[84313] trunk/dports/lang/clisp/Portfile

[Ryan Schmidt]

On Sep 22, 2011, at 03:01, and.damore at macports.org wrote:

> Revision: 84313
>          http://trac.macports.org/changeset/84313
> Author:   and.damore at macports.org
> Date:     2011-09-22 01:01:17 -0700 (Thu, 22 Sep 2011)
> Log Message:
> -----------
> Removed clisp port's overridden phases
> 
> Modified Paths:
> --------------
>    trunk/dports/lang/clisp/Portfile

> +post-extract {
> +    file mkdir ${workpath}/home
> +}

> +use_parallel_build  no
> +build.dir           ${worksrcpath}/src
> +build.env-append    HOME=${workpath}/home
> +build.env-delete    LD_PREBIND LD_PREBIND_ALLOW_OVERLAP
> +build.cmd           "ulimit -s 16384 && make"
> +build.target 
> 
> -build {
> -    file mkdir ${workpath}/home
> -    set cmdstring "cd ${build.dir} && ulimit -s 16384 && \
> -                   unset LD_PREBIND LD_PREBIND_ALLOW_OVERLAP && \
> -                   HOME=${workpath}/home make"
> -    ui_debug "EXECUTING: $cmdstring"
> -    system $cmdstring
> -    set cmdstring "$cmdstring check"
> -    ui_debug "EXECUTING: $cmdstring"
> -    system $cmdstring
> -}
> +test.run            yes
> +test.target         check

The difference here is that before, "sudo port build clisp" would have included running "make check", and installation would have been aborted if any of the tests failed, whereas now users will have to explicitly run "sudo port test clisp" if they want the tests done. On the plus side that should make clisp faster to build, but it opens up the possibility that clisp will be built wrong on some unusual system and the problem not noticed. This is of course how most other ports work too, so I think it's fine. Some projects' developers are adamant that you must run the tests, but clisp's unix/INSTALL file does show the tests as optional (see step 8).