Re: [Fink-devel] Invariant splitoffs of variant packages

[Brendan Cully]

On Thursday, 29 December 2005 at 15:09, Daniel Macks wrote:
> Okay, let's resolve this...
> 
>   Package: foo-py%type_pkg[python]
>   Version: 1
>   Revision: 1
>   Type: python (2.3 2.4)
>   SplitOff: <<
> Package: foo-doc
>   <<
> 
> SplitOff packages are first-class package objects internally, with
> pointers that link parent and children and sibling packages. Fink and
> Debian use "package-version-revision" as a unique key for a package.
> 
> In CVS HEAD, a single "foo-1-1" package object is created in the
> database and it is arbitrarily assigned as the child of one of the
> foo-pyXX of which it is a SplitOff. That means it is functionally
> broken for all other foo-pyXY variants because foo-py23's child's
> parent might be foo-py24 instead. As a result, the foo-py23 build
> process tries to form foo-doc, foo-doc has incorrect data for that
> situation (%I, %D, and %b point to the non-existent foo-py24 paths,
> etc.).
> 
> In released versions of fink, the above .info fragment reportedly
> works correctly, and we have a few packages that have this construct.
> 
> As an under-the-hood core guy, it doesn't make sense to me that this
> should ever work or be supported, but I can see why a maintainer would
> expect it to work. Should we support it? Can we support it? Should we
> say "make a separate .info file for foo-doc" instead?

as an over-the-hood i-like-fink-because-it's-red-and-shiny guy, I have
no idea whether or how hard it is to support it, but in theory I like
the idea of a single doc package if the contents would be identical
anyway. separate .info works, but is more maintenance work and one
more thing to get out of sync when a new upstream is released.

I don't know exactly what "arbitrarily assigned as the child" means,
but I guess in theory it'd be cool if the splitoff could be the child
of all of the parent packages. monday, tuesday, friday with py23,
wednesday, thursday, saturday with py24, and alternate sundays it
switches, squid and whale style. Or alternatively, it could be the
child of whichever parent is building it?


---
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
for problems?  Stop!  Download the new AJAX search engine that makes
searching your log files as easy as surfing the  web.  DOWNLOAD SPLUNK!
http://ads.osdn.com/?ad_id=7637&alloc_id=16865&op=click
___
Fink-devel mailing list
Fink-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/fink-devel

[Fink-devel] compface help with patch

[Rogue Marechal]

Hi ML -

I thought I'd have a go at updating an existing patch for an abandoned
package (compface) as an exercise. I posted it in the tracker.

Anyway I got the thing to compile and as far as I can see working fine but
I'm sure the patch is less than perfect.

For a start I am getting a fuzz warning of 1. I read and reread the man
page and I more or less understand what that means but still I can't find
the offending spot. I guess finding it would help clear out a lot of
things.

If anyone can see me through this (it's a really small patch and fast
compile), I'd be eternally grateful.

RM



---
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
for problems?  Stop!  Download the new AJAX search engine that makes
searching your log files as easy as surfing the  web.  DOWNLOAD SPLUNK!
http://ads.osdn.com/?ad_id=7637&alloc_id=16865&op=click
___
Fink-devel mailing list
Fink-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/fink-devel

[Fink-devel] Need help with python linking

[Michèle Garoche]

Hello,

Would somebody know how to make the system find the right python  
framework in the case explained below:


It is with Mac OS X 10.4.3, XCode 2.2, latest libtool in fink 10.4- 
transitional unstable



The configure.ac file is as follows:

AC_ARG_ENABLE(python,
AC_HELP_STRING([--disable-python],[don't build the python plugin]),
python=$enableval,
python=yes
)
if test "$python" = yes; then
AC_PATH_PROG(pythonpath, python)
if test "_$pythonpath" = _ ; then
AC_PATH_PROG(pythonpath, python2)
fi
if test "_$pythonpath" = _ ; then
python=no
else
AC_MSG_CHECKING(Python version)
changequote(<<, >>)dnl
		PY_VER=`$pythonpath -c 'import distutils.sysconfig; print  
distutils.sysconfig.get_config_vars("VERSION")[0];'`
		PY_LIB=`$pythonpath -c 'import distutils.sysconfig; print  
distutils.sysconfig.get_python_lib(standard_lib=1);'`
		PY_INC=`$pythonpath -c 'import distutils.sysconfig; print  
distutils.sysconfig.get_config_vars("INCLUDEPY")[0];'`
		$pythonpath -c "import sys; map(int,sys.version[:3].split('.'))>=  
[2,2] or sys.exit(1)"

changequote([, ])dnl
AC_MSG_RESULT($PY_VER)
if test "$?" != "1"; then
AC_MSG_CHECKING(Python compile flags)
PY_PREFIX=`$pythonpath -c 'import sys; print 
sys.prefix'`
PY_EXEC_PREFIX=`$pythonpath -c 'import sys; print 
sys.exec_prefix'`
if test -f $PY_INC/Python.h; then
AC_MSG_RESULT([-I${PY_INC}])
AC_MSG_CHECKING(Python link flags)
PYSHARED=`$pythonpath -c 'import distutils.sysconfig;print  
distutils.sysconfig.get_config_var("LINKFORSHARED")'`

