Sorry, a bit of confusion here... I should have cc'd my reply to Ulf's
message to openssl-dev... but if I'm not mistaken, this diff is covered
by
> If you copy the file crypto/sha/asm/sha1-586.pl from 0.9.5 to
> 0.9.5beta1 and rebuild it, does that break beta1?
Which being confused, I did it the other way around :-) which fixed
0.9.5... but I'll go back first thing this morning when I get in to work
and verify that this breaks beta1 too ;-)...
Hope this isn't too confusing for everybody else... (when should stuff
be kept private and when should it go to openssl-dev?)
keep up the good work,
Sean // [EMAIL PROTECTED]
Sean O Riordain <[EMAIL PROTECTED]>:
> I went back to the openssl-0.9.5beta1 and double checked... it does work
> when I do a plain "./config" etc...
>
> OpenSSL 0.9.5beta1 24 Feb 2000
> built on: Thu Mar 2 07:48:27 GMT 2000
> platform: FreeBSD
> options: bn(64,32) md2(int) rc4(idx,int) des(ptr,risc1,16,long)
> idea(int) blowfish(idx)
> compiler: gcc -DTERMIOS -DL_ENDIAN -fomit-frame-pointer -O3 -m486 -Wall
> -DSHA1_ASM -DMD5_ASM -DRMD160_ASM
> `test' is up to date.
>
> ==============
>
> For OpenSSL 0.9.5, I did a "./config 386" and edited the Makefile to add
> a "-g" to CFLAGs, did a make, make test... same problem... I ran it in
> debug and the screen dump is attached... (file t.t ;-)
[...]
> (gdb) n
> SHA1_Update (c=0xefbfdb90, data_=0xefbfdb8c, len=3230007500)
> at ../md32_common.h:415
[...]
> (gdb)
> 474 if (sw > 0)
> (gdb)
> 502 HASH_BLOCK_DATA_ORDER(c,data,sw);
> (gdb) n
>
> Program received signal SIGSEGV, Segmentation fault.
> 0x1f7a in .L000start ()
> (gdb)
What happens if you change crypto/sha/asm/sha1-586.pl as follows:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
diff -u -u -r1.4 sha1-586.pl
--- sha1-586.pl 2000/02/27 01:15:25 1.4
+++ sha1-586.pl 2000/03/02 18:56:42
@@ -392,7 +392,7 @@
&X_expand("esi");
&mov(&wparam(1),"esi");
- &set_label("shortcut", 1);
+ &set_label("shortcut");
&comment("");
&comment("Start processing");
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
After doing so, delete crypto/sha/asm/sx86unix.cpp (which is created
by that Perl script) and run "make" again. (You have to manually
delete that file because some dependencies are missing in the
Makefiles.)
This seems to be the only thing that changed in the SHA1 assembler
implementation, maybe the result confuses your assembler and/or linker.
Ulf Möller wrote:
>
> On Thu, Mar 02, 2000 at 08:34:45AM +0000, Sean O Riordain wrote:
>
> > I went back to the openssl-0.9.5beta1 and double checked... it does work
> > when I do a plain "./config" etc...
>
> Stranger and stranger.
>
> If you copy the file crypto/sha/asm/sha1-586.pl from 0.9.5 to
> 0.9.5beta1 and rebuild it, does that break beta1?
Ulf, that two char change does the trick... all fixed...
.
.
.
./sha1test
test 1 ok
test 2 ok
test 3 ok
.
.
.
OpenSSL self-test report:
OpenSSL version: 0.9.5
Last change: PKCS7_encrypt() was adding text MIME headers twice
beca...
OS (uname): FreeBSD isis.merrion.nua.net 2.2.8-STABLE FreeBSD
2.2.8-STABLE #0: Mon Dec 14 17:40:15 GMT 1998
[EMAIL PROTECTED]:/usr/src/sys/compile/NUASERVER i386
OS (config): i686-pc-freebsd2.2.8
Target (default): FreeBSD
Target: FreeBSD
Compiler: gcc version 2.7.2.1
Test passed.
