Iptables incompatible with linux-headers!
I have been able to reproduce the strange phenomenon that a buildroot setting of
gcc-3.4.4
binutils-2.17
buildroot-20070514
uClibc-0.9.28.3
busybox-1.5.0
linux-headers-2.4.27
dropbear
iptables-1.3.7 (and 1.3.5)
cannot compile
iptables-1.3.7/extensions/libipt_string.c
due to non-existing structure members of ipt_string_info. However, when I finally got the idea to alter the value inside
package/iptables/iptables.mk
to explicitely replace the header reference by the directory
KERNEL_DIR=/tmp2/linux-2.4.27
where I separately compiled the kernel, the previous failure is changed into success.
Until today I have only used Buildroot for 2.6-kernels and I did never encounter the same failure. Would someone venture to call this a bug, or is it my handling that could have caused my troubles? Aside from one complaint on getopt.h when I aimed at a SuSv3-version, the above was the only trouble I had in this setting.
Regards, Mats Erik Andersson
Quoted 20 lines · Mats Erik Andersson
I have been able to reproduce the strange phenomenon that a buildroot setting of
gcc-3.4.4
binutils-2.17
buildroot-20070514
uClibc-0.9.28.3
busybox-1.5.0
linux-headers-2.4.27
dropbear
iptables-1.3.7 (and 1.3.5)cannot compile
iptables-1.3.7/extensions/libipt_string.cdue to non-existing structure members of
ipt_string_info. However, when I finally got the idea to alter the value inside
Quoted 8 lines
package/iptables/iptables.mkto explicitely replace the header reference by the directory
KERNEL_DIR=/tmp2/linux-2.4.27where I separately compiled the kernel, the previous failure is changed into success.
IIRC iptables needs to pickup some kernel-headers. If you use old versions of linux (old as in not in buildroot), then you should be able to pass KERNEL_DIR=/there/ to the toplevel make and all should be well. I did never try if this works, though, so it's possible that you have to tweak it a little bit (and send the resulting patch which makes sure that KERNEL_DIR is properly seen in package/* to the list).
Quoted 7 lines
Until today I have only used Buildroot for 2.6-kernels and I did never encounter the same failure. Would someone venture to call this a bug, or is it my handling that could have caused my troubles? Aside from one complaint on
getopt.hwhen I aimed at a SuSv3-version, the above was the only trouble I had in this setting.
I don't remember any details about this getopt thing, i fear..