deleting all default configure flags causes error.

[tonytung@…]

In the tcl port, -I${prefix}/include is removed from cppflags:

configure.cppflags-delete  -I${prefix}/include

When handle_option-delete processes this, it yields an empty string and unsets the option.

##                                                                                                              
# Handle option-delete                                                                                          
#                                                                                                               
# @param option name of the option                                                                              
# @param args arguments                                                                                         
proc handle_option-delete {option args} {
    global $option user_options option_procs

    if {![info exists user_options($option)] && [info exists $option]} {
        set temp [set $option]
        foreach val $args {
            set temp [ldelete $temp $val]
        }
        if {$temp eq ""} {
            unset $option
        } else {
            set $option $temp
        }
    }
}

This results in portconfigure.tcl declaring ${configure.cppflags} as an unknown variable in configure_main(..), yielding this error message:

Error: Target org.macports.configure returned: can't read "configure.cppflags": no such variable

Replacing these lines:

append_list_to_environment_value configure "CPPFLAGS" ${configure.cppflags}
append_list_to_environment_value configure "LDFLAGS" ${configure.ldflags}

with:

catch {append_list_to_environment_value configure "CPPFLAGS" ${configure.cppflags}}
catch {append_list_to_environment_value configure "LDFLAGS" ${configure.ldflags}}

seems to do the trick (i.e., everything builds fine).

[tonytung@…]

Cc Me!

[MarcusCalhoun-Lopez (Marcus Calhoun-Lopez)]

Cc Me!

[blb@…]

Note that I committed a kludge to the tcl port in r44859 to work around this for now.

[tobypeterson]

r44901 ... this is simpler and more reasonable behavior for "*-delete", and doesn't appear to cause any problems.