Mailing List Archive: ANNOUNCE: openssh-1.2.1pre25

ANNOUNCE: openssh-1.2.1pre25

Jan 6, 2000, 11:43 PM

Post #1 of 6 (1640 views)

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

1.2.1pre25 is out. Please use a mirror:

http://violet.ibs.com.au/openssh/files/MIRRORS.html

The following mirrors already have it:

ftp://ftp.localhost.ca/pub/openssh/files/
ftp://thermo.stat.ncsu.edu/pub/openssh/files/
http://www.firedrake.org/openssh/files/

Changes:

- - "Corrupted check bytes on input" when using triple DES has been
fixed

- - Added support for directory based lastlogs. This should make Irix
as functional as the other platforms.

- - Compilation fixes

- - Documentation updates

- - ssh-agent now properly cleans up after itself.

- - Beginnings of SCO support

Open Issues:

- - manpages on Solaris and other platforms

- - AIX status is unknown

- - snprintf for NeXT and older Solaris systems

- - Older Linux systems lack the poll() function. Replacement needed.

- - Connection stalls over forwarded connections

- - Hang on logout. Refer to TODO for details.

Detailed changelog:

20000107
- New config.sub and config.guess to fix problems on SCO. Supplied
by Gary E. Miller <gem@rellim.com>
- SCO build fix from Gary E. Miller <gem@rellim.com>

20000106
- Documentation update & cleanup
- Better KrbIV / AFS detection, based on patch from:
Holger Trapp <Holger.Trapp@Informatik.TU-Chemnitz.DE>

20000105
- Fixed annoying DES corruption problem. libcrypt has been
overriding symbols in libcrypto. Removed libcrypt and crypt.h
altogether (libcrypto includes its own crypt(1) replacement)
- Added platform-specific rules for Irix 6.x. Included warning that
they are untested.

20000103
- Add explicit make rules for files proccessed by fixpaths.
- Fix "make install" in RPM spec files. Report from Tenkou N. Hattori
<tnh@kondara.org>
- Removed "nullok" directive from default PAM configuration files.
Added information on enabling EmptyPasswords on openssh+PAM in
UPGRADING file.
- OpenBSD CVS updates
- [ssh-agent.c]
cleanup_exit() for SIGTERM/SIGHUP, too. from fgsch@ and
dgaudet@arctic.org
- [sshconnect.c]
compare correct version for 1.3 compat mode

20000102
- Prevent multiple inclusion of config.h and defines.h. Suggested
by Andre Lucas <andre.lucas@dial.pipex.com>
- Properly clean up on exit of ssh-agent. Patch from Dean Gaudet
<dgaudet@arctic.org>

19991231
- Added support for directory-based lastlogs
- Really fix typedefs, patch from Ben Taylor <bent@clark.net>

Regards,
Damien Miller

- --
| "Bombay is 250ms from New York in the new world order" - Alan Cox
| Damien Miller - http://www.mindrot.org/
| Email: djm@mindrot.org (home) -or- djm@ibs.com.au (work)

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.0 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iD8DBQE4dZkvormJ9RG1dI8RAoTzAJ4nkp65WW8hNO0alIrE8My0Rci0xQCglQ6m
Ls3xdOymKybAU+p795e4XnE=
=RWBk
-----END PGP SIGNATURE-----

Re: ANNOUNCE: openssh-1.2.1pre25 [ In reply to ]

gem at rellim

Jan 7, 2000, 11:42 AM

Post #2 of 6 (1627 views)

Permalink

Yo Damien!

It works much better now. I can confirm the "corrupted bytes" thing
is fixed on SCO.

Somehow part of my SCO patch did not make it in. The following is
required to make UnixWare 7.1 work in configure:

*-*-sysv*)
LIBS="$LIBS -lgen -lsocket"

cat >> confdefs.h <<\EOF
#define CRYPT_H_BREAKS_BUILD 1
#define HAVE_INNETGR 1
#define HAVE_U_INTXX_T 1
EOF

The "cat" part disappeared. Did you try to fix the tests for those
defines instead of just forcing the conditions as I proposed? If so maybe
we can work together on getting the tests fixed?

RGDWS
GARY

On Fri, 7 Jan 2000, Damien Miller wrote:

> 1.2.1pre25 is out. Please use a mirror:

RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 20340 Empire Ave, Suite E-3, Bend, OR 97701
gem@rellim.com Tel:+1(541)382-8588 Fax: +1(541)382-8676

Re: ANNOUNCE: openssh-1.2.1pre25 [ In reply to ]

djm at mindrot

Jan 7, 2000, 2:32 PM

Post #3 of 6 (1623 views)

Permalink

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Fri, 7 Jan 2000, Gary E. Miller wrote:

> Yo Damien!
>
> It works much better now. I can confirm the "corrupted bytes" thing
> is fixed on SCO.
>
> Somehow part of my SCO patch did not make it in. The following is
> required to make UnixWare 7.1 work in configure:
>
> *-*-sysv*)
> LIBS="$LIBS -lgen -lsocket"
>
> cat >> confdefs.h <<\EOF
> #define CRYPT_H_BREAKS_BUILD 1
> #define HAVE_INNETGR 1
> #define HAVE_U_INTXX_T 1
> EOF
>
> The "cat" part disappeared. Did you try to fix the tests for those
> defines instead of just forcing the conditions as I proposed? If so
> maybe we can work together on getting the tests fixed?

I would rather fix the tests so that these are auto-detected.

The CRYPT_H_BREAKS_BUILD isn't there any more.

Do you need to include any special libraries for the innetgr()
function?

Why did you need to define HAVE_U_INTXX_T again?

