[Bug gold/18959] New: gold doesn't respect alignment of .rodata.str.* section

[koriakin at 0x04 dot net]

https://sourceware.org/bugzilla/show_bug.cgi?id=18959

            Bug ID: 18959
           Summary: gold doesn't respect alignment of .rodata.str.*
                    section
           Product: binutils
           Version: 2.26 (HEAD)
            Status: NEW
          Severity: normal
          Priority: P2
         Component: gold
          Assignee: ccoutant at gmail dot com
          Reporter: koriakin at 0x04 dot net
                CC: ian at airs dot com
  Target Milestone: ---

Here's a test case:

$ cat t.s
.section .rodata.str1.2, "aMS",1

.balign 2
.global a
a:
.asciz "ab"

.balign 2
.global b
b:
.asciz ""
$ as t.s -o t.o
$ ld.gold t.o -shared -o t.so
$ nm t.so
00000000000001fc R a
00000000000001ff R b
00000000000012b0 A __bss_start
0000000000001200 d _DYNAMIC
00000000000012b0 A _edata
00000000000012b0 A _end
$ ld.bfd t.o -shared -o t.so
$ nm t.so
00000000000001dc R a
00000000000001de R b
0000000000200290 D __bss_start
00000000002001e0 d _DYNAMIC
0000000000200290 D _edata
0000000000200290 D _end
0000000000200290 d _GLOBAL_OFFSET_TABLE_

b is properly aligned to 2 bytes with bfd ld, but not with gold.

This happens with any architecture, I found it while working on s390, but it's
readily reproducible with x86_64, i386, etc.

Could possibly be a reason for #18855 if (like s390) relocations with alignment
restrictions are involved, but I have no way of checking.

I'm going to try making a patch for this issue right now.

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