Propagate SETENV to forward model validation

The settings implied through SETENV can affect how forward
model behave, and may also change how validation should be
performed.

This is in particular true for the Eclipse forward model steps
which try to validate the requested version. If a different
configuration of Eclipse is set through

  SETENV ECL100_SITE_CONFIG somefile.yml

(or for ECL300_SITE_CONFIG) this information must be conveyed
to the validation code.

src/ert/config/ert_config.py

@@ -766,10 +766,15 @@ def _create_list_of_forward_model_steps_to_run(
         cls,
         installed_steps: Dict[str, ForwardModelStep],
         substitutions: Substitutions,
-        config_dict,
+        config_dict: dict,
     ) -> List[ForwardModelStep]:
         errors = []
         fm_steps = []
+
+        env_vars = {}
+        for key, val in config_dict.get("SETENV", []):
+            env_vars[key] = val
+
         for fm_step_description in config_dict.get(ConfigKeys.FORWARD_MODEL, []):
             if len(fm_step_description) > 1:
                 unsubstituted_step_name, args = fm_step_description
@@ -836,7 +841,7 @@ def _create_list_of_forward_model_steps_to_run(
                         skip_pre_experiment_validation=True,
                     )
                     job_json = substituted_json["jobList"][0]
-                    fm_step.validate_pre_experiment(job_json)
+                    fm_step.validate_pre_experiment(job_json, env_vars)
                 except ForwardModelStepValidationError as err:
                     errors.append(
                         ConfigValidationError.with_context(

src/ert/config/forward_model_step.py

@@ -181,7 +181,9 @@ def convert_to_substitutions(cls, v: Dict[str, str]) -> Substitutions:
             return v
         return Substitutions(v)
 
-    def validate_pre_experiment(self, fm_step_json: ForwardModelStepJSON) -> None:
+    def validate_pre_experiment(
+        self, fm_step_json: ForwardModelStepJSON, env_vars: Dict[str, str]
+    ) -> None:
         """
         Raise errors pertaining to the environment not being
         as the forward model step requires it to be. For example

src/ert/plugins/hook_implementations/forward_model_steps.py

@@ -3,7 +3,7 @@
 import subprocess
 from pathlib import Path
 from textwrap import dedent
-from typing import List, Literal, Optional, Type
+from typing import Dict, List, Literal, Optional, Type
 
 import yaml
 
@@ -14,7 +14,6 @@
     ForwardModelStepValidationError,
     plugin,
 )
-from ert.plugins import ErtPluginManager
 
 
 class CarefulCopyFile(ForwardModelStepPlugin):
@@ -220,13 +219,17 @@ def __init__(self) -> None:
             default_mapping={"<NUM_CPU>": 1, "<OPTS>": ""},
         )
 
-    def validate_pre_experiment(self, _: ForwardModelStepJSON) -> None:
+    def validate_pre_experiment(
+        self, _: ForwardModelStepJSON, env_vars: Dict[str, str]
+    ) -> None:
         if "<VERSION>" not in self.private_args:
             raise ForwardModelStepValidationError(
                 "Forward model step ECLIPSE100 must be given a VERSION argument"
             )
         version = self.private_args["<VERSION>"]
-        available_versions = _available_eclrun_versions(simulator="eclipse")
+        available_versions = _available_eclrun_versions(
+            simulator="eclipse", env_vars=env_vars
+        )
 
         if available_versions and version not in available_versions:
             raise ForwardModelStepValidationError(
@@ -278,13 +281,17 @@ def __init__(self) -> None:
             default_mapping={"<NUM_CPU>": 1, "<OPTS>": "", "<VERSION>": "version"},
         )
 
