Pin Java version + add support for Java 9, 10, 11

[hadim]

@hanslovsky: Here is a first try to pin the Java version.

[conda-forge-linter]

Hi! This is the friendly automated conda-forge-linting service.

I wanted to let you know that I linted all conda-recipes in your PR (recipe) and found some lint.

Here's what I've got...

For recipe:

• The source section contained an unexpected subsection name. string is not a valid subsection name.

[conda-forge-linter]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[hadim]

@jakirkham & @sodre: does this approach make sense to you?

[sodre]

@hadim, if you are trying to build pyjnius against different versions of java, e.g. 8, 9, 10, 11, etc.. here is how I did it within the context of postgis and postgres.

1. I kept the postgis version and build strings intact as you would normally do with any package.

2. I added postgres as a host and run time dependencies in the meta.yaml file without any version information.

3. I created the conda_build_config.yaml file pinning the postgres versions.

4. Finally, I reran the conda smithy rerender command to get the additional ci_support files for CircleCI and Travis.

[sodre]

I just saw your conda_build_config.yaml and meta.yaml files. I think you just need to add the max_pin to conda_build_config.yaml, remove the build.string field, and rerun codna smithy rerender to get this to work.

[hadim]

@hanslovsky : ready for review/merge.

[hadim]

Good news: the next version of pyjnius will be compatible with Java >=9: kivy/pyjnius#363

[hadim]

This is a work in progress: DO NOT MERGE.

[hadim]

But since it does not depend on a compiler, why it is not available on some of them?

[sodre]

I think it is because it has not gone through the GCC7 Migration yet, e.g. the build number is <1000

[hadim]

Ok, so it's an easy fix I guess.

Can I migrate myself? Or do I have to wait for the bot?

[hadim]

I checked https://conda-forge.org/status/ and couldn't find openjdk in it. I guess it's because no compiler is declared in its recipe.

As already said the compiler is only declared in the test section. Is that why a compiler is declared in the yaml configuration?

A solution would be to remove tests that use C compiler, that would make the package compiler-independent. On the other side, I think it's a good practice to check that JNI compilation works.

[hadim]

This PR should fix the issue conda-forge/openjdk-feedstock#42

Once it's merged I will remove the hard-coded C compilers from this PR.

[hadim]

To do before merging here:

• Java 8: Java8 openjdk-feedstock#43
• Java 9: Java9 openjdk-feedstock#44
• Java 10: Java10 openjdk-feedstock#45
• Java 11: Java11 openjdk-feedstock#46
• Add more Java version to this recipe.
• Wait for pyjnius to release a new version.

[hadim]

It's almost good. Only OSX builds for toolchain_c are failing when solving env:

conda.exceptions.ResolvePackageNotFound: 
  - openjdk=10

It's weird because conda-forge has those builds. Example for Java 11: https://anaconda.org/conda-forge/openjdk/files?version=11.0.1. There is OSX builds for both toolchain_c and clang with main and gcc7 labels. I also checked inside those packages and info/hash_input.json contains (for the main labeled package):

{
  "c_compiler": "toolchain_c"
}

Does Anaconda repos have an index that could have been confused by the java branches I have created on the Java feedstock repo? See https://github.com/conda-forge/openjdk-feedstock/branches

[hadim]

I guess conda servers have been fixed because the missing OSX OpenJDK packages are back.

It's all green now.

Now waiting for a pyjnius release.

[hadim]

@sodre could you review, please?

[sodre]

@hadim this looks fine to me minus the build_number jinja2 expression and the fact that we are still pulling from github.

Keep in mind that I am not a maintainer of this package so I can't really merge this into the master branch.

[hadim]

@sodre Is that ok to release a dev version in master here? I feel it could take a while before we see a git tag on the pyjnius side.

Something like 1.1.3-dev0-GIT_HASH?

[conda-forge-linter]

Hi! This is the friendly automated conda-forge-linting service.

I wanted to let you know that I linted all conda-recipes in your PR (recipe) and found some lint.

Here's what I've got...

For recipe:

• Package version 1.1.3_dev0-b212d5 doesn't match conda spec

[conda-forge-linter]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[sodre]

@sodre Is that ok to release a dev version in master here? I feel it could take a while before we see a git tag on the pyjnius side.

Something like 1.1.3-dev0-GIT_HASH?

That is probably a question better asked to @conda-forge/core ...

[hadim]

I got a green light on gitter from @CJ-Wright to merge here.

@hanslovsky only can merge this PR and I know he is AFK for two weeks.

@CJ-Wright: could you merge this PR (if you can...)?

[CJ-Wright]

@hadim any interest in joining the maintainers of this package?

[hadim]

I will be once this PR is merged :-)

[CJ-Wright]

sorry I missed that bit