On Wed, Jan 31, 2001 at 02:23:50PM +0100, Schmitt, Martin wrote:
> Hi Dave and all!
>
> > Do a ps to find the pid of the other rsync process and see if
> > truss shows
> > it is still functioning.
> ...
> > As already pointed out add /usr/ccs/bin to your PATH, and
> > hopefully you'll
> > be able to find the compiler in /opt/SUNWspro/bin.
>
> Okay, after I found that gcc was already was in my path, I finally managed
> to compile my own rsync. Knowing the path to "make", it was actually quite
> straightforward.
>
> It looks different from the original problem. This time, rsync starts taking
> up 40 Megs (used to be 20) before beginning to loop.
At first I thought you were the person who had the binary compatibility
problem with solaris 2.4.5 & 2.4.6 binaries but in checking my email
trail I see I was confused. Perhaps you were wondering why I was talking
about version 2.4.5 in my Friday message to you. So now I believe what
changed for you is that you've switched from 2.4.1.
> ps -ef|grep rsync and truss on the parent and child pid returns:
>
> 19425: waitid(P_PID, 19428, 0xEFFFFBA8, WEXITED|WTRAPPED|WNOWAIT)
> (sleeping...)
> 19428: poll(0xEFFFC1D8, 1, 60000) (sleeping...)
> 19428: poll(0xEFFFC1D8, 1, 60000) = 0
> 19428: poll(0xEFFFC1D8, 1, 60000) (sleeping...)
>
> With the last 2 lines alternating roughly every Minute.
>
> Additionally, I executed rsync with "--dry-run" to see what would happen.
>
> 16518: read(0, " ;\v\0\0", 4) = 4
> 16518: poll(0xEFFFC3C8, 1, 60000) = 1
> 16518: write(1, "04\0\007 ;\v\0\0", 8) Err#32 EPIPE
> 16518: Received signal #13, SIGPIPE [caught]
> 16518: sigaction(SIGUSR1, 0xEFFFDEB0, 0xEFFFDF30) = 0
> 16518: sigaction(SIGUSR2, 0xEFFFDEB0, 0xEFFFDF30) = 0
> 16518: llseek(0, 0, SEEK_CUR) Err#29 ESPIPE
> 16518: _exit(20)
> 16515: waitid(P_PID, 16518, 0xEFFFFBA8, WEXITED|WTRAPPED|WNOWAIT) = 0
> 16515: ioctl(0, TIOCGPGRP, 0xEFFFFB64) Err#22 EINVAL
> 16515: getpgid(16518) = 460
> 16515: ioctl(0, TIOCGPGRP, 0xEFFFFB64) Err#22 EINVAL
> 16515: waitid(P_PID, 16518, 0xEFFFFBA8, WEXITED|WTRAPPED) = 0
> 16515: llseek(0, 0, SEEK_CUR) Err#29 ESPIPE
> 16515: _exit(20)
>
> What is this "caught SIGPIPE" in Line 4? Could it be that the underlying
> SSH2 is failing on me?
Yes, that could be.
> This was the output of truss on the source (remote end of rsync) machine.
>
> Once again, my Equipment: Twice Sun E450, SunOS 5.6, Commercial SSH2,
> current rsync 2.4.6 compiled from source.
>
> Any options that I might be able to set at compile time?
I don't think so.
On Wed, Jan 31, 2001 at 04:48:16PM +0100, Schmitt, Martin wrote:
> > What is this "caught SIGPIPE" in Line 4? Could it be that the
> > underlying
> > SSH2 is failing on me?
>
> I dug out an ancient post from Tridge where he talks about a socketpair()
> call that should be used instead of pipe() when rsyncing over SSH.
>
> http://lists.samba.org/pipermail/rsync/2000-January/001743.html
>
> Is this still valid?
Possibly. Not many people have reported using ssh2 on this mailing list
so I don't know what it uses. Can you try OpenSSH?
- Dave
