root6 @6.04.12: internal GSL is not using the right compiler

[andre.david@…]

As per instructions in #49935 (sorry for littering the closed ticket):

sudo port uninstall root6
sudo port sync
sudo port install root6

fails with conflicting +python27 in root5 as expected, so

sudo port clean root6
sudo port install root6 +python34

fails with first attached log.

:info:build configure: error: in `/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_science_root6/root6/work/build/GSL-prefix/src/GSL-build':
:info:build configure: error: C compiler cannot create executables

Since I saw something like this before, I also tried

sudo port install root6 +python34 build.jobs=1

that fails with second attached log.

[andre.david@…]

The two logs mentioned in original submission

[andre.david@…]

I have also deactivated root5 to run

sudo port install root6

which seems to install a precompiled binary (that works). As soon as I ask for

sudo port install root6 +python34

and compilation is triggered, the same thing as the first log above happens.

[andre.david@…]

Just to report that

sudo port install root6 -gsl

has been compiling since some 10 minutes now (so I assume it will work) and this all looks similar to #50007.

failed: third.log attached.

[mojca (Mojca Miklavec)]

Please remember to CC the maintainer(s) of the port when filing new tickets.

ROOT 6 has just been upgraded to version 6.06.00 in r143637. On one hand this should hopefully solve the problems with GSL. On the other hand this problem:

:info:build In file included from /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_science_root6/root6/work/root-6-04-12/math/mathmore/src/GSLNLSMinimizer.cxx:19:
:info:build /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_science_root6/root6/work/root-6-04-12/math/mathmore/src/GSLMultiFit.h:146:38: error: no member named 'J' in 'gsl_multifit_fdfsolver'
:info:build       gsl_multifit_gradient(fSolver->J, fSolver->f,fVec);
:info:build                             ~~~~~~~  ^
:info:build /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_science_root6/root6/work/root-6-04-12/math/mathmore/src/GSLMultiFit.h:157:45: error: no member named 'J' in 'gsl_multifit_fdfsolver'
:info:build       int ret = gsl_multifit_covar(fSolver->J, kEpsrel, fCov);
:info:build                                    ~~~~~~~  ^

might have to be addressed in the same way as it was in r142498 for root5: by disabling mathmore when GSL is disabled.

I didn't test it yet, but quite likely problems with Python 3 persist. I'm not sure if that ever worked properly.

[cjones051073 (Chris Jones)]

You didn't do what I said to for the first two builds, which was to 'port select gcc none' first.

In any case please try again with 6.06.00, as the underlying issue should be fixed there.

[mojca (Mojca Miklavec)]

Chris, are you sure that you are writing this reply to the right ticket? My first impression was that the reporter had two different problems, both unrelated to the chosen gcc port.

[cjones051073 (Chris Jones)]

No, its the correct ticket. I still believe (until proven otherwise) the issue in the 'first' and 'second' logs are because the internal GCC build is picking up the gcc version the OP has 'port selected', which is GCC 5.x, which the internal GSL version does not support.

[mojca (Mojca Miklavec)]

I'm sorry, I forgot that you gave additional instructions in the previous ticket which led to this report.

I'm closing the ticket now. I believe that upgrade to 6.06.00 should have solved these problems as well as introduced new/unrelated ones that should be discussed separately (#50066).

In case that C compiler cannot create executables didn't disappear, please reopen the ticket. In all other cases please open a new ticket.