On Tue, 2007-02-06 at 16:25 -0600, Randy McMurchy wrote:
> Andrew Beverley wrote these words on 02/06/07 16:02 CST:
> 
> > Hopefully I'll get this in before 6.2 is released!
> 
> No promises. :-)
> 
> 
> > Is it possible to update iptables to 1.3.7? I've checked the 'current
> > development' version of the book and iptables is still at 1.3.5
> > 
> > I tried a few weeks ago to compile 1.3.7 on an LFS 6.2 system and found
> > that some of the modules didn't compile properly and weren't available
> > to use.
> 
> I'm guessing that you mean 1.3.5 in the last sentence. However, I
> just compiled and installed the package (1.3.5) and as best as I can
> tell, everything seemed to build. There were only a couple of minor
> warnings during the building, and no messages during the install
> other than the files being installed.

I've spent some time trying to get 1.3.7 working but get a "FATAL:
Module ip_tables not found" when trying to use a MARK target. I've sent
a request to the netfilter-devel list to try and get it working.

> At this point, without something better to go on other than "found
> that some of the modules didn't compile properly and weren't available
> to use", it's going to be hard to break the package freeze.

The modules weren't compiling because iptables was using a default
include path of /usr/src/linux/include. This doesn't exist in LFS so
needs to be specified when making iptables as follows:

make PREFIX=/usr LIBDIR=/lib BINDIR=/sbin KERNEL_DIR=/usr/src/linux-2...

If it's not specified the compilation of some modules are not even
attempted due to missing headers. As an example, ipt_recent, (which uses
extensions/.recent-test to check for the headers) is not built.

> > I can provide tests and/or try to replicate if required - I am just
> > writing this from memory.
> 
> If you could just identify what didn't work, that would be great.
> However, I don't really have any way of testing an IPTables
> installation right now, so I'm hesitant to update the book this late
> in the release cycle.

At this point I'm wondering why the BLFS book states to only use the
KERNEL_DIR variable if on a non-x86 architecture. I've tried 1.3.5,
1.3.6 and 1.3.7 and they all need the variable set, otherwise (as I
found) some of the modules do not get built because of the missing
headers.

So unless there is a really good reason for not using the KERNEL_DIR
variable, I would suggest that, at a minimum, the book is changed to
state that KERNEL_DIR should always be set.

There is my current problem with iptables 1.3.7 but it would also be
nice to see the version upped to 1.3.6 which I am content works (with
KERNEL_DIR set)

Many thanks,

Andy Beverley


-- 
http://linuxfromscratch.org/mailman/listinfo/blfs-dev
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Reply via email to