netatalk

On Sat, Oct 12, 2002 at 04:24:20AM +0000, didier wrote:

> Jens-Uwe Mager wrote:
> >You can ignore the hint on incoming calls and on info calls you can
> >simply set it to the constant kTextEncodingUTF8 0x08000103).
> Ok for unicode <--> unicode. But when the client or the server don't use 
> unicode.

Then they will probably not use AFP 3.0 but some older AFP protocol rev.
As far as my understanding goes MacOS X is the only AFP 3.0 client today
and this will always use kTextEncondingUTF8. I did ask the Apple AFP
folks and they said they never tested (and never expected to use) afp
3.0 without Unicode. The protocol spec makes one believe that this is
possible, but from my tests it does simply not work.

-- 
Jens-Uwe Mager	<pgp-mailto:62CFDB25>

On Sat, 12 Oct 2002 04:35:57 +0000, didier wrote:

> I've commited more changes and it should go a litle further with guest,
> afp_login_ext is not finished (but you can put big return and short cut
> password checking).
> Only ascii filenames,as utf8 and ascii are the same.

But aren't the UTF-8 pathnames completely different?

> Changes are not tested, I've taken a notebook without OS X for the
> weekend, but it compiles.

Tried it with

     --with-tcp-wrappers \
     --with-shadow \
     --enable-afp3 \
     --with-did=cnid \
     --with-catsearch \
     --with-mangling \
     --with-cdb \
     --without-logfile

but it stops (see make.txt). Then with DID=last

     --with-tcp-wrappers \
     --with-shadow \
     --enable-afp3 \
     --with-did=last \
     --without-logfile

it went through. The connection will be established via AFP 3.0 but drops
immediately (the clients issues a error 57), when logging in as guest:

    Oct 12 12:01:18 notebook-tk afpd[15110]: login noauth
    Oct 12 12:01:18 notebook-tk afpd[15110]: login nobody (uid 65534, gid
        65534) AFPX03
    Oct 12 12:01:20 notebook-tk afpd[15110]: Parsing volset (null)
    Oct 12 12:01:20 notebook-tk afpd[15110]: Parsing volset (null)
    Oct 12 12:01:22 notebook-tk afpd[15084]: server_child[1] 15110 killed

Then tried it with DHCAST128 and got both -5024 and -5023.

    Oct 12 12:07:28 notebook-tk afpd[15117]: bad function 3F
    Oct 12 12:07:28 notebook-tk afpd[15117]: logout
    Oct 12 12:07:28 notebook-tk afpd[15117]: 0.10KB read, 0.05KB written
    Oct 12 12:07:28 notebook-tk afpd[15084]: server_child[1] 15117 done

For the client's output see afp-debug.txt

Regards,

Thomas

Thomas Kaiser wrote:
> On Sat, 12 Oct 2002 04:35:57 +0000, didier wrote:
> 
> 
>>I've commited more changes and it should go a litle further with guest,
>>afp_login_ext is not finished (but you can put big return and short cut
>>password checking).
>>Only ascii filenames,as utf8 and ascii are the same.
> 
> 
> But aren't the UTF-8 pathnames completely different?
Yes but there's some code for decoding an returning UTF-8 pathnames.

> 

> 
> Tried it with
> 
>      --with-tcp-wrappers \
>      --with-shadow \
>      --enable-afp3 \
>      --with-did=cnid \
>      --with-catsearch \
>      --with-mangling \
>      --with-cdb \
>      --without-logfile
Did you make clean first?

> 
> it went through. The connection will be established via AFP 3.0 but drops
> immediately (the clients issues a error 57), when logging in as guest:
> 
>     Oct 12 12:01:18 notebook-tk afpd[15110]: login noauth
>     Oct 12 12:01:18 notebook-tk afpd[15110]: login nobody (uid 65534, gid
>         65534) AFPX03
>     Oct 12 12:01:20 notebook-tk afpd[15110]: Parsing volset (null)
>     Oct 12 12:01:20 notebook-tk afpd[15110]: Parsing volset (null)
>     Oct 12 12:01:22 notebook-tk afpd[15084]: server_child[1] 15110 killed
Can you mount the volume read-only. Can you make a traffic capture?

> 
> Then tried it with DHCAST128 and got both -5024 and -5023.
> 
>     Oct 12 12:07:28 notebook-tk afpd[15117]: bad function 3F
Fixed in cvs (was in 67 not 63) but it will always returns an error 
because it doesn't check the password.

>     Oct 12 12:07:28 notebook-tk afpd[15117]: logout
>     Oct 12 12:07:28 notebook-tk afpd[15117]: 0.10KB read, 0.05KB written
>     Oct 12 12:07:28 notebook-tk afpd[15084]: server_child[1] 15117 done
> 
> For the client's output see afp-debug.txt
>

There is test code under the util directory in libatalk.  This tests the basics, I'm just not very hot on forking or multithreading stuff.

Simon

OK, I suspected as much.  Perhaps its ok to just keep the file open, as like I say, the first call to the logger should be during connect, and before logon, and hence will be done as root (so this is the user that will open the file).

Simon