This message is in MIME format.  The first part should be readable text,
  while the remaining parts are likely unreadable without MIME-aware tools.
  Send mail to mime@docserver.cac.washington.edu for more info.

---559023410-824023566-946508457=:22340
Content-Type: TEXT/PLAIN; charset=US-ASCII

One of the ideas for helping the elf transition is to make a modified
ld.so for a.out. Well, to do that, we need to be able to make a.out
objects on an elf system.

I've got something which links, but doesn't work. And some strange things
are going on in the resulting binary which I don't understand.

From what I can tell, a whole lot of the symbols which get put into the
a.out file are at 0x1000 less than they were in the ELF version, even
though the disassembled code hasn't moved.

Here's what I'm doing now, I make all of the .o's using gcc. Then:

ld -o ld_aout.so.mrg -Ttext 0x1020   -L/usr/src/lib/libc/obj.i386
mdprologue.o rtld.o malloc.o shlib.o md.o vfprintf.o -lc_pic
crunchide -v -u ld_aout.so.mrg
objcopy -R .note -R .ident ld_aout.so.mrg ld_aout.so.mrg1
ld -o ld_aout.so --oformat a.out-i386-netbsd  -T \
/usr/src/libexec/ld.aout_so/elf_aout_merge.x -Bshareable \
-Bsymbolic -assert nosymbolic  ld_aout.so.mrg1

First, I use ld to make one object file, which hopefully contains all of
the code needed. Then I run a modified version of crunchide which prepends
"_" to all of the variable names. Third I use objcopy to get rid of the
.note and .ident sections. Lastly, I use ld to make an a.out file, merging
a bunch of the sections together, and doing the final linking.

I'm attaching the merge file to this note.

When I diff the results of objdump -dx for ld_aout.so.mrg1 and ld_aout.so,
I see a bunch of:

     1023:      53              pushl  %ebx
-    1024:      e8 00 00 00 00  call   1029 <__rtl+0x9>
+    1024:      e8 00 00 00 00  call   1029 <_rtld+0xfa1>
     1029:      5b              popl   %ebx

in the disassembly. When I look at the symbols, I see:

Name			old value		new value
__tcgetattr		0x00007e24		0x00006e24
_printf			0x000054f4		0x000044f4
_remove_search_dir	0x00003624		0x00002624

This list was done by had. Doing an automatic compare, I find that a few
symbols have the same value (___bss_start), and a few are off by other
than 0x1000.

I noticed that some symbols disapeared. I guess that the linker is
throwing away unused routines. That's fine. But the error in symbol values
don't correspond to just that. For instance, _binder (one of the early
routines) is 26b4 in the Elf, and 16b4 in the a.out. ??

Thoughts?

Take care,

Bill

---559023410-824023566-946508457=:22340
Content-Type: TEXT/PLAIN; charset=US-ASCII; name="elf_aout_merge.x"
Content-Transfer-Encoding: BASE64
Content-ID: <Pine.SOL.3.96.991229150057.22340C@marcy.nas.nasa.gov>
Content-Description: elf section merge-up file

