Yeah! Thanks a lot for your work. I fixed some things there, and everything should work properly now. Also I left qca-tls.. cause Psi must have SSL-connection-ability by default, that's why our OS is so secured :)
On 3/24/07, Ingo Schwarze <[EMAIL PROTECTED]> wrote:
Hi Vlad, hi Bernd,
Vlad Glagolev wrote on Sat, Mar 24, 2007 at 12:48:06PM +0300:
> On 3/23/07, viq <[EMAIL PROTECTED]> wrote:
>> On 15/03/07, Vlad Glagolev <[EMAIL PROTECTED]> wrote:
>>> Tested @ i386. Also the current maintainer of Psi for OpenBSD
>>> told me, that I could take Psi maintainership :-)
>> i386, seems to do everything as it should.
> Nice! I think it should be committed.
thanks a lot for your offer to take over the net/psi port
abandoned by Gerardo Santana.
Not paying enough attention on ports@, i did the same update last
week, so now we can compare notes. :-/
Please consider the following points:
- Are you *really* sure you can safely remove Bernd's patch
patch-iris_libidn_rfc3454_c? I still need that patch!
The file iris/libidn/rfc3454.c gets compiled with
gcc -c [...] -I. -I../../../../../../../local/include [...]
-I../iris/libidn -o .obj/rfc3454.o ../iris/libidn/rfc3454.c
In case the libidn package is installed, the preprocessor picks
up /usr/local/include/stringprep.h, which is *not* compatible
with iris/libidn/rfc3454.c, instead of the required
iris/libidn/stringprep.h.
- Removing patch-iris_xmpp-core_jid_cpp also means that the
include file /usr/local/include/stringprep.h gets used for
iris/xmpp-core/jid.cpp. As that one is only using
int stringprep(...)
typedef enum {...} Stringprep_profile_flags
and the stringprep-#defines
stringprep_nameprep
stringprep_xmpp_nodeprep
stringprep_xmpp_resourceprep
which are consistently defined across both versions of the
header, this *might* be safe for now, but i doubt using
two different versions of the same header file is wise.
- Why do you patch the files
psics.pri
src/psiaccount.cpp
iris/jabber/s5b.cpp
iris/xmpp-core/connector.cpp
The port works for me without these patches, and i fail to
imagine why they might be needed.
- The net/psi port can be built and run without qca-tls.
Therefore, i think that qca-tls should not be in LIB_DEPENDS.
What is more, there is also qca-sasl as an alternative,
even though it has not yet been ported to OpenBSD, so
forcing qca-tls on people might be a bad idea.
- Even if you should decide to add qca-tls a a dependency,
IMHO qca should not be removed from LIB_DEPENDS.
The qca libs is what we link against, so it should be there.
In theory, it could happen that qca might be removed
as a dependency from qca-tls at some future time, thus
breaking psi. Ok, i do admit this is not terribly probable
in this particular case, but anyway...
In case you put qca back to LIB_DEPENDS, don't forget to
remove it from WANTLIB.
- As long as growl is not ported to OpenBSD, i think
CONFIGURE_ARGS should have --disable-growl.
Imagine someone suddenly ports growl.
At that point, psi will probably pick up the dependency
without having it in *_DEPENDS!
In case we want growl later, we should enable it explicitely.
- The distinfo file should also have SHA256 now.
- The file patches/patch-configure needs updated line numbers.
Finally, i see three minor points:
- In the file
/usr/ports/infrastructure/templates/Makefile.template,
MASTER_SITES and EXTRACT_SUFX is behind PERMIT_*.
I see no reason to change the standard order.
- I suggest to put PORTHOME after the standard variables,
not somewhere in the middle.
- AFAIK, patching the $OpenBSD$ tags is not needed,
so i left that out of my diff.
My suggested patch to -current CVS is attached.
It is also available at
http://www.studis.de/Software/psi-0.10.diff
Could you cross-check and possibly resubmit?
Thank you very much for your work and your consideration,
and sorry for bursting in so late,
Ingo
--
Ingo Schwarze <[EMAIL PROTECTED]>
usta.de / studis.de system operation
-- Cheerz, Vlad / Stelz
update-psi--0.10.diff
Description: Binary data
