On Thu, Dec 4, 2008 at 9:00 PM, Allan McRae <allan@archlinux.org> wrote:
Dan McGee wrote:
On Thu, Dec 4, 2008 at 8:48 PM, Dan McGee <dan@archlinux.org> wrote:
Fix the issue uncovered by FS#12344. In this instance, the dotglob shopt was being set in the build() function but never cleared, causing issues in the remaining parts of the makepkg script.
Signed-off-by: Dan McGee <dan@archlinux.org> --- 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 4cc255c..25e4cc8 100644 --- a/scripts/makepkg.sh.in +++ b/scripts/makepkg.sh.in @@ -675,6 +675,8 @@ run_build() {
# ensure all necessary build variables are exported export CFLAGS CXXFLAGS MAKEFLAGS CHOST + # save our shell options so build() can't override what we need + local $shellopts=$(shopts -p)
I could have used this as a litmus test to see if anyone actually tried this, but assume I put "shopt" (drop the s) here instead. Fixed locally. :)
Ha, i was just trying to figure out why I had not "shopts" command !
local ret=0 if [ "$LOGGING" = "1" ]; then @@ -695,6 +697,8 @@ run_build() { else build 2>&1 || ret=$? fi + # reset our shell options + eval "$shellopts"
if [ $ret -gt 0 ]; then error "$(gettext "Build Failed.")" -- 1.6.0.4
Strange problem.... Patch looks good to me.
With even stranger symptoms! Thayer build a package that had 2 .PKGINFO files in it. I didn't even know tar could do that. This actually caused goofy issues on gerolde when validating that the package's arch was the same as the repo it was going to - because the check basically cat's the .PKGINFO and looks for the arch setting. Instead of 'i686' it was getting 'i686\ni686'