AC_MSG_RESULT([-L${PY_LIB}/config 
-lpython${PY_VER} ${PYSHARED}])
LIBS="$LIBS -L${PY_LIB}/config -lpython${PY_VER} 
${PYSHARED}"
CFLAGS="$CFLAGS -I$PY_INC"
else
python=no
AC_MSG_RESULT([Can't find Python.h])
fi
else
echo "Python too old. Only 2.2 or above is supported."
python=no
fi
fi
fi
if test "$python" = yes; then
AC_DEFINE(HAVE_PYTHON)
fi



It resolves at configure time as:

- Location of python

checking for python... /usr/bin/python

- Version of python
checking Python version... 2.3

- Compile flags

checking Python compile flags... -I/System/Library/Frameworks/ 
Python.framework/Versions/2.3/include/python2.3


Up to now everything seems OK.

- Linking flags

checking Python link flags... -L/System/Library/Frameworks/ 
Python.framework/Versions/2.3/lib/python2.3/config -lpython2.3 -u  
__dummy -u _PyMac_Error -framework System Python.framework/Versions/ 
2.3/Python -framework CoreServices -framework Foundation


That's where it fails at link stage, not finding Python.framework/ 
Versions/2.3/Python:


powerpc-apple-darwin8-gcc-4.0.1: Python.framework/Versions/2.3/ 
Python: No such file or directory




- Brute method to make it compiled.

Not sure it is good. Is it?

I've changed:

AC_MSG_RESULT([-L${PY_LIB}/config -lpython${PY_VER} ${PYSHARED}])
LIBS="$LIBS -L${PY_LIB}/config -lpython${PY_VER} 
${PYSHARED}"

to:

AC_MSG_RESULT([-L${PY_LIB}/config -lpython${PY_VER}])
LIBS="$LIBS -L${PY_LIB}/config -lpython${PY_VER} 
"


Cheers,
Michèle





---
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
for problems?  Stop!  Download the new AJAX search engine that makes
searching your log files as easy as surfing the  web.  DOWNLOAD SPLUNK!
http://ads.osdn.com/?ad_idv37&alloc_id865&op=click
___
Fink-devel mailing list
Fink-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/fink-devel

Re: [Fink-devel] net-snmp on Tiger

[Cian Hughes]

-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1

Yes, I found that out yesterday, I have tracked down the header I was  
missing (if_var.h) I just need to brush up on how to get the fink  
build-system to place a file into the source tree after un- 
compressing it, as opposed to just patching existing files.

I am glad to report that I have net-snmp 5.3 working on Mac OS X 10.4.
Regards,
Cian Hughes
[EMAIL PROTECTED]


On 2 Jan 2006, at 22:57, Martin Costabel wrote:


Cian Hughes wrote:

-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
Ok, so I'm currently trying to do two things with regard to net_snmp:
I need to figure out exactly what apple keeps in /System/Library/ 
Frameworks/System.framework/PrivateHeaders (which is referenced  
from http://darwinsource.opendarwin.org/10.4.3/net_snmp-18/Makefile


There are a couple hundred files in there, almost 20% of what is  
in /usr/include. Some of them are the same as in /usr/include, some  
are different, of interest for net_snmp mainly those in netinet and  
netinet6. You get them if you install and run darwinbuild opendarwin.org/projects/darwinbuild/>


Most of them are in the kernel tarball which you can download from

a few come from other projects, but these are probably not needed  
by net_snmp.


--
Martin


-BEGIN PGP SIGNATURE-
Version: GnuPG v1.4.1 (Darwin)

iD8DBQFDupDRaVVfOlCF0TQRAu3OAJ91LNDIghKN837gaDtZTan0AdZ7rACfXC9G
/8kpYizPDh4G95mLdxyriug=
=P/HJ
-END PGP SIGNATURE-


---
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
for problems?  Stop!  Download the new AJAX search engine that makes
searching your log files as easy as surfing the  web.  DOWNLOAD SPLUNK!
http://ads.osdn.com/?ad_id=7637&alloc_id=16865&op=click
___
Fink-devel mailing list
Fink-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/fink-devel