Apparently on my local fc33 VM I’m getting glade3-libgladeui-devel-3.8.6-8.fc32.x86_64 from rawhide (note: still fc32)
I need to do some more checking, possibly i shouldn’t be using rawhide at this time …
Re Docker image builds … I’m not capable of doing that at this time, so relying on you. Some more documentation on how to build Docker images for graphviz runners
would be appreciated.
That glade3-libgladeui-devel is listed under A.4. Removed packages, but I don’t know the relation between RHEL and Fedora versions so it might not be significant.
It turns out that we no longer use anything from that package, so i removed the yum install for it from ci/fedora33/Dockerfile. Will need you help again to rebuild the Docker image.
But I’m hitting some unrelated problem now on fedora33. i think this is in SWIG generated code, so probably a SWIG issue.
gv_ruby.cpp: In function ‘void SWIG_RubyInitializeTrackings()’:
gv_ruby.cpp:1267:85: error: call of overloaded ‘rb_define_virtual_variable(const char [21], VALUE (&)(…), NULL)’ is ambiguous
1267 | rb_define_virtual_variable(“SWIG_TRACKINGS_COUNT”, swig_ruby_trackings_count, NULL);
| ^
In file included from /usr/include/ruby/ruby.h:2863,
from /usr/include/ruby.h:33,
from gv_ruby.cpp:880:
/usr/include/ruby/backward/cxxanyargs.hpp:59:1: note: candidate: ‘void ruby::backward::cxxanyargs::rb_define_virtual_variable(const char*, VALUE ()(…), void ()(…))’
59 | rb_define_virtual_variable(const char *q, type *w, void_type *e)
| ^~~~~~~~~~~~~~~~~~~~~~~~~~
/usr/include/ruby/backward/cxxanyargs.hpp:90:1: note: candidate: ‘void ruby::backward::cxxanyargs::rb_define_virtual_variable(const char*, VALUE ()(…), void ()(VALUE, ID, VALUE*))’
90 | rb_define_virtual_variable(const char *q, type *w, rb_gvar_setter_t *e)
| ^~~~~~~~~~~~~~~~~~~~~~~~~~
make[4]: *** [Makefile:2531: libgv_ruby_la-gv_ruby.lo] Error 1
Maybe… But i think I triggered the problem by doing a distro-sync and possibly picking an older
version of SWIG without fully distcleaning my build tree.
A bit of clean up and a second attempt resulted in a clean build. So it was just a local problem, i think.