On 12/16/18 9:41 AM, Max Horn wrote:
Hi all,

thanks for the feedback, I have fixed the ARCHFLAGS now and made the code 
generally more robust, adding a new helper Fink::Services::get_sdkpath() which 
returns an empty string on Darwin < 18 and the SDK path as returned by `xcrun 
--sdk macosx --show-sdk-path`. I believe it would make sense to change that Darwin 
version cutoff to a lower value, but for now wanted to play it safe.

The perlmodule builds fine now, but python is still broken. Hanspeter hinted 
that normally it finds zlib; so I am guessing this might be another case where 
the SDK path needs to be inserted somewhere. I'll take a look.

In general, I noticed that in various .info files, we might have to start using 
the SDK path now. Perhaps time to introduce %{sdkpath} for that? Any concerns 
about doing just that?

E.g. in libcurl4.info, we have this in the PatchScript:

        /bin/cp /usr/include/lber*.h .
        /bin/cp /usr/include/ldap*.h .
        /bin/cp /usr/lib/libldap.dylib .

That is of course easy to adjust (and indeed, I just see that somebody did it). 
But there is also this in ConfigureParams:

ConfigureParams: <<
        --with-darwinssl \
        --without-ssl \
        --without-gnutls \
        --with-libssh2=%p \
        --with-gssapi=/usr \

and that last line could be easily turned into

        --with-gssapi=%{sdkpath}/usr \

fink-buildenv-modules currently can set SDK_PATH (tested through Xcode9) [1] but this could easily be folded into fink code if it would help more there. The above snippet --with-pkg=/path is very common throughout our .info jungle.


[1] https://github.com/nieder/fink-buildenv-modules/blob/master/modules/base.sh.in#L56

fink-core mailing list
List archive:
Subscription management:

Reply via email to