Native Elixir Fixes

[artman41]

• Cause failed compilation of .ex files to prevent creation of .app file
  • Replace wildcard for ex beams for a shell command
  • To prevent failed .ex compilation from continuing to create a $(PROJECT).app file, I've moved the elixir generation code to be ran by the actual shell. Doing this though meant the Module names of the compiled .ex files needed to be found separately. To solve this, I've tried to use a shell command to manually build the Module name based on the __aliases__ attribute in the elixir ast of the code
• Allow support for Elixir v1.13 (supported by OTP 23)
  • I'm happy to remove this commit from the branch if you don't want to support OTP 23 essen
• Replace my email
• Account for .ex files which reference mix.exs config
  • I noticed when using the dependency mint that it referenced the Mix.Project.config() at compile time, hence the need to load the mix.exs project during .ex compilation
• Simplify Ex compilation failure
  • As suggested, modified the compile_ex.erl to return the compiled modules for easier var binding. The erl code will now redirect the errors printed by 'Elixir.Kernel.ParallelCompiler':compile_to_path/2 to stderr & print to stdout only for the module names or the word _ERROR_ on a failed compilation

[essen]

You can't do that, the wildcard function operates on cache that is updated at the start of executing the target, so it won't know about the new Elixir beam files.

[artman41]

good point, leave it with me and I'll double check against what I have locally - could be the case that I need to rollback that change

[essen]

I need more info about the two most recent commits because I don't know what they do.

[artman41]

@essen tried a different way of getting the elixir beams, let me know what you think

I'll add an explanation under the commits in the PR description to say why it's needed 👍

[essen]

I don't think you need to go that far. It is much better to get the modules list from the compilation, we only need a way to signal success or failure. That could be signaled as a word in the output, which we would then check, before making MODULES be everything except that (first or last) word. The check would error out if failure is indicated.

See word related functions in https://www.gnu.org/software/make/manual/html_node/Text-Functions.html especially firstword, wordlist.

[artman41]

I don't think you need to go that far. It is much better to get the modules list from the compilation, we only need a way to signal success or failure. That could be signaled as a word in the output, which we would then check, before making MODULES be everything except that (first or last) word. The check would error out if failure is indicated.

See word related functions in https://www.gnu.org/software/make/manual/html_node/Text-Functions.html especially firstword, wordlist.

I feel dumb - I forgot entirely about firstword, glad you're able to offer a different perspective 😄

I'll revert that and get a change pushed up now 👍

Only thing I can think of that might be a problem though, shell will capture the compilation failure output - is there a way you can think of that would accurately redirect that to the shell?

Ignore the previous question - had a brain fart