Hi Ben, On Thu, Jul 21, 2016 at 09:43:57PM +1000, Ben Woodcroft wrote: > On 21/07/16 11:42, Pjotr Prins wrote: > >Sorry for the new thread. But this time the patch is readable from all > >mailers (supporting UTF8). Please comment. > > Maybe to fix the $HOME issues by including a phase to set > environment variables during testing as in e.g. blast+ ? i.e. > > (setenv "HOME" "/tmp")
So we can assume /tmp is there? > Is there a reason to include 'patch' as a native-dependency, rather > than include them in the origin in 'patches' as usual? If you read my writeup on the packaging process at https://github.com/pjotrp/guix-notes/blob/master/ELIXIR.org you can see the build phase fails on introducing *any* patch at source. I am in the dark on the *why*, but my guess is that the mix build tool does some dark magic checking. > I'm left confused why the mix tests fail. If they run git say, git > is present and able to be used. Why doesn't it find it if it is in > the PATH? Is it because it attempts to use the network? Some of it is git - which I patched, see the writeup - but there also may be some network stuff going on. > Thanks for your work on this and please excuse any naivety as I am > completely ignorant of elixir. We are working on fixing texts, see https://github.com/elixir-lang/elixir/issues/5043 meanwhile I would like the package accepted as is. Only 11 tests out of 2000+ are skipped and we know they are minor. That means Elixir passes pretty much as is. 33 Mix tests are skipped, not much either (mostly git stuff). Mix is a package manager so I am not too concerned. In fact it is working fine in our development environments. We'll work with the community on closing these tests. Meanwhile it would be a good thing to accept the package so we can move forward in deployments. Pj.
