Based on an old email on llvm's ML - http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20161212/411904.html
- handle the R_X86_64_RELATIVE case for amd64 in lld.
With this patch the linker is able to link the postgresql
binary with dtrace.
Even if this isn't the proper solution, at least gives the
ability to compile portsgresql.
Bug report: http://s.verknowsys.com/ccc1cbd1fafb3d4e9138e10d73bee772.png
Old patch: http://lists.llvm.org/pipermail/llvm-commits/attachments/20161212/4d669479/attachment.bin
16:47 < dmilith> it linked with dtrace using lld
16:47 < dmilith> so that patch works
16:47 < dmilith> now let's try if dtrace agrees ;)
17:01 < dmilith> this is getting more and more interesting ;))
17:02 < dmilith> are hardlinks limited somehow?
17:02 < dmilith> DEBUG: creating and filling new WAL file
17:02 < dmilith> LOG: could not link file "pg_xlog/xlogtemp.12780" to "pg_xlog/000000010000000000000001":
Operation not permitted
17:02 < dmilith> FATAL: could not open file "pg_xlog/000000010000000000000001": No such file or directory
17:02 < dmilith> on initdb
17:02 < dmilith> I just love postgres build process.. heh
17:05 <@Zer0P> dmilith: yes
17:06 <@Zer0P> security.bsd.hardlink_check_uid: 1
17:06 <@Zer0P> security.bsd.hardlink_check_gid: 1
17:06 <@Zer0P> -> 0
17:06 <@Zer0P> security.bsd.hardlink_check_uid: Unprivileged processes cannot create hard links to files owned
by other users
17:06 < dmilith> ah!
17:06 < dmilith> great info! <3
17:08 < dmilith> like a charm..
17:08 <@Zer0P> so with the 2 liner patch, everything works fine? :P
17:09 < dmilith> yes
17:09 < dmilith> link passes, at least for postgresql