[Bug gold/18288] linker -s output suboptimal

[dilyan.palauzov at aegee dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=18288

dilyan.palauzov at aegee dot org <dilyan.palauzov at aegee dot org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ian at airs dot com
          Component|binutils                    |gold
           Assignee|unassigned at sourceware dot org   |ccoutant at gmail dot 
com

--- Comment #10 from dilyan.palauzov at aegee dot org <dilyan.palauzov at aegee 
dot org> ---
I put all the libraries under opedkim/opendkim/LIBS , compressed the file and
put it under http://mail.aegee.org/dpa/bz18288/opendkim-with-libs4.tar.xz .

When linking, only the crt-files are picked from /usr , the rest is taken from
the provided opendkim/opendkim/LIBS folder.


I compile with

gcc -LLIBS -Wl,--verbose -std=gnu99 -pthread -pipe -O3 -fno-fat-lto-objects
-flto -pthread -Wl,-O1 -Wl,-z -Wl,relro -Wl,-s -o .libs/opendkim
opendkim-opendkim.o opendkim-opendkim-ar.o opendkim-opendkim-arf.o
opendkim-opendkim-crypto.o opendkim-opendkim-db.o opendkim-opendkim-dns.o
opendkim-opendkim-lua.o opendkim-config.o opendkim-flowrate.o
opendkim-reputation.o opendkim-stats.o opendkim-test.o opendkim-util.o
../libopendkim/.libs/libopendkim.so ./LIBS/libtre.so ./LIBS/libmilter.a
./LIBS/libgnutls.so -lz ./LIBS/libtspi.so ./LIBS/libpthread.so
./LIBS/libtasn1.so ./LIBS/libnettle.so ./LIBS/libhogweed.so ./LIBS/libgmp.so
./LIBS/libdb.so ./LIBS/libopendbx.so ./LIBS/libmemcachedutil.so
./LIBS/libmemcached.so ./LIBS/libsasl2.so ./LIBS/libstdc++.so ./LIBS/liblua.a
./LIBS/libm.so ./LIBS/libunbound.so ./LIBS/libssl.so ./LIBS/libcrypto.so
../libvbr/.libs/libvbr.so ../librbl/.libs/librbl.so ./LIBS/libresolv.so
./LIBS/libdl.so ./LIBS/libstrl.so  -Wl,-rpath -Wl,/usr/lib64

I installed ld and gold from master (commit 8900d71e3880)

After compiling with ld (master) the resulting binary is 471 648 bytes and
strip does not change its size.  Doing readelf --all on the unstripped and
stripped binary has this diffs:

--- stripped    2015-04-23 13:36:57.879247346 +0000
+++ unstripped    2015-04-23 13:36:49.264247644 +0000
@@ -41,7 +41,7 @@
   [ 8] .rela.dyn         RELA             0000000000403aa0  00003aa0
        00000000000000c0  0000000000000018   A       4     0     8
   [ 9] .rela.plt         RELA             0000000000403b60  00003b60
-       0000000000001ff8  0000000000000018  AI       4    11     8
+       0000000000001ff8  0000000000000018  AI       4    23     8
   [10] .init             PROGBITS         0000000000405b58  00005b58
        0000000000000024  0000000000000000  AX       0     0     4
   [11] .plt              PROGBITS         0000000000405b80  00005b80

Using gold from master produces binary of size 475 880 bytes -> suboptimal
size.  The size is the same, when I use gold 2.25/1.11

On my system --hash-style=gnu is applied automatically.

-- 
You are receiving this mail because:
You are on the CC list for the bug.