Andreas Fink wrote:
Here is how I configure
export PATH="/usr/X11R6/bin:/usr/bin:/bin:/usr/local/bin:/sbin:/usr/
sbin:/usr/local/sbin"
export LDFLAGS="-lz -L/usr/local/lib -Wl,-search_paths_first -Wl,-
single_module"
So why is the "-lz" necessary? The configure script should add that for
you.
export CFLAGS="-arch ppc -arch i386 -g -Os -fno-omit-frame-pointer -
Wl,-syslibroot,/Developer/SDKs/MacOSX10.4u.sdk -I/usr/local/include/
cairo"
I assume the -I/usr/local/include/cairo is for building GTK+, not for
building Wireshark.
Are the -Os and -fno-omit-frame-pointer just preferences for your build,
or are they necessary?
export CPPFLAGS=""
export CXXFLAGS="-arch ppc -arch i386 -g -Os fno-omit-frame-pointer -
fno-exceptions -fno-rtti -Wl,-syslibroot,/Developer/SDKs/
MacOSX10.4u.sdk -I/usr/local/include/cairo"
Is that for the libraries being built along with Wireshark? (There's
currently no C++ code in Wireshark, unless I've missed something.)
./configure --mandir=/usr/share/man --with-ssl --enable-threads --
enable-usr-local --with-lua --enable-adns --disable-dependency-tracking
Should the man pages be installed in /usr/share/man, or
/usr/local/share/man?
Problem #1: LDFLAGS make ./configure fail.
I would need to add -Wl,-single_module to LDFLAGS (otherwise we get
errors speaking about multi module)
The page at
http://developer.apple.com/releasenotes/DeveloperTools/RN-dyld/index.html
says
The static linker ld(1) supports building dylibs as either
-multi_module or -single_module. Conceptually, a multi-module dylib is
like a group of smaller dylibs - one per source file. The extra
meta-data in a multi-module dylib allowed (pre-10.4) dyld to delay
binding and running initializers on the individual modules. In Mac OS X
10.4 dyld ignores all multi-module meta-data and completely binds and
runs all initializers the same as if it were built single-module. In
general, building dylibs -single_module is preferred. The one case where
multi-module is still useful is it allows internal functions of a dylib
to be interposable. For instance, libSystem.dylib is built multi_module
so that malloc() can be overridden and all uses of malloc in libSystem
(e.g. strdup) would be redirected to use the overridden version.
so it appears that it's the right flag to use when building dylibs.
Unfortunately, it's *only* usable when building a dylib, which is why...
I can not do this like above because ./configure not wanting to compile.
...that fails. I'll look into whether there's a way to force it in
libtool, as that's probably the right way to ensure that it's used when
building only dylibs (although building plugins as bundles might have to
leave it out).
Problem #2: Missing LUA files
The files wslua/dtd_gen.lua and wslua/console.lua seem to be missing
(Make looks for them)
Copying them over from the previous install directory solves this.
Why are those missing?
Missing from SVN, or from the release tarball? I assume it's the
release tarball, as they're both in the epan/wslua directory; perhaps
epan/wslua/Makefile.am has to put them in the right macro.
Problem #3: IRDA plugin fails complaining about
_proto_reg_handoff_sir multiple defined.
...
fix: edit file packet-sir.c and rename the function
"proto_reg_handoff_sir to irda_proto_reg_handoff_sir
Or proto_reg_handoff_irsir(), to parallel proto_register_irsir().
and the only
call to it in proto_register_irsir as well
I've checked in a change to do the aforementioned rename - and to get
rid of the call in proto_register_irsir(), as handoff registration
routines are supposed to be called after *all* registration routines are
called.
