On Wed, Aug 14, 2013 at 04:25:57AM +0900, Hiroki Sato wrote: > gj> +.It Va SRCREVISION > gj> +The revision of the > gj> +.Li src/ > gj> +tree to use. > gj> +Defaults to the current top of tree revision. > gj> +.It Va DOCREVISION > gj> +The revision of the > gj> +.Li doc/ > gj> +tree to use. > gj> +Defaults to the current top of tree revision. > gj> +.It Va PORTREVISION > > Why separation between revision and branch is needed? I > intentionally dropped this part from your old patch to > generate-release.sh because branch and revision number can be > specified like releng/9.2@NNN in a single variable. An incorrect > configuration of the two variables do not always cause a fatal error, > so specification in fewer number of variables is more foolproof. >
I personally do not like using branch@rNNNNNN, but will update to remove
the {SRC,DOC,PORT}REVISION variables.
> gj> +.It Va TARGET
> gj> +The target machine type for the release.
> gj> +Defaults to the current machine type.
> gj> +.It Va TARGET_ARCH
> gj> +The target machine architecture for the release.
> gj> +Defaults to the value of
> gj> +.Va TARGET .
> gj> +.Pp
>
> Please remove default configuration of these variables from
> release.sh. The reasonable defaults are already set in src/Makefile
> and setting TARGET_ARCH=$TARGET by default is simply wrong. Also,
> get_rev_branch() is redundant.
>
Will be changed shortly.
Why is get_rev_branch() redundant?
> gj> +Defaults to setting the number of
> gj> +.Xr make 1
> gj> +jobs
> gj> +.Pq Ar -j
> gj> +to half the number of CPUs available on the system.
>
> Did you try this on a uniprocessor machine?
Ugh. No, and now I see why it will not work.
Glen
pgpwh6WEnmRI1.pgp
Description: PGP signature
