xtables 1.31 compilation errors

xtables 1.31 compilation errors

[Mr Dash Four]

When trying to build the above from source I get the following error:

gcc -Wp,-MMD,./.libxt_gradm.oo.d,-MT,libxt_gradm.oo 
-D_LARGEFILE_SOURCE=1 -D_LARGE_FILES -D_FILE_OFFSET_BITS=64     
-D_REENTRANT -Wall -Waggregate-return -Wmissing-declarations     
-Wmissing-prototypes -Wredundant-decls -Wshadow -Wstrict-prototypes     
-Winline -pipe     -I../include  
-I/home/zeek/rpmbuild/BUILD/xtables-addons-1.31/extensions -DPIC -fPIC 
-O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions 
-fstack-protector --param=ssp-buffer-size=4 -m64 -mtune=core2 -o 
libxt_gradm.oo -c libxt_gradm.c;
libxt_gradm.c:84: warning: implicit declaration of function 'ALIGN'
libxt_gradm.c:84: error: initializer element is not constant
libxt_gradm.c:84: error: (near initialization for 'gradm_mt_reg.size')
libxt_gradm.c:85: error: initializer element is not constant
libxt_gradm.c:85: error: (near initialization for 
'gradm_mt_reg.userspacesize')
make[3]: *** [libxt_gradm.oo] Error 1
make[3]: *** Waiting for unfinished jobs....
make[3]: Leaving directory 
`/home/zeek/rpmbuild/BUILD/xtables-addons-1.31/extensions'
make[2]: *** [user-all-local] Error 2
make[2]: Leaving directory 
`/home/zeek/rpmbuild/BUILD/xtables-addons-1.31/extensions'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/home/zeek/rpmbuild/BUILD/xtables-addons-1.31'

Any ideas? From what I remember there was a similar problem with a 
couple of files in 1.30!

Re: xtables 1.31 compilation errors

[Jan Engelhardt]

On Saturday 2010-12-11 01:41, Mr Dash Four wrote:

> When trying to build the above from source I get the following error:
>
> gcc -Wp,-MMD,./.libxt_gradm.oo.d,-MT,libxt_gradm.oo -D_LARGEFILE_SOURCE=1
> -D_LARGE_FILES -D_FILE_OFFSET_BITS=64     -D_REENTRANT -Wall -Waggregate-return
> -Wmissing-declarations     -Wmissing-prototypes -Wredundant-decls -Wshadow
> -Wstrict-prototypes     -Winline -pipe     -I../include
> -I/home/zeek/rpmbuild/BUILD/xtables-addons-1.31/extensions -DPIC -fPIC -O2 -g
> -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector
> --param=ssp-buffer-size=4 -m64 -mtune=core2 -o libxt_gradm.oo -c libxt_gradm.c;
> libxt_gradm.c:84: warning: implicit declaration of function 'ALIGN'
> libxt_gradm.c:84: error: initializer element is not constant
> libxt_gradm.c:84: error: (near initialization for 'gradm_mt_reg.size')
> libxt_gradm.c:85: error: initializer element is not constant
> libxt_gradm.c:85: error: (near initialization for 'gradm_mt_reg.userspacesize')

> Any ideas? From what I remember there was a similar problem with a couple of
> files in 1.30!

Hooray. Another victim to ALIGN. The problem is not xtables-addons,
though. (Please complain to your distro if they have not updated
linux-glibc-devel yet.)
Further info:
http://www.google.de/search?q=ALIGN+xtables+kernel
http://bugs.gentoo.org/325257

Re: xtables 1.31 compilation errors

[Mr Dash Four]

>> Any ideas? From what I remember there was a similar problem with a couple of
>> files in 1.30!
>>     
>
> Hooray. Another victim to ALIGN. The problem is not xtables-addons,
> though. (Please complain to your distro if they have not updated
> linux-glibc-devel yet.)
> Further info:
> http://www.google.de/search?q=ALIGN+xtables+kernel
> http://bugs.gentoo.org/325257
>   
Erm, I am not sure! I have no problems compiling 1.30 and XT_ALIGN is 
used there. For reference:

uname -r = 2.6.34.7-61.fc13.x86_64
rpm -q glibc-devel = 2.12.1-4 (for both x86_64 and i686 as I am doing 
cross compilation a lot)

So, is there something wrong with the above versions, which needs to be 
fixed?

