smbsh progress

Edan Idzerda edan at mtu.edu
Sun Oct 4 17:15:18 GMT 1998


On Sun, 4 Oct 1998, Andrew Tridgell wrote:

> I did a bit more work on smbwrapper/smbsh today. The main new stuff
> is:

I did a little, too :)

 
> I think the next step is portability. Maybe a Solaris port?

That's what I was trying to see if I could get going.  I have
it working, almost, under Solaris 2.6.  At the moment 'ls'
seg faults if:

    $ ls /smb

    $ cd /; ls

But!  

    $ ls /

works.  The only thing promising so far is that I
did once get

    smbw_chdir (/smb)
    cleaning /smb
    cleaned /smb (fname=/smb cwd=/)
    parsed path name=/smb cwd=/ [MEEM#1D] [IPC$] []
    set SMBW_CWD to /smb
    setenv failed

Where "MEEM" is our workgroup name.  Whoa!  I take that back, I also
appear to have gotten a browse list back from Samba.  Neato.

'ls' is dying on readdir64.  My experience with implementing 
my own getdents() for a Solaris VFS was that getdents() HAD to
return a struct dirent64, so I wonder if this is related.
Actually, duh, maybe we just need to define an smb_readdir64
as readdir64() must be getting a smb fd.

At the moment, 'ls /smb' appears to be getting a browse list
from samba, when readdir64() dies and smbw.log is left with

[350] 00 00 00 00 00 00 00 00  00 00 03 10 01 00 16 07  ........ ........
[360] 00 00 50 45 52 43 48 32  32 00 00 00 00 00 00 00  ..PERCH2 2.......
[370] 00 00 00 00 03 10 01 00  17 07 00 00 50 45 52 43  ........ ....PERC
[380] 48 32 33 00 00 00 00 00  00 00 00 00 00 00 03

"PERCH22" is an NT machine, and PERCH23 should have been next.  So
the last call that smbwrapper.so was able to log was smbw_dir_open/dir_open.
I've messed around with this a little more than the above sounds
like, but that's all that's really interesting so far.  If I'm
wasting my time... hey, that's ok.  It's the weekend.  :)

- edan




More information about the samba-technical mailing list