empty host section is discarded, host and build envs merged

[mbargull]

Actual Behavior

host and build environments are merged if requirements/host is present but empty, i.e., an empty list or None.

Expected Behavior

build and host should not be merged for all cases were host is present, as described in the docs:

The only time that build and host are merged is when the host section is absent, and no {{ compiler() }} jinja2 functions are used in meta.yaml.

Steps to Reproduce

(base) $ conda --version
conda 4.5.4
(base) $ conda-build --version
conda-build 3.10.2
(base) $ cat meta.yaml
package:
  name: pkg
requirements:
  build:
    - bzip2
  host:
(base) $ conda render .
package:
    name: pkg
requirements:
    build:
        - libgcc-ng 7.2.0 hdf63c60_3
        - bzip2 1.0.6 h9a117a8_4
    run:
        - bzip2 >=1.0.6,<2.0a0
extra:
    copy_test_source_files: true
    final: true

(base) $ echo '    - unzip' >> meta.yaml 
(base) $ cat meta.yaml 
package:
  name: pkg
requirements:
  build:
    - bzip2
  host:
    - unzip
(base) $ conda render .
package:
    name: pkg
requirements:
    build:
        - libgcc-ng 7.2.0 hdf63c60_3
        - bzip2 1.0.6 h9a117a8_4
    host:
        - libgcc-ng 7.2.0 hdf63c60_3
        - unzip 6.0 h611a1e1_0
extra:
    copy_test_source_files: true
    final: true

Expected behavior: No run: ['bzip2 >=1.0.6,<2.0a0'] in both cases.

Output of conda info

     active environment : base
    active env location : /opt/conda
            shell level : 1
       user config file : /root/.condarc
 populated config files : 
          conda version : 4.5.4
    conda-build version : 3.10.2
         python version : 3.6.5.final.0
       base environment : /opt/conda  (writable)
           channel URLs : https://repo.anaconda.com/pkgs/main/linux-64
                          https://repo.anaconda.com/pkgs/main/noarch
                          https://repo.anaconda.com/pkgs/free/linux-64
                          https://repo.anaconda.com/pkgs/free/noarch
                          https://repo.anaconda.com/pkgs/r/linux-64
                          https://repo.anaconda.com/pkgs/r/noarch
                          https://repo.anaconda.com/pkgs/pro/linux-64
                          https://repo.anaconda.com/pkgs/pro/noarch
          package cache : /opt/conda/pkgs
                          /root/.conda/pkgs
       envs directories : /opt/conda/envs
                          /root/.conda/envs
               platform : linux-64
             user-agent : conda/4.5.4 requests/2.18.4 CPython/3.6.5 Linux/4.14.39-1-MANJARO debian/9 glibc/2.24
                UID:GID : 0:0
             netrc file : None
           offline mode : False

[msarahan]

I think #2882 should fix this. Can you try it?

[mbargull]

Same result with latest master:

(base) $ conda-build --version                                                                                                                                                                                                                                                   
conda-build 3.10.2+8.g44821450
(base) $ cat meta.yaml                                                                                                                                                                                                                                                           
package:
  name: pkg
requirements:
  build:
    - bzip2
  host:
(base) $ conda-render .                                                                                                                                                                                                                                                          
package:
    name: pkg
requirements:
    build:
        - libgcc-ng 7.2.0 hdf63c60_3
        - bzip2 1.0.6 h9a117a8_4
    run:
        - bzip2 >=1.0.6,<2.0a0
extra:
    copy_test_source_files: true
    final: true

(base) $ echo '    - unzip' >> meta.yaml                                                                                                                                                                                                                                         
(base) $ cat meta.yaml                                                                                                                                                                                                                                                           
package:
  name: pkg
requirements:
  build:
    - bzip2
  host:
    - unzip
(base) $ conda render .                                                                                                                                                                                                                                                          
package:
    name: pkg
requirements:
    build:
        - libgcc-ng 7.2.0 hdf63c60_3
        - bzip2 1.0.6 h9a117a8_4
    host:
        - libgcc-ng 7.2.0 hdf63c60_3
        - unzip 6.0 h611a1e1_0
extra:
    copy_test_source_files: true
    final: true

[msarahan]

OK, thanks. I'll take another look.

[msarahan]

Sorry, this is not going to be fixed. Empty keys are stripped from recipes. I can't keep an empty host section. If you need to specify that the build and host envs should not be merged, and you're not using the new {{compiler()}} functions, you should use the (currently undocumented) build/merge_build_host key in meta.yaml.

[mbargull]

Sorry, this is not going to be fixed. Empty keys are stripped from recipes. I can't keep an empty host section.

Hm... Not the answer I was hoping for.

---

But regardless of that decision, I'll have to ask you to reopen the issue (or open another one) since build/merge_build_host doesn't seem to work for the False case:

(base) $ cat > meta.yaml <<'EOF'
> package:
>   name: pkg
> build:
>   merge_build_host: False
> requirements:
>   build:
>     - bzip2
>   host:
> EOF
(base) $ conda-build --version
conda-build 3.10.2
(base) $ conda-render .
package:
    name: pkg
build:
    merge_build_host: 'False'
requirements:
    build:
        - libgcc-ng 7.2.0 hdf63c60_3
        - bzip2 1.0.6 h9a117a8_4
    run:
        - bzip2 >=1.0.6,<2.0a0
extra:
    copy_test_source_files: true
    final: true

(base) $ cat > meta.yaml <<'EOF'
package:
  name: pkg
build:
  merge_build_host: False
requirements:
  build:
    - bzip2
  host:
EOF
(base) $ conda-build --version
conda-build 3.10.2+15.g7d407c5f
(base) $ conda-render .
--------------
Hash contents:
--------------
{}
----------
meta.yaml:
----------
package:
    name: pkg
build:
    merge_build_host: 'False'
requirements:
    build:
        - libgcc-ng 7.2.0 hdf63c60_3
        - bzip2 1.0.6 h9a117a8_4
    run:
        - bzip2 >=1.0.6,<2.0a0
extra:
    copy_test_source_files: true
    final: true

Given that its value is being rendered as 'False', I assume it's not parsed correctly.
(I even tried merge_build_host: '' to get another "falsy" value, but that one got completely stripped from the output, too.)

[msarahan]

It is parsed correctly, it's a bug elsewhere. The build section is still being treated as the host section (this is why you're getting the run_exports). I think you might need to use

build:
  ignore_run_exports:
    - bzip2

The problem is that your use case runs head-long into the use case of maintaining backwards compatibility with old recipes that have not yet split build and host. I might be able to add more checks for merge_build_host, but I'd also really like to understand why this edge case is necessary.

[mbargull]

but I'd also really like to understand why this edge case is necessary.

It offers consistency in the build process (e.g., always having those two envs active, weak run_exports not from build) and also consistency in the recipes. That way we can tell the users, for non-legacy recipes, to always use/add requirements/host instead of requirements/build and use build really ever only for compilers/build tools and get consistent behavior if they add host unconditionally.

---

I really appreciate that you added that change in gh-2896!
I think the changes to discern empty from non-existent values are very helpful going forward -- though I feel a bit sorry it was my request that made you make those profound changes, i.e., caused extended work :-O . But again, really appreciated! 👍

[github-actions]

Hi there, thank you for your contribution!

This issue has been automatically locked because it has not had recent activity after being closed.

Please open a new issue if needed.

Thanks!