-    def validate_pre_experiment(self, _: ForwardModelStepJSON) -> None:
+    def validate_pre_experiment(
+        self, _: ForwardModelStepJSON, env_vars: Dict[str, str]
+    ) -> None:
         if "<VERSION>" not in self.private_args:
             raise ForwardModelStepValidationError(
                 "Forward model step ECLIPSE300 must be given a VERSION argument"
             )
         version = self.private_args["<VERSION>"]
-        available_versions = _available_eclrun_versions(simulator="e300")
+        available_versions = _available_eclrun_versions(
+            simulator="e300", env_vars=env_vars
+        )
         if available_versions and version not in available_versions:
             raise ForwardModelStepValidationError(
                 f"Unavailable ECLIPSE300 version {version} current supported "
@@ -628,14 +635,15 @@ def installable_forward_model_steps() -> List[Type[ForwardModelStepPlugin]]:
     return [*_UpperCaseFMSteps, *_LowerCaseFMSteps]
 
 
-def _available_eclrun_versions(simulator: Literal["eclipse", "e300"]) -> List[str]:
+def _available_eclrun_versions(
+    simulator: Literal["eclipse", "e300"], env_vars: Dict[str, str]
+) -> List[str]:
     if shutil.which("eclrun") is None:
         return []
-    pm = ErtPluginManager()
     ecl_config_path = (
-        pm.get_ecl100_config_path()
+        env_vars.get("ECL100_SITE_CONFIG")
         if simulator == "eclipse"
-        else pm.get_ecl300_config_path()
+        else env_vars.get("ECL300_SITE_CONFIG")
     )
 
     if not ecl_config_path:

test-data/ert/poly_example/poly_eval.py

@@ -1,5 +1,6 @@
 #!/usr/bin/env python3
 import json
+import time
 
 
 def _load_coeffs(filename):
@@ -16,3 +17,4 @@ def _evaluate(coeffs, x):
     output = [_evaluate(coeffs, x) for x in range(10)]
     with open("poly.out", "w", encoding="utf-8") as f:
         f.write("\n".join(map(str, output)))
+    time.sleep(10)

tests/ert/unit_tests/config/test_forward_model.py

@@ -5,6 +5,7 @@
 import stat
 from pathlib import Path
 from textwrap import dedent
+from typing import Dict
 from unittest.mock import patch
 
 import pytest
@@ -595,9 +596,12 @@ def test_that_eclipse_jobs_check_version(eclipse_v, mock_eclrun):
 
     ecl100_config_content = f"eclrun_env:\n  PATH: {os.getcwd()}\n"
     ecl300_config_content = f"eclrun_env:\n  PATH: {os.getcwd()}\n"
-    ert_config_contents = (
-        f"NUM_REALIZATIONS  1\nFORWARD_MODEL ECLIPSE{eclipse_v} (<VERSION>=1)\n"
-    )
+    ert_config_contents = f"""NUM_REALIZATIONS  1
+
+        SETENV ECL100_SITE_CONFIG {ecl100_config_file_name}
+        SETENV ECL300_SITE_CONFIG {ecl300_config_file_name}
+
+        FORWARD_MODEL ECLIPSE{eclipse_v} (<VERSION>=1)"""
 
     # Write config file
     config_file_name = "test.ert"
@@ -606,17 +610,11 @@ def test_that_eclipse_jobs_check_version(eclipse_v, mock_eclrun):
     Path(ecl100_config_file_name).write_text(ecl100_config_content, encoding="utf-8")
     # Write ecl300_config file
     Path(ecl300_config_file_name).write_text(ecl300_config_content, encoding="utf-8")
-    with patch(
-        "ert.plugins.hook_implementations.forward_model_steps.ErtPluginManager"
-    ) as mock:
-        instance = mock.return_value
-        instance.get_ecl100_config_path.return_value = ecl100_config_file_name
-        instance.get_ecl300_config_path.return_value = ecl300_config_file_name
-        with pytest.raises(
-            ConfigValidationError,
-            match=rf".*Unavailable ECLIPSE{eclipse_v} version 1 current supported versions \['4', '2', '8'\].*",
-        ):
-            _ = ErtConfig.with_plugins().from_file(config_file_name)
+    with pytest.raises(
+        ConfigValidationError,
+        match=rf".*Unavailable ECLIPSE{eclipse_v} version 1 current supported versions \['4', '2', '8'\].*",
+    ):
+        ErtConfig.with_plugins().from_file(config_file_name)
 
