[pacman-dev] [PATCH] makepkg: Place packages symlinks in build dir when DESTDIR is used

[Ray Rashif]

2009/11/3 Eric Bélanger <snowmaniscool at gmail.com>

> On Tue, Nov 3, 2009 at 12:59 AM, Allan McRae <allan at archlinux.org> wrote:
> > Dan McGee wrote:
> >>
> >> On Mon, Nov 2, 2009 at 11:00 PM, Eric Bélanger <snowmaniscool at gmail.com
> >
> >> wrote:
> >>
> >>>
> >>> When DESTDIR is used, symlinks to the packages will be put in the build
> >>> directory.  This combines the convenience of a global package cache
> with the
> >>> ease of having a package (i.e. a symlink) in the build directory for
> testing
> >>> and installation purpose.
> >>>
> >>
> >> This seems really reasonable. It gets a +1 from me for the idea; I
> >> didn't test the code or anything though.
> >>
> >>
> >>>
> >>> Signed-off-by: Eric Bélanger <snowmaniscool at gmail.com>
> >>> ---
> >>>  scripts/makepkg.sh.in |    6 ++++++
> >>>  1 files changed, 6 insertions(+), 0 deletions(-)
> >>>
> >>> diff --git a/scripts/makepkg.sh.in b/scripts/makepkg.sh.in
> >>> index 92b0454..dc9124b 100644
> >>> --- a/scripts/makepkg.sh.in
> >>> +++ b/scripts/makepkg.sh.in
> >>> @@ -1000,6 +1000,8 @@ create_package() {
> >>>               "$PKGEXT" ; EXT=$PKGEXT ;;
> >>>       esac
> >>>       local
> >>> pkg_file="$PKGDEST/${nameofpkg}-${pkgver}-${pkgrel}-${PKGARCH}${EXT}"
> >>> +       local
> >>>
> pkglinks_name="${startdir}/${nameofpkg}-${pkgver}-${pkgrel}-${PKGARCH}${PKGEXT}"
> >>> +       local
> >>>
> pkglinks_target="$PKGDEST/${nameofpkg}-${pkgver}-${pkgrel}-${PKGARCH}${PKGEXT}"
> >>>
> >>
> >> Since we repeat the same
> >> "${nameofpkg}-${pkgver}-${pkgrel}-${PKGARCH}${PKGEXT}" part three
> >> times, we should really pull that into a local variable and then set
> >> the three others as necessary.
> >>
>
> Actually it was repeated twice (EXT vs. PKGEXT).  I removed one of my
> new variables. See new patch (inline and attached).
>
> >
> > On the gpg branch, there is (going to be...) a patch creating "tar_file"
> > (=current pkg_file)  and pkg_file (pkglinks_target here).  I think we can
> > get away with just those two varabiles...  see below.
> >
> >>
> >>>
> >>>       local ret=0
> >>>
> >>> @@ -1018,6 +1020,10 @@ create_package() {
> >>>               ret=$?
> >>>       fi
> >>>
> >>> +       if [ $ret -eq 0 ]; then
> >>> +               [ ! -f "${pkglinks_name}" ] && ln -s
> "${pkglinks_target}"
> >>> "${pkglinks_name}" && ret=$?
> >>>
> >>
> >> Will this work with err traps? I thought you need to be explicit with
> >> the if; then; fi blocks.
> >>
>
> I don't really know how err traps works but I've made the if statement
> explicit as it seemed that's what you prefered.
>
> >
> > How about testing if $PKGDEST is the same as $startdir and then make the
> > symlink using
> > ln -s ${pkg_file} ${pkgfile/$PKGDEST/$startdir/}
> >
>
> Thanks. I did that in my new patch.  I also added the -f option to ln.
>  So if you switch to undefined PKGDEST to defined PKGDEST, the package
> in the build directory will be overwritten by the symlink instead of
> having the ln error out.
>
> Eric
>
> =======================
> diff --git a/scripts/makepkg.sh.in b/scripts/makepkg.sh.in
> index 92b0454..e2b28c1 100644
> --- a/scripts/makepkg.sh.in
> +++ b/scripts/makepkg.sh.in
> @@ -1000,6 +1000,7 @@ create_package() {
>                 "$PKGEXT" ; EXT=$PKGEXT ;;
>        esac
>        local
> pkg_file="$PKGDEST/${nameofpkg}-${pkgver}-${pkgrel}-${PKGARCH}${EXT}"
> +       local
> pkglinks_target="${startdir}/${nameofpkg}-${pkgver}-${pkgrel}-${PKGARCH}${PKGEXT}"
>
>        local ret=0
>
> @@ -1018,6 +1019,13 @@ create_package() {
>                 ret=$?
>        fi
>
> +       if [ $ret -eq 0 ]; then
> +               if [ "$PKGDEST" != "${startdir}" ]; then
> +                       ln -sf "${pkg_file/$EXT/$PKGEXT}"
> "$pkglinks_target"
> +               fi
> +               ret=$?
> +       fi
> +
>        if [ $ret -ne 0 ]; then
>                error "$(gettext "Failed to create package file.")"
>                exit 1 # TODO: error code
> --
> 1.6.5.2
>
>
>
>
This is really convenient, but would it not be good if the symlink(s) are
removed upon --clean?