Re: xtables 1.31 compilation errors

[Mr Dash Four]

>>
>> Hooray. Another victim to ALIGN. The problem is not xtables-addons,
>> though. (Please complain to your distro if they have not updated
>> linux-glibc-devel yet.)
>> Further info:
>> http://www.google.de/search?q=ALIGN+xtables+kernel
>> http://bugs.gentoo.org/325257
> Erm, I am not sure! I have no problems compiling 1.30 and XT_ALIGN is 
> used there. For reference:
>
> uname -r = 2.6.34.7-61.fc13.x86_64
> rpm -q glibc-devel = 2.12.1-4 (for both x86_64 and i686 as I am doing 
> cross compilation a lot)
>
> So, is there something wrong with the above versions, which needs to 
> be fixed?
Further to the above: The 'problematic' line, which SHOULD be included 
in libxt_gradm.c is:
#include "compat_user.h" (it actually states in the header file 
"linux-glibc-devel 2.6.34 header screwup", so at least someone admits to 
a screw up for once!)

With this however, the problems are NOT fixed, I am getting this further 
down the line:

ipset_iphash.c: In function ‘iphash_create_init’:
ipset_iphash.c:42: error: ‘IP_NF_SET_HASHSIZE’ undeclared (first use in 
this function)
ipset_iphash.c:42: error: (Each undeclared identifier is reported only once
ipset_iphash.c:42: error: for each function it appears in.)
make[5]: *** [libipset_iphash.oo] Error 1
make[5]: *** Waiting for unfinished jobs....
make[5]: Leaving directory 
`/home/zeek/rpmbuild/BUILD/xtables-addons-1.31/extensions/ipset'
make[4]: *** [user-all-local] Error 2
make[4]: *** Waiting for unfinished jobs....
make[4]: Leaving directory 
`/home/zeek/rpmbuild/BUILD/xtables-addons-1.31/extensions/ipset'

Re: xtables 1.31 compilation errors

[Mr Dash Four]

> Further to the above: The 'problematic' line, which SHOULD be included 
> in libxt_gradm.c is:
> #include "compat_user.h" (it actually states in the header file 
> "linux-glibc-devel 2.6.34 header screwup", so at least someone admits 
> to a screw up for once!)
>
> With this however, the problems are NOT fixed, I am getting this 
> further down the line:
>
> ipset_iphash.c: In function ‘iphash_create_init’:
> ipset_iphash.c:42: error: ‘IP_NF_SET_HASHSIZE’ undeclared (first use 
> in this function)
> ipset_iphash.c:42: error: (Each undeclared identifier is reported only 
> once
> ipset_iphash.c:42: error: for each function it appears in.)
> make[5]: *** [libipset_iphash.oo] Error 1
> make[5]: *** Waiting for unfinished jobs....
> make[5]: Leaving directory 
> `/home/zeek/rpmbuild/BUILD/xtables-addons-1.31/extensions/ipset'
> make[4]: *** [user-all-local] Error 2
> make[4]: *** Waiting for unfinished jobs....
> make[4]: Leaving directory 
> `/home/zeek/rpmbuild/BUILD/xtables-addons-1.31/extensions/ipset'
...Also, IP_NF_SET_HASHSIZE is in ipset_ipportiphash.c, 
ipset_ipporthash.c and ipset_nethash.c, in neither of which 
IP_NF_SET_HASHSIZE is defined, apparently!

There is a definition of CONFIG_IP_NF_SET_HASHSIZE in ip_set.h which is 
set to 1024.

In 1.30 the value of 1024 is directly used, instead of 
IP_NF_SET_HASHSIZE (which SHOULD have been CONFIG_IP_NF_SET_HASHSIZE!), 
so someone (Jozsef?) needs to fix this.

I am amazed nobody checks these things - how do you compile this code 
people?

Re: xtables 1.31 compilation errors

[Jan Engelhardt]

On Saturday 2010-12-11 02:47, Mr Dash Four wrote:
>>
>> ipset_iphash.c: In function ‘iphash_create_init’:
>> ipset_iphash.c:42: error: ‘IP_NF_SET_HASHSIZE’ undeclared (first use in this
>> function)
>> ipset_iphash.c:42: error: (Each undeclared identifier is reported only once
>> ipset_iphash.c:42: error: for each function it appears in.)
>
>In 1.30 the value of 1024 is directly used, instead of
>IP_NF_SET_HASHSIZE (which SHOULD have been
>CONFIG_IP_NF_SET_HASHSIZE!), so someone (Jozsef?) needs to fix this.