 
 @pytest.mark.skipif(shutil.which("eclrun") is not None, reason="eclrun is available")
@@ -669,7 +667,9 @@ def __init__(self):
                 command=["something", "<arg1>", "-f", "<arg2>", "<arg3>"],
             )
 
-        def validate_pre_experiment(self, fm_step_json: ForwardModelStepJSON) -> None:
+        def validate_pre_experiment(
+            self, fm_step_json: ForwardModelStepJSON, env_vars: Dict[str, str]
+        ) -> None:
             if set(self.private_args.keys()) != {"<arg1>", "<arg2>", "<arg3>"}:
                 raise ForwardModelStepValidationError("Bad")
 
@@ -893,6 +893,33 @@ def validate_pre_realization_run(
         )
 
 
+def test_that_plugin_forward_model_validation_sees_setenv(tmp_path):
+    (tmp_path / "test.ert").write_text(
+        """
+        NUM_REALIZATIONS  1
+        SETENV FOO bar
+        FORWARD_MODEL FM1()
+        """
+    )
+
+    class ExceptionThatWeWant(ForwardModelStepValidationError):
+        pass
+
+    class FM1(ForwardModelStepPlugin):
+        def __init__(self):
+            super().__init__(name="FM1", command=["dummy.sh"])
+
+        def validate_pre_experiment(
+            self, _: ForwardModelStepJSON, env_vars: Dict[str, str]
+        ) -> None:
+            raise ExceptionThatWeWant(f'Found FOO={env_vars["FOO"]}')
+
+    with pytest.raises(ConfigValidationError, match=".*Found FOO=bar.*"):
+        ErtConfig.with_plugins(forward_model_step_classes=[FM1]).from_file(
+            tmp_path / "test.ert"
+        )
+
+
 def test_that_plugin_forward_model_raises_pre_experiment_validation_error_early(
     tmp_path,
 ):
@@ -911,7 +938,9 @@ class FM1(ForwardModelStepPlugin):
         def __init__(self):
             super().__init__(name="FM1", command=["the_executable.sh"])
 
-        def validate_pre_experiment(self, fm_step_json: ForwardModelStepJSON) -> None:
+        def validate_pre_experiment(
+            self, fm_step_json: ForwardModelStepJSON, _: Dict[str, str]
+        ) -> None:
             if self.name != "FM1":
                 raise ForwardModelStepValidationError("Expected name to be FM1")
 
@@ -924,7 +953,9 @@ def __init__(self):
                 command=["the_executable.sh"],
             )
 
-        def validate_pre_experiment(self, fm_step_json: ForwardModelStepJSON) -> None:
+        def validate_pre_experiment(
+            self, fm_step_json: ForwardModelStepJSON, _: Dict[str, str]
+        ) -> None:
             if self.name != "FM2":
                 raise ForwardModelStepValidationError("Expected name to be FM2")
 
@@ -1011,7 +1042,9 @@ class FMWithAssertionError(ForwardModelStepPlugin):
         def __init__(self):
             super().__init__(name="FMWithAssertionError", command=["the_executable.sh"])
 
-        def validate_pre_experiment(self, fm_step_json: ForwardModelStepJSON) -> None:
+        def validate_pre_experiment(
+            self, fm_step_json: ForwardModelStepJSON, _: dict
+        ) -> None:
             raise AssertionError("I should be a warning")
 
     class FMWithFMStepValidationError(ForwardModelStepPlugin):
@@ -1021,7 +1054,9 @@ def __init__(self):
                 command=["the_executable.sh"],
             )
 
-        def validate_pre_experiment(self, fm_step_json: ForwardModelStepJSON) -> None:
+        def validate_pre_experiment(
+            self, fm_step_json: ForwardModelStepJSON, _: dict
+        ) -> None:
             raise ForwardModelStepValidationError("I should not be a warning")
 
     with (