Summary: | -rpath fails with binutils 2.15.90.0.1.1-r3 | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Ingo Krabbe <ikrabbe.ask> |
Component: | [OLD] Development | Assignee: | Gentoo Toolchain Maintainers <toolchain> |
Status: | RESOLVED FIXED | ||
Severity: | major | CC: | arnetheduck |
Priority: | High | ||
Version: | unspecified | ||
Hardware: | x86 | ||
OS: | Linux | ||
URL: | http://sources.redhat.com/bugzilla/show_bug.cgi?id=570 | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: |
a small test program with library. You will need cweb for the make process.
rpath.tar.bz2 |
Description
Ingo Krabbe
2004-12-01 00:57:37 UTC
could you post a small test case ? does 2.15.90.0.3-r3 fail too ? Created attachment 45117 [details]
a small test program with library. You will need cweb for the make process.
I don't have the 2.15.90.0.3-r3 version here so it would be nice if you could
test that.
Created attachment 45244 [details]
rpath.tar.bz2
even simpler test app
the test application you attached was a mess ... no idea how to use it :p so i made one ... it builds a sample .so with a single func and then builds a small program which basically dlloads the .so, grabs the func with dlsym, and then runs the func $ ld --version GNU ld version 2.15.90.0.1.1 20040303 $ make cc -Wall -ldl -Wl,-rpath ./lib/ test-rpath.c -o test-rpath cc -shared -o libtest.so libtest.c root@vapier 0 rpath # make clean all test rm -f test-rpath libtest.so rm -f lib/* cc -Wall -ldl -Wl,-rpath ./lib/ test-rpath.c -o test-rpath cc -shared -o libtest.so libtest.c cp libtest.so lib/libtest-rpath.so ./test-rpath hello! $ readelf -d test-rpath | grep rpath 0x0000000f (RPATH) Library rpath: [./lib/] *** Bug 75914 has been marked as a duplicate of this bug. *** Bug 75914 contains some additional info (confirms the bug for amd64 at least=) and another simple test case (largely equivalent to this one). binutils-2.15.92.0.2-r1 is x86 stable now (technically, that leaves at least amd64 unresolved, unless I missed an ebuild update) that technically is probably correct, but non-x86 arches update their packages on their own schedule |