| Message ID | 1488550733-3956-9-git-send-email-wg@grandegger.com |
|---|---|
| State | Superseded |
| Headers | show |
Hello, On Fri, 3 Mar 2017 15:18:52 +0100, Wolfgang Grandegger wrote: > $ORIGIN/../../usr/bin is also a valid rpath for binaries in > "$hostdir/usr/bin". I guess you meant $ORIGIN/../../usr/lib and $hostdir/usr/lib here, right? Isn't this patch 8/9 actually a fix, which should come first in the series, and that we could/should apply quickly, independent from the rest? Thanks, Thomas
Hello, Am 03.03.2017 um 15:46 schrieb Thomas Petazzoni: > Hello, > > On Fri, 3 Mar 2017 15:18:52 +0100, Wolfgang Grandegger wrote: >> $ORIGIN/../../usr/bin is also a valid rpath for binaries in >> "$hostdir/usr/bin". > > I guess you meant $ORIGIN/../../usr/lib and $hostdir/usr/lib here, > right? "$ORIGIN/../../usr/lib" in ELF files in "$hostdir/usr/bin" *or* "$hostdir/usr/lib". IIRC, I think the problem was with some binaries in "usr/bin". > Isn't this patch 8/9 actually a fix, which should come first in the > series, and that we could/should apply quickly, independent from the > rest? Yes! I got it after an initial build. Can't remember what package made the problem. Going to a fresh re-make... Wolfgang.
Hello, Am 03.03.2017 um 15:56 schrieb Wolfgang Grandegger: > Hello, > > Am 03.03.2017 um 15:46 schrieb Thomas Petazzoni: >> Hello, >> >> On Fri, 3 Mar 2017 15:18:52 +0100, Wolfgang Grandegger wrote: >>> $ORIGIN/../../usr/bin is also a valid rpath for binaries in >>> "$hostdir/usr/bin". >> >> I guess you meant $ORIGIN/../../usr/lib and $hostdir/usr/lib here, >> right? > > "$ORIGIN/../../usr/lib" in ELF files in "$hostdir/usr/bin" *or* > "$hostdir/usr/lib". IIRC, I think the problem was with some binaries in > "usr/bin". > >> Isn't this patch 8/9 actually a fix, which should come first in the >> series, and that we could/should apply quickly, independent from the >> rest? > > Yes! I got it after an initial build. Can't remember what package made > the problem. Going to a fresh re-make... That's not correct. At the *second* make I get: *** ERROR: package host-libxml-parser-perl installs executables without proper RPATH: *** /work/agco/output/dcu_host/host/usr/bin/gtester *** /work/agco/output/dcu_host/host/usr/bin/xgettext *** /work/agco/output/dcu_host/host/usr/bin/msgexec *** /work/agco/output/dcu_host/host/usr/bin/msggrep *** /work/agco/output/dcu_host/host/usr/bin/msgcat *** /work/agco/output/dcu_host/host/usr/bin/toe *** /work/agco/output/dcu_host/host/usr/bin/msgconv *** /work/agco/output/dcu_host/host/usr/bin/infocmp In my case, this is due to the sanitized rpath from the previous make. With my config there is no ELF file with an initial rpath "$ORIGIN/../../usr/lib". Well, this check is not necessary at all after rpath sanitation. I'm going to remove it in my patch series. Wolfgang.
diff --git a/support/scripts/check-host-rpath b/support/scripts/check-host-rpath index 6ce547c..020c123 100755 --- a/support/scripts/check-host-rpath +++ b/support/scripts/check-host-rpath @@ -58,7 +58,7 @@ check_elf_has_rpath() { for dir in ${rpath//:/ }; do # Remove duplicate and trailing '/' for proper match dir="$( sed -r -e 's:/+:/:g; s:/$::;' <<<"${dir}" )" - [ "${dir}" = "${hostdir}/usr/lib" ] && return 0 + [ "${dir}" = "${hostdir}/usr/lib" -o "${dir}" = "\$ORIGIN/../../usr/lib" ] && return 0 done done < <( readelf -d "${file}" \ |sed -r -e '/.* \(R(UN)?PATH\) +Library r(un)?path: \[(.+)\]$/!d' \
$ORIGIN/../../usr/bin is also a valid rpath for binaries in "$hostdir/usr/bin". Signed-off-by: Wolfgang Grandegger <wg@grandegger.com> --- support/scripts/check-host-rpath | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)