On Thu, 2016-10-06 at 11:41 -0400, Alan Conway wrote: > On Thu, 2016-10-06 at 10:10 -0400, Andrew Stitcher wrote: > > > > On Thu, 2016-09-29 at 09:20 -0400, Alan Conway wrote: > > > > > > > > > On Thu, 2016-09-22 at 23:48 +0000, Jeff Donner wrote: > > > > > > > > > > > > > > > > Could you please add dependencies: > > > > > > > > rubygem-rspec > > > > rubygem-simplecov > > > > > > > > for Fedora 24 at least, to the installation documentation? > > > > Really this is not a documentation issue - it is a build issue. > > > > The CMake build should just not try to test ruby if these are not > > present (and in fact istr that there is code to do exactly this). > > > > The build should never just fail because of some lack of > > dependency. > > Either it should complain at the configure phase if the dependency > > is > > mandatory; or it should just not build whatever bit of proton > > depends > > on an absent optional dependency. > > > > Agreed, will fix and add a note to PROTON-570. It seems this has been > fixed, broken, re-fixed and re-broken. >
So here's the story: the build is already correct in that lack of optional components does not stop the whole build. The question is what constitutes "presence" of the optional ruby component. As it stands we skip ruby by default unless ruby-devel AND the rspec/simplecov gems are present. We error if you try to force BUILD_RUBY=ON and any of these are missing. Now we could make the tests optional, but that makes me very queasy. If someone builds the ruby binding on a new platform/version, then they really need to run the tests to make sure it actually works, so these gems are build requirements (if you are building ruby) not options. So I propose to leave this as a doc bug (fixed) but I would be happy to improve the cmake warning/error messages if anyone has suggestions. Thoughts? Alan. --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
