netatalk

Hi,

I have no concrete idea.

2009/2/11 HAT <hat@...>:
> I'm trying to build CVS HEAD on Solaris 10.
>
> LD_LIBRARY_PATH=/opt/sfw/lib ; export LD_LIBRARY_PATH
> autogen.sh \
>    --enable-srvloc=/usr \
>    --with-ssl-dir=/usr/sfw \
>    --with-bdb=/opt/sfw \
>    --sysconfdir=/etc \
>    --with-logfile=/var/log/netatalk.log
> gmake

Try Solaris make.

>>Undefined                       first referenced
>> symbol                             in file
>>dirfd                               directory.o
>>ld: fatal: Symbol referencing errors. No output written to afpd
>>collect2: ld returned 1 exit status

That definitely is in libc:

frank@...$ nm /lib/libc.so | grep dirfd
[7054]  |    299048|        10|FUNC |GLOB |3    |13     |dirfd

> I wrote a patch.

Afaict it shouldn't be necessary.

> But, it still failed.
>
>>In file included from /usr/include/sys/systm.h:20,
>>                 from /usr/include/sys/conf.h:40,
>>                 from tpi.c:6:
>>/usr/include/sys/proc.h:202: error: syntax error before "p_t1_lgrpid"
>>/usr/include/sys/proc.h:203: error: syntax error before "p_tr_lgrpid"
>>/usr/include/sys/proc.h:337: error: syntax error before '}' token
>>/usr/include/sys/proc.h:345: error: syntax error before '*' token
>>/usr/include/sys/proc.h:349: error: syntax error before '*' token
>>/usr/include/sys/proc.h:350: error: syntax error before '*' token

Which netatalk files caused these include and errors?

Andrew reportedly can build CVS HEAD on Solaris 10 and I can build on
a stock Opensolaris 10 just fine. So there might be something with
your installation. Have you read Andrews original mail (afair from
December'08) where he wrote how some generated header file was
foobared ?

-Frank

-Frank

2009/2/11 Frank Lahm <franklahm@...>:
> Andrew reportedly can build CVS HEAD on Solaris 10 and I can build on
> a stock Opensolaris 10 just fine. So there might be something with
> your installation. Have you read Andrews original mail (afair from
> December'08) where he wrote how some generated header file was
> foobared ?

http://marc.info/?l=netatalk-devel&m=122841746024062&w=2

-Frank

Where you able to resolve this issue ?

-Frank

> Where you able to resolve this issue ?

Because I have been examining how to mangle for these several days,
I am not examining Solaris still so much.

It fails in any of /usr/ccs/bin/make, /usr/xpg4/bin/make and /usr/sfw/bin/gmake.

-- 
HAT

2009/2/16 HAT <hat@...>:
>> Where you able to resolve this issue ?
>
> Because I have been examining how to mangle for these several days,
> I am not examining Solaris still so much.

Its worth it: it now has ACLs and EAs support ! ;o)

It fails here:
>In file included from /usr/include/sys/systm.h:20,
>                 from /usr/include/sys/conf.h:40,
>                 from tpi.c:6:

...which really sounds very much like what Andrew saw:
http://marc.info/?l=netatalk-devel&m=122841746024062&w=2

-Frank

>...which really sounds very much like what Andrew saw:
>http://marc.info/?l=netatalk-devel&m=122841746024062&w=2

Oh!
I executed mkheaders with Solaris 9 some time ago.
I recalled it now.

I will try it after tomorrow because midnight in Japan.

-- 
HAT

Hi,

>> I'm trying to build CVS HEAD on Solaris 10.

>>>Undefined                       first referenced
>>> symbol                             in file
>>>dirfd                               directory.o
>>>ld: fatal: Symbol referencing errors. No output written to afpd
>>>collect2: ld returned 1 exit status

>That definitely is in libc:
>
>frank@...$ nm /lib/libc.so | grep dirfd
>[7054]  |    299048|        10|FUNC |GLOB |3    |13     |dirfd

The libc.so of Solaris 10 doesn't contain dirfd!

hat@...:~ >nm /lib/libc.so | grep dirfd
hat@...:~ >

not finded.

-- 
HAT

2009/4/7 HAT <hat@...>:
>>> I'm trying to build CVS HEAD on Solaris 10.
>
>>>>Undefined                       first referenced
>>>> symbol                             in file
>>>>dirfd                               directory.o
>>>>ld: fatal: Symbol referencing errors. No output written to afpd
>>>>collect2: ld returned 1 exit status
>
>>That definitely is in libc:
>>
>>frank@...$ nm /lib/libc.so | grep dirfd
>>[7054]  |    299048|        10|FUNC |GLOB |3    |13     |dirfd
>
> The libc.so of Solaris 10 doesn't contain dirfd!
>
> hat@...:~ >nm /lib/libc.so | grep dirfd
> hat@...:~ >
>
> not finded.

ACK. It's in Opensolaris, but in Solaris 10. It's not in SUS3.
We have to rewrite that differently. Will you take this?

-Frank

>> hat@...:~ >nm /lib/libc.so | grep dirfd
>> hat@...:~ >
>>
>> not finded.
>
>ACK. It's in Opensolaris, but in Solaris 10. It's not in SUS3.
>We have to rewrite that differently. Will you take this?

I hardly know Solaris.

usual technique:

#ifndef HAVE_DIRFD
#define dirfd(dir) (dir->dd_fd)
#endif /* HAVE_DIRFD */

I don know whether this method is right.

-- 
HAT

2009/4/8 HAT <hat@...>:
>>> hat@...:~ >nm /lib/libc.so | grep dirfd
>>> hat@...:~ >
>>>
>>> not finded.
>>
>>ACK. It's in Opensolaris, but in Solaris 10. It's not in SUS3.
>>We have to rewrite that differently. Will you take this?
>
> I hardly know Solaris.
>
> usual technique:
>
> #ifndef HAVE_DIRFD
> #define dirfd(dir) (dir->dd_fd)
> #endif /* HAVE_DIRFD */
>
> I don know whether this method is right.

I'd say the problem centers around whether calling fsync for a
directory fd is usefull on any other planet then planet penguin or
planet Apple. I guess its not. POSIX has nothing to say about fsycin'
dir fds.

-Frank