[Subversion-users] Can't execute svn commands through the network

Let me see if I understand:

- you have a working copy on a Samba network disk

- the Samba server is FreeBSD 8.2

- the Samba client is FreeBSD 9.0

On Linux I'd run "strace svn st -u" to find out which system call is
failing. Do you have strace installed? Perhaps truss can provide the
information?

It appears that your Samba setup doesn't support some POSIX sematics.
Perhaps mounting nobrl will help?

No I don't. strace is only for i386 while I'm running amd64. DTrace
isn't also an option for me, because I would need to rebuild my kernel.
But I'm just a beginner, so I don't think this would end well :)
I tried truss, but I can't do anything with the results. Maybe
they are interesting for you?

http://www.file-upload.net/download-4381573/truss.log.html
Unfortunately mount_smbfs doesn't support the nobrl option.

It seems odd in general to use samba between unix-type machines
instead of nfs - or just using ssh, remote X, or freenx to log into
the machine that has the disk space. Can you use nfs instead (you can
still share the same space with samba if there are also windows
clients) ?
Linux has a cifs mount option that preserves more of the unix
semantics than smb, but I don't know if that exists in freebsd.

It's something to do with locking:

open("/usr/home/michael/mount/pw-websites/PW-Websites/baumarkt.de/Website-online
/.svn/wc.db",O_RDWR|0x100000,00) = 3 (0x3)

[...]

fcntl(3,F_SETLK,0x7fffffffbaf0) = 0 (0x0)
fcntl(3,F_SETLK,0x7fffffffc2a0) = 0 (0x0)
fcntl(3,F_SETLK,0x7fffffffc2a0) ERR#11 'Resource deadlock avoid
ed'
fcntl(3,F_SETLK,0x7fffffffc2a0) = 0 (0x0)
write(2,"svn: E200030: sqlite: disk I/O e"...,37) = 37 (0x25)
write(2,"svn: E200030: sqlite: disk I/O e"...,37) = 37 (0x25)
I suspect that sqlite is attempting byte range locking and your server
doesn't support it. I don't know what to do about that.

Perhaps tell sqlite to use a different locking strategy by default?

(I know the default strategy can be set at compile-time. I don't recall
whether it can also be set later.)

Hello Michael,

Do you have a virus scanner involved?
Try to disable it, just for the tests.

I have seen the virus scanner interfere with
the sqlite engine of subversion 1.7.x in very interesing ways.
The new (1.7.0) sqlite-based client writes some journal file into
the working copy, which may fail if the virus scanner takes too long
to check the file.
In my case a subsequent fopen() call fails, which is being reported as sqlite
disk IO errors.

I have written a small C program that does
fopen()
write several k of data
fclose()
and repeat.
My box normally fails within 100..1000 iterations ... it takes some
time to trigger it, but *very* rarely it has made all 64000 iterations
without error.
Interestingly, it is fopen() that fails with EINVAL.

Mit freundlichen Grüßen
Dr.
Hartmut Niemann

Siemens AG
Infrastructure & Cities Sector
Rail Systems Division
Locomotives and Components
IC RL LOC EN LE 8
Werner-von-Siemens-Str. 67
91052 Erlangen, Deutschland
Tel.: +49 9131 7-34264
Fax: +49 9131 7-26254
mailto:[email protected]

Siemens Aktiengesellschaft: Vorsitzender des Aufsichtsrats: Gerhard Cromme; Vorstand: Peter Löscher, Vorsitzender; Roland Busch, Brigitte Ederer, Klaus Helmrich, Joe Kaeser, Barbara Kux, Hermann Requardt, Siegfried Russwurm, Peter Y. Solmssen, Michael Süß; Sitz der Gesellschaft: Berlin und München, Deutschland; Registergericht: Berlin Charlottenburg, HRB 12300, München, HRB 6684; WEEE-Reg.-Nr. DE 23691322