Don't point fingers when you don't know who to blame.
This is not something Jozsef caused.

>I am amazed nobody checks these things - how do you compile this
>code people?

Releases are checked for whether the kernel modules compile with all
kernel versions. Userspace compilation testing is limited to what I
run. The rest is done happily by people more familiar with specific
choices of distros. That is how the ecosystem works, and it usually
works.

[Not to mention again that the IP_NF_SET_HASHSIZE issue does not appear
in a released tarball. So while everybody tries to avoid git states
that do not compile — it's a hindrance for git-bisect too — slips
do happen.]

Of course, there are also commercial support options available if you
prefer more QA.

Re: xtables 1.31 compilation errors

[Mr Dash Four]

>>> Of course, there are also commercial support options available if you
>>> prefer more QA.
>>>       
>> So, in order to get xtables to at least compile I have to get a
>> commercial support?! Really?! I dread the day you take over from
>> Patrick McHardy then - I really do!
>>     
>
> You mix up the Xtables core with Xtables-addons. The latter had the
> compile issue, but Patrick has nothing to do with Xtables-addons.
>   
You know exactly what I meant, so don't try to wriggle out of it.

> N.B.: To bring you up to date with the beer-free software: if someone
> feels a project needs more attention, they are to invest in it. If
> they do not come forward with patches, the public assumes they have
> other capabilities, and so, as a running gag, money spending is
> proposed.
>   
Save your beer-free-software teachings for someone else - you are in no 
position to try and educate me!

I've spotted 2 bugs in the current git tree of xtables (one of which was 
in the *released* version of ipset) and I asked for help on how to 
resolve them. In the meantime I did find a solution, which I posted earlier.

You tried to send me down blind alleys and basically acting smart a**e 
instead of providing a solution - the aforementioned bug with ALIGN 
wasn't something new to you - you already knew what the problem was and, 
most importantly, how to fix it, though you did not consider it 
necessary to let me know that, but to point me to google search and some 
obscure bug submission on gentoo instead.

>>> Releases are checked for whether the kernel modules compile with
>>> all kernel versions. Userspace compilation testing is limited to
>>> what I run. The rest is done happily by people more familiar with
>>> specific choices of distros. That is how the ecosystem works, and
>>> it usually works.
>>>       
>> The key word being 'usually' I presume, eh? Don't tell me you were
>> able to compile this code, because I won't believe you. I have
>> downloaded xtables with the *released* version of ipset (which
>> happens to be 4.5)
>>     
>
> I was meaning: a _released_ version _of xtables-addons_.
>   
The bug with IP_NF_SET_HASHSIZE was in the released version if ipset and 
not xtables and that is what I meant with my previous post.

> Yes, I know about the current git HEAD state. Happened to slip on
> compiling state 5bcdf7f. Big deal — Errare humanum est.
>   
I have no issue with errors being present in the git source tree - no 
software is ever perfect. I, however, have issues with you asking for 
'commercial support' (i.e. begging for money) in order to get a code 
which simply compiles and also sending me down blind alleys when you 
knew how to fix the problem I requested help for.

> Fixed now..
>   
So, wouldn't it have been easier for you to just point out how to fix 
this as soon as I posted about this bug, instead of harping on about 
commercial support, google search, gentoo bug submission and the 
difference between xtables and xtables-addons?!

Re: xtables 1.31 compilation errors

[Jan Engelhardt]

On Saturday 2010-12-11 02:34, Mr Dash Four wrote:
>
> Further to the above: The 'problematic' line, which SHOULD be included in
> libxt_gradm.c is:
> #include "compat_user.h" (it actually states in the header file
> "linux-glibc-devel 2.6.34 header screwup", so at least someone admits to a
> screw up for once!)

The screwup is on behalf of a79ff731a1b277d0e92d9453bdf374e04cec717a.
Me, I am just adding workarounds to xt-a.

> With this however, the problems are NOT fixed, I am getting this further down
> the line:
>
> ipset_iphash.c: In function ‘iphash_create_init’:
> ipset_iphash.c:42: error: ‘IP_NF_SET_HASHSIZE’ undeclared (first use in this
> function)

This is a different one, and not in 1.31, just git master -
and will be fixed soon.