Regards,
Damien

- --
| "Bombay is 250ms from New York in the new world order" - Alan Cox
| Damien Miller - http://www.mindrot.org/
| Email: djm@mindrot.org (home) -or- djm@ibs.com.au (work)

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.0 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iD8DBQE4dml7ormJ9RG1dI8RAknoAJ9zL3dOA2X5v41wru5Aal/RFN+LCwCgmnMM
env3R1jUgHj5HLZbAw2YJf4=
=JGIH
-----END PGP SIGNATURE-----

Re: ANNOUNCE: openssh-1.2.1pre25 [ In reply to ]

gem at rellim

Jan 7, 2000, 2:55 PM

Post #4 of 6 (1623 views)

Permalink

Yo Damien!

On Sat, 8 Jan 2000, Damien Miller wrote:

> I would rather fix the tests so that these are auto-detected.
If you are up for it, then I am.

> The CRYPT_H_BREAKS_BUILD isn't there any more.
Confirmed. SCO works OK without it.

> Do you need to include any special libraries for the innetgr()
> function?
Just -lsocket

BTW, getspnam() is what requires -lgen

> Why did you need to define HAVE_U_INTXX_T again?

Otherwise I get this warning message a whole lot:
UX:acomp: WARNING: "./defines.h", line 77: typedef redeclared: u_int16_t
UX:acomp: WARNING: "./defines.h", line 78: typedef redeclared: u_int32_t

I hate warning messages because the bugs are usually hiding amoung them
somewhere...

The u_intXX_t is in /usr/include/sys/bitypes.h.

bitypes.h is pulled in directly by:
netdb.h
resolv.h
arpa/inet.h
arpa/nameser.h
net/if_dl.h
netinet/if_ether.h
netinet/in.h
netinet/in6.h
netinet/in6_f.h
sys/bitypes.h
sys/convsa.h
sys/socket.h
sys/un.h

And sys/socket.h is always called by defines.h

So, for at least the SCO case, if you included sys.socket.h in your
test for u_int16_t that should work.

Other traffic on this list has mentioned that this is a fatal error
on Solaris.

RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 20340 Empire Ave, Suite E-3, Bend, OR 97701
gem@rellim.com Tel:+1(541)382-8588 Fax: +1(541)382-8676

Re: ANNOUNCE: openssh-1.2.1pre25 [ In reply to ]

djm at mindrot

Jan 7, 2000, 3:04 PM

Post #5 of 6 (1628 views)

Permalink

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Fri, 7 Jan 2000, Gary E. Miller wrote:

> Yo Damien!
>
> On Sat, 8 Jan 2000, Damien Miller wrote:
>
> > I would rather fix the tests so that these are auto-detected.
> If you are up for it, then I am.
>
> > The CRYPT_H_BREAKS_BUILD isn't there any more.
> Confirmed. SCO works OK without it.
>
> > Do you need to include any special libraries for the innetgr()
> > function?
> Just -lsocket
>
> BTW, getspnam() is what requires -lgen

Are they detected OK now?

> > Why did you need to define HAVE_U_INTXX_T again?
>
> Otherwise I get this warning message a whole lot:
> UX:acomp: WARNING: "./defines.h", line 77: typedef redeclared: u_int16_t
> UX:acomp: WARNING: "./defines.h", line 78: typedef redeclared: u_int32_t
>
> I hate warning messages because the bugs are usually hiding amoung them
> somewhere...
>
> The u_intXX_t is in /usr/include/sys/bitypes.h.
>
> bitypes.h is pulled in directly by:
> netdb.h
> resolv.h
> arpa/inet.h
> arpa/nameser.h
> net/if_dl.h
> netinet/if_ether.h
> netinet/in.h
> netinet/in6.h
> netinet/in6_f.h
> sys/bitypes.h
> sys/convsa.h
> sys/socket.h
> sys/un.h

... but not sys/types.h? That seems broken.

> And sys/socket.h is always called by defines.h
>
> So, for at least the SCO case, if you included sys.socket.h in your
> test for u_int16_t that should work.

I now detect and include u_intXX_t declarations in sys/bitypes.h.

Regards,
Damien Miller

- --
| "Bombay is 250ms from New York in the new world order" - Alan Cox
| Damien Miller - http://www.mindrot.org/
| Email: djm@mindrot.org (home) -or- djm@ibs.com.au (work)

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.0 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iD8DBQE4dnDyormJ9RG1dI8RAqHsAJ981rNlY79er//cbjTkx+QbTOXDAQCgnnCH
69+a4np1J/OSp/sM2T5yEWg=
=YuKi
-----END PGP SIGNATURE-----

Re: ANNOUNCE: openssh-1.2.1pre25 [ In reply to ]

gem at rellim

Jan 7, 2000, 3:38 PM

Post #6 of 6 (1628 views)

Permalink

Yo Damien!

On Sat, 8 Jan 2000, Damien Miller wrote:

> > > Do you need to include any special libraries for the innetgr()
> > > function?
> > Just -lsocket
> >
> > BTW, getspnam() is what requires -lgen
>
> Are they detected OK now?
-lsocket is detected OK.
-lgen is added by the previous patch.

> > The u_intXX_t is in /usr/include/sys/bitypes.h.
[...]
> ... but not sys/types.h? That seems broken.

hehehe, It is SCO, what can I say?

> I now detect and include u_intXX_t declarations in sys/bitypes.h.
Great! I look forward to pre26!

RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 20340 Empire Ave, Suite E-3, Bend, OR 97701
gem@rellim.com Tel:+1(541)382-8588 Fax: +1(541)382-8676