U0VDVElPTlMNCnsNCiAgLyogUmVhZC1vbmx5IHNlY3Rpb25zLCBtZXJnZWQg
aW50byB0ZXh0IHNlZ21lbnQ6ICovDQogIC4gPSArIFNJWkVPRl9IRUFERVJT
Ow0KICAuaW50ZXJwIDogeyAqKC5pbnRlcnApIH0NCiAgLmhhc2ggICAgICAg
ICAgOiB7ICooLmhhc2gpCQl9DQogIC5keW5zeW0gICAgICAgIDogeyAqKC5k
eW5zeW0pCQl9DQogIC5keW5zdHIgICAgICAgIDogeyAqKC5keW5zdHIpCQl9
DQogIC5yZWwudGV4dCAgICAgIDogeyAqKC5yZWwudGV4dCkJCX0NCiAgLnJl
bGEudGV4dCAgICAgOiB7ICooLnJlbGEudGV4dCkgCX0NCiAgLnJlbC5kYXRh
ICAgICAgOiB7ICooLnJlbC5kYXRhKQkJfQ0KICAucmVsYS5kYXRhICAgICA6
IHsgKigucmVsYS5kYXRhKSAJfQ0KICAucmVsLnJvZGF0YSAgICA6IHsgKigu
cmVsLnJvZGF0YSkgCX0NCiAgLnJlbGEucm9kYXRhICAgOiB7ICooLnJlbGEu
cm9kYXRhKSAJfQ0KICAucmVsLmdvdCAgICAgICA6IHsgKigucmVsLmdvdCkJ
CX0NCiAgLnJlbGEuZ290ICAgICAgOiB7ICooLnJlbGEuZ290KQkJfQ0KICAu
cmVsLmN0b3JzICAgICA6IHsgKigucmVsLmN0b3JzKQl9DQogIC5yZWxhLmN0
b3JzICAgIDogeyAqKC5yZWxhLmN0b3JzKQl9DQogIC5yZWwuZHRvcnMgICAg
IDogeyAqKC5yZWwuZHRvcnMpCX0NCiAgLnJlbGEuZHRvcnMgICAgOiB7ICoo
LnJlbGEuZHRvcnMpCX0NCiAgLnJlbC5ic3MgICAgICAgOiB7ICooLnJlbC5i
c3MpCQl9DQogIC5yZWxhLmJzcyAgICAgIDogeyAqKC5yZWxhLmJzcykJCX0N
CiAgLnJlbC5wbHQgICAgICAgOiB7ICooLnJlbC5wbHQpCQl9DQogIC5yZWxh
LnBsdCAgICAgIDogeyAqKC5yZWxhLnBsdCkJCX0NCiAgLmluaXQgICAgICAg
ICAgOiB7ICooLmluaXQpCX0gPTANCiAgLnBsdCA6IHsgKigucGx0KSB9DQog
IC50ZXh0ICAgICAgOg0KICB7DQogICAgKigudGV4dCkNCiAgICAqKC5yb2Rh
dGEpDQogICAgKigucm9kYXRhMSkNCiAgICAqKC5nb3QxKQ0KICB9DQogIC5m
aW5pICAgICAgOiB7ICooLmZpbmkpICAgIH0gPTANCiAgLmN0b3JzICAgICA6
IHsgKiguY3RvcnMpICAgfQ0KICAuZHRvcnMgICAgIDogeyAqKC5kdG9ycykg
ICB9DQogIF9ldGV4dCA9IC47DQogIFBST1ZJREUgKGV0ZXh0ID0gLik7DQog
IC8qIFJlYWQtd3JpdGUgc2VjdGlvbiwgbWVyZ2VkIGludG8gZGF0YSBzZWdt
ZW50OiAqLw0KICAuID0gKC4gKyAweDBGRkYpICYgMHhGRkZGRjAwMDsNCiAg
LmRhdGEgICAgOg0KICB7DQogICAgKiguZGF0YSkNCiAgICAqKC5kYXRhMSkN
CiAgICAqKC5zZGF0YSkNCiAgICAqKC5zZGF0YTIpDQogICAgKiguZ290LnBs
dCkgKiguZ290KQ0KICAgICooLmR5bmFtaWMpDQogICAgQ09OU1RSVUNUT1JT
DQogIH0NCiAgX2VkYXRhICA9ICAuOw0KICBQUk9WSURFIChlZGF0YSA9IC4p
Ow0KICBfX2Jzc19zdGFydCA9IC47DQogIC5ic3MgICAgICAgOg0KICB7DQog
ICAqKC5zYnNzKSAqKC5zY29tbW9uKQ0KICAgKiguZHluYnNzKQ0KICAgKigu
YnNzKQ0KICAgKihDT01NT04pDQogIH0NCiAgX2VuZCA9IC4gOw0KICBQUk9W
SURFIChlbmQgPSAuKTsNCn0NCg==
---559023410-824023566-946508457=:22340--