tcpdump mailing list archives

Re: 'bogus savefile header'


From: Guy Harris <guy () alum mit edu>
Date: Tue, 24 Aug 2010 17:46:23 -0700


On Aug 23, 2010, at 8:30 PM, Aaron Turner wrote:

So building the latest tcpdump from git and it won't link against the
latest libpcap from git:

ld: warning: in /usr/local/lib/libpcap.dylib, file was built for
unsupported file format which is not the architecture being linked
(i386)

running file against the actual file (libpcap.dylib is a symlink) returns:

/usr/local/lib/libpcap.1.2-PRE-GIT.dylib: Mach-O 64-bit dynamically
linked shared library x86_64

Just ran ./configure && make && make install for both and verified
that tcpdump is linking against the git version of libpcap.  I don't
see any special CFLAGS or configure args for the current
libpcap/tcpdump MacPorts builds (1.1.1/4.1.1) so maybe something
changed?

libpcap is building with: -arch x86_64 -arch i386

I tried it and it built libpcap 3-way fat (x86_64, i386, ppc), which is the default when building on Snow Leopard.  The 
default when building on Leopard is to build it 4-way fat; in both cases, it matches the way it's built in OS X.

What version of OS X are you building it on?  Did you set any environment variables when running the configure script 
for libpcap, in an attempt to override the default fat build?  Did you use "--disable-universal" to make it not build 
fat at all?

I tried:

CFLAGS="-arch x86_64 -arch i386" ./configure

for tcpdump, but still no worky.

If you used --disable-universal for libpcap, you have to use --disable-universal for tcpdump, otherwise it'll try to 
build a fat tcpdump, which will fail if the libpcap against which it's being linked isn't fat enough to support all the 
architectures for which tcpdump was built.

ideas?

Unless you have a strong reason *NOT* to build fat, just go with the flow and let the configure scripts figure out how 
to build stuff.

If you *DO* have a strong reason not to build fat, don't build libpcap *OR* tcpdump fat.

-
This is the tcpdump-workers list.
Visit https://cod.sandelman.ca/ to unsubscribe.


Current thread: