[Slackbuilds-users] Are we there yet? (krb5)

[Christoph Willing]

On 22/11/2013, at 7:57 PM, Willy Sudiarto Raharjo <willysr at gmail.com> wrote:

>> sed -e "s at python2.5/Python.h@& python2.7/Python.h at g" \
>>    -e "s at -lpython2.5]@&,\n AC_CHECK_LIB(python2.7,main,[P
>> YTHON_LIB=-lpython2.7])@g" \
>>    -i configure.in &&
>> sed -e "s at interp->result at Tcl_GetStringResult(interp)@g" \
>>    -i kadmin/testing/util/tcl_kadm5.c &&
>> 
>> Makes it build. I am not sure of heath of the build right now, though.
> 
> Unfortunately that alone is not enough to make it build on my machine,
> but i have finally make it buildable here using the latest version of
> krb5, 1.11.4 and some TCL compatibility hack
> 
> Here's the changed files if you want to test it
> If you can confirmed that it's working there, i might push it to my
> branch and probably dig
> network/mod_auth_kerb: Removed (missing dep)
> network/openssh-krb5: Removed (missing dep)


I don't suggest that the patch is wrong but (just like the correction using sed) that style of change seems somewhat of an intrusion on the original source code. Although this code is "just" in the testing area of the source, we're discussing a security sensitive project. It is regularly updated yet the authors have chosen not to adopt this particular workaround. I would be very cautious about changing it.

An alternative to changing the code is to enable the existing tcl code to be accepted. This can be achieved by changing the empty "DEFS=" declaration in src/kadmin/testing/util/Makefile.in to be:
    DEFS= -DUSE_INTERP_ERRORLINE -DUSE_INTERP_RESULT
(both defined in /usr/include/tcl.h and so, in my mind, more or less official)

That change enables both the current SBo version (1.9.4) and the current kerberos release version (1.11.4) to compile to completion.

It doesn't address the python version problem but the change above to configure.in, then autoconf, certainly works and seems harmless.

chris