Re: gnustep-base tests

[Richard Frith-Macdonald]

> On 9 May 2016, at 19:57, Riccardo Mottola <riccardo.mottola@libero.it> wrote:
>> 

>>> 
>>> PS
>>> I build these packets in debian unstable under Vitualbox (amd64)
>> gnustep-make:
>> 
>> ./configure \
>>        --with-layout=fhs-system \
>>        --enable-native-objc-exceptions \

Native exceptions are enabled by default, so -enable-native-objc-exceptions is 
not needed (but harmless)

>>        --disable-strict-v2-mode
>> 
> 
> why do you need to disable v2 strict mode? Which packages fail? few to none 
> should.
> In case I think upstream packages have been updated and we can backport the 
> patch.

Definitely.  From the latest release notes:

   The '-enable-strict-v2-mode' option is now, after eight years, turned
on by default (in anticipation of finally removing backward
compatibility with version one).  WARNING; Packagers please ensure that
you update any old gnustep-make version one makefiles.

So -disable-strict-v2-mode is only for building/installing ancient software, 
and will cease to be an option.

>> gnustep-base:
>>        # Override the test for libkvm to ensure that /proc is used on
>>        # GNU/kFreeBSD even if libkvm-dev is installed (#593898).
>>    ./configure \
>>        ac_cv_lib_kvm_kvm_getenvv=no \
>>        --enable-libffi \
>>        --disable-openssl \
>>        --disable-bfd
> 
> disable-bfd ? disable-openssl? I never needed those options, don't even know 
> what the first does.
> Also, enable-libffi should not be needed: if it is present it gets used by 
> default. On contrary, if ffcall is preferred (on some rare platforms) you 
> need to explicit it.

Yes, those options are the default settings, so should have no effect (but are 
harmless).

I would encourage the use of --enable-bfd on development systems (it gives 
improved stacktraces for diagnostics), but it should be disabled (the default) 
by packagers, since it alters the license of the gnustep-base library (and 
things linked with it).