[Bug ld/18836] New: ld handles .gnu.lto_ prefixed sections oddly, /DISCARD/ semantics unclear

[rguenth at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=18836

            Bug ID: 18836
           Summary: ld handles .gnu.lto_ prefixed sections oddly,
                    /DISCARD/ semantics unclear
           Product: binutils
           Version: 2.25
            Status: NEW
          Severity: normal
          Priority: P2
         Component: ld
          Assignee: unassigned at sourceware dot org
          Reporter: rguenth at gcc dot gnu.org
  Target Milestone: ---

For t.s

        .section        .gnu.lto_.debug_info,"",@progbits
.hidden t.c.f0aceea6
.globl  t.c.f0aceea6
t.c.f0aceea6:
        .long   0x31
        .value  0x4
        .long   .Ldebug_abbrev1

        .section        .gnu.lto_.debug_abbrev,"",@progbits
.Ldebug_abbrev1:
        .uleb128 0x1

        .section        .gnu.lto_.debug_str,"MS",@progbits,1
        .string "/abuild/rguenther/trunk-g/gcc"

and the linker script

SECTIONS {
  .debug_info : { t.o(.gnu.lto_.debug_info) }
  .debug_abbrev : { t.o(.gnu.lto_.debug_abbrev) }
  .debug_str : { t.o(.gnu.lto_.debug_str) }
}

desired to link .gnu.lto_ debug-info sections from LTO input to a final
executable by linking them as regular debug-info sections like with

> as -o t.o t.s
> ld -o x script
ld: warning: cannot find entry symbol _start; defaulting to 0000000000400040

ld produces

> readelf -S x
There are 5 section headers, starting at offset 0x138:

Section Headers:
  [Nr] Name              Type             Address           Offset
       Size              EntSize          Flags  Link  Info  Align
  [ 0]                   NULL             0000000000000000  00000000
       0000000000000000  0000000000000000           0     0     0
  [ 1] .debug_str        PROGBITS         0000000000000000  00000040
       000000000000001e  0000000000000001  MS       0     0     1
  [ 2] .shstrtab         STRTAB           0000000000000000  0000005e
       0000000000000026  0000000000000000           0     0     1
  [ 3] .symtab           SYMTAB           0000000000000000  00000088
       0000000000000090  0000000000000018           4     2     8
  [ 4] .strtab           STRTAB           0000000000000000  00000118
       0000000000000020  0000000000000000           0     0     1

so it dropped all but .debug_str, presumably because of the default
/DISCARD/ : { *(.gnu.lto*) }
rule in effect for non-partial links.  But oddly this rule didn't apply
to .debug_str.

Also there is no way to "override" this DISCARD as it seems to apply
in "order" of the linker scripts(?) rather than in the order suggested
by -M:

Discarded input sections

 .gnu.lto_.debug_info
                0x0000000000000000        0xa t.o
 .gnu.lto_.debug_abbrev
                0x0000000000000000        0x1 t.o

Memory Configuration

Name             Origin             Length             Attributes
*default*        0x0000000000000000 0xffffffffffffffff

Linker script and memory map

...
.debug_info
 *(.debug_info .gnu.linkonce.wi.*)
 t.o(.gnu.lto_.debug_info)

.debug_abbrev
 *(.debug_abbrev)
 t.o(.gnu.lto_.debug_abbrev)

...
.debug_str      0x0000000000000000       0x1e
 *(.debug_str)
 t.o(.gnu.lto_.debug_str)
 .gnu.lto_.debug_str
                0x0000000000000000       0x1e t.o
...
/DISCARD/
 *(.note.GNU-stack)
 *(.gnu_debuglink)
 *(.gnu.lto_*)
OUTPUT(x elf64-x86-64)


which I hope is a bug.  Otherwise I'd need a new feature to be able to
get at inputs discarded sections...  (doing a partial link now as a
workaround, but that creates additional I/O and a quite big temporary file)

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