversal?

[Ryan Schmidt]

On Sep 15, 2010, at 08:50, Michael Dickens wrote:

> (2) In doing these experiments, I found that if I append the same environment variable more than once to build.env (or any other *.env), then only the last version is used -- and, no warnings or such are printed.  It's simple enough to check for this issue and print a warning; I'd also propose concatenating the various flags -- seems like a better solution than ignoring some.  Here's the diff that does both:
> 
> ++++++++++++++++++
> Index: base/src/port1.0/portutil.tcl
> ===================================================================
> --- base/src/port1.0/portutil.tcl       (revision 71479)
> +++ base/src/port1.0/portutil.tcl       (working copy)
> @@ -753,6 +753,12 @@
> 
>         while {[regexp "^(?: *)(\[^= \]+)=(\"|'|)(\[^\"'\]*?)\\2(?: +|$)(.*)$" ${the_environment} matchVar key delimiter value remaining]} {
>             set the_environment ${remaining}
> +            if {[info exists ${command}.env_array(${key})]} {
> +                ui_debug "Warning: Multiple values provided for ${command} environment variable '${key}'; concatenating them."
> +                set previous_key_val [set ${command}.env_array(${key})]
> +                set concat_val [concat ${previous_key_val} ${value}]
> +                set value [join ${concat_val} " "]
> +            }
>             set ${command}.env_array(${key}) ${value}
>         }
>     } else {
> 
> ++++++++++++++++++

Existing ports might depend on the existing behavior.

Though IMHO MacPorts is within its right to do it either way; behavior is IMHO undefined if you ask for the same env var to be appended multiple times with different values, and port authors should refrain from doing that.