OpenSSH on RH SPARC

Gregory Leblanc GLeblanc at cu-portland.edu
Thu Jul 6 15:09:20 EST 2000


> -----Original Message-----
> From: R P Herrold [mailto:herrold at owlriver.com]
> Sent: Wednesday, July 05, 2000 9:38 PM
> To: Gregory Leblanc
> Cc: OpenSSH List (E-mail)
> Subject: Re: OpenSSH on RH SPARC
> 
> On Wed, 5 Jul 2000, Gregory Leblanc wrote:
> 
> > Anybody made OpenSSH work on a SPARC machine yet?  
> <snip>
> > Now, I can make these errors go away by creating /etc/ssh, 
> but it seems that
> > since it's building into someplace, shouldn't it be 
> creating these darn
> > things there?  
> 
> ummm, if all you got was an install error, regarding the
> missing directory, make the directory, and re-install the RPMs
> -- no need to recomplie for that.  There is some error in the
> OpenSSH spec file you are using --- you don't specify which,
> and whose.  I am aware of at least two competing versions, in
> general circulation.

Whoops, sorry for being so vague.  I'm using the version from the OpenSSH
ftp site.  The maintainer is somebody at mindrot.org, so I figured this had
to be the right list, although I've never gotten an acknowledgement of any
of my emails to him.  Is the other RPM better, and if so, where do I get it?
The error that I'm getting is part of the RPM rebuild, and so the binary RPM
doesn't get written.  :-(  As I said, I can get a binary RPM by creating
that directory beforehand, but is that kosher?  I think that it should be
creating that in the Build-root directory, and NOT where the RPM is going to
be installed.  Maybe I should point this back at the RPM list...

> > BTW, anybody know why OpenSSL is so slow to build?  I think 
> I could almost
> > do a complete Linux kernel re-compile in the amount of time 
> it takes to
> > rebuild that RPM.
> 
> On the sparc, I think the numeric handling is not well up to
> snuff, compared to that which one might be used to on the i386
> line.  A lot is done in math routines rather than at the chip
> level, and it does seem to drag ...  
> 
> ... Unfortunately this carries over to runtime --- I am quite
> unhappy with sshd performance on the Sparcs.

Can you elaborate, off list if it's not appropriate?  Is this a limitation
with the Linux kernel, or with the OpenSSH code, or with the GCC compiler?
If it's with the kernel, are the *BSD's or Solaris any better?  If with the
compiler, have you tried any others?  If OpenSSH, uhm, bummer.  Not going to
try anything else, as OpenSSH kicks butt.  Maybe somebody should look at the
code, but it won't be me.  Thanks for the help,
	Grego





More information about the openssh-unix-dev mailing list