On Sun, Aug 14, 2011 at 09:00:44PM +1000, Allan McRae wrote:
On 14/08/11 06:07, Dave Reisner wrote:
Signed-off-by: Dave Reisner<dreisner@archlinux.org> --- bsdtar can't extract a tarball with the same path/file in it twice, but it will gladly pack it that way... odd. Credit for the AUR for finding this.
scripts/makepkg.sh.in | 4 ++++ 1 files changed, 4 insertions(+), 0 deletions(-)
diff --git a/scripts/makepkg.sh.in b/scripts/makepkg.sh.in index f464014..e5840f1 100644 --- a/scripts/makepkg.sh.in +++ b/scripts/makepkg.sh.in @@ -1540,6 +1540,10 @@ check_sanity() { error "$(gettext "%s file (%s) does not exist.")" "$i" "$file" ret=1 fi + if in_array "$file" "${source[@]}"; then + error "$(gettext "%s file found in source array: %s")" "$i" "$file" + ret=1 + fi done done
I am sure we already "fixed" this in the past. Did the fix get lost with some of that re-factoring that happened with handling install/changelog files?
Seems that this was dealt with in 64c325, and it caused some regressions which were dealt with in ac5c2f. And... now I can't reproduce this. It definitely still works [1]. Looking at the tarball again, all I see is insanity: -rw-r--r-- tealeg/users 1237 2011-06-20 09:15 shunit2/PKGBUILD -rw-r--r-- tealeg/users 367 2011-06-20 07:14 shunit2/shunit2.install hrw-r--r-- tealeg/users 0 2011-06-20 07:14 shunit2/shunit2.install link to shunit2/shunit2.install Not created by makepkg --source ... disregard...
As background, there was a big discussion a couple of years back about whether we should support the inclusion of install files (changelog was different then) in the source array. The majority opinion (of which I was not part...) then was to do so. So even though this patch goes with my line of thinking about the handling of install files, I think we should fix the actual issue and not pack the same file twice.