https://sourceware.org/bugzilla/show_bug.cgi?id=22150
Bug ID: 22150
Summary: [2.29 Regression] ld.bfd keeps a version reference in
.gnu.version_r for symbols which are optimized out
Product: binutils
Version: 2.29
Status: NEW
Severity: normal
Priority: P2
Component: ld
Assignee: unassigned at sourceware dot org
Reporter: doko at debian dot org
Target Milestone: ---
[forwarded from https://launchpad.net/bugs/1715641]
test files at https://people.canonical.com/~doko/tmp/lp1715641.tar.xz
feel free to call that a regression or behavior change ;) ... This is seen when
building NetworkManager using ld 2.29 using glibc 2.26, the dpkg-shlibs tool
not seeing a versioned dependency on a glibc 2.26 symbol, and then the
packaging tools declaring NetworkManager be ready for use with glibc 2.24, and
then the dynamic loader complaining with
# ldd /usr/sbin/NetworkManager
/usr/sbin/NetworkManager: /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.25'
not found (required by /usr/sbin/NetworkManager)
The behaviour change between ld 2.28 and ld 2.29 seems to be
$ readelf -V NetworkManager-ld-2.28 | fgrep GLIBC_2.25
$ readelf -V NetworkManager-ld-2.29 | fgrep GLIBC_2.25
0x0170: Name: GLIBC_2.25 Flags: none Version: 11
$ readelf -V NetworkManager-gold-2.28 | fgrep GLIBC_2.25
000: 0 (*local*) 2 (GLIBC_2.25) d (CURL_GNUTLS_3) d
(CURL_GNUTLS_3)
0x0010: Name: GLIBC_2.25 Flags: none Version: 2
$ readelf -V NetworkManager-gold-2.29 | fgrep GLIBC_2.25
000: 0 (*local*) 2 (GLIBC_2.25) d (CURL_GNUTLS_3) d
(CURL_GNUTLS_3)
0x0010: Name: GLIBC_2.25 Flags: none Version: 2
$ objdump -T NetworkManager-ld-2.29|grep GLIBC_2.25
$ objdump -T NetworkManager-gold-2.29|grep GLIBC_2.25
0000000000000000 DF *UND* 0000000000000000 GLIBC_2.25
__explicit_bzero_chk
There is no behavior change in gold, and gold correctly marks the symbol as
undefined. ld-2.28 otoh is able to detect that the symbol isn't used, doesn't
emit a reference, but in 2.29 it keeps a reference in the .gnu.version_r
section. Is this expected behavior?
There is a bug in NetworkManager as well, as it defines a backup for the
explicit_bzero function which is new in glibc 2.25, but it fails to define that
for explicit_bzero_chk which is used when building with -D_FORTIFY_SOURCE.
--
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
bug-binutils mailing list
bug-binutils@gnu.org
https://lists.gnu.org/mailman/listinfo/bug-binutils
