Upgrade miniwdl to 1.15.0 and refresh WDL unit test skip lists (#5377)

requirements-wdl.txt

@@ -1,2 +1,2 @@
-miniwdl==1.13.1
+miniwdl==1.15.0
 wdlparse==0.1.0

src/toil/test/wdl/wdltoil_test.py

@@ -49,58 +49,28 @@
 # These tests (in the same order as in SPEC.md) are known to fail
 WDL_11_UNIT_TESTS_UNSUPPORTED_BY_TOIL = [
     "test_object",  # Objects are not supported
-    "map_to_struct",  # miniwdl cannot coerce map to struct, https://github.com/chanzuckerberg/miniwdl/issues/712
-    "relative_and_absolute_task",  # needs root to run
     "test_gpu_task",  # needs gpu to run, else warning
-    "hisat2_task",  # This needs way too many resources (and actually doesn't work?), see https://github.com/DataBiosphere/wdl-conformance-tests/blob/2d617b703a33791f75f30a9db43c3740a499cd89/README_UNIT.md?plain=1#L8
-    "gatk_haplotype_caller_task",  # same as above
     "input_ref_call",  # Inputs refering into workflow body not yet implemented: see https://github.com/DataBiosphere/toil/issues/4993
     "call_imported",  # Same as input_ref_call since it imports it
-    "call_imported_task",  # Same as input_ref_call since it imports it
-    "test_sub",  # MiniWDL does not handle metacharacters properly when running regex, https://github.com/chanzuckerberg/miniwdl/issues/709
-    "read_bool_task",  # miniwdl bug, see https://github.com/chanzuckerberg/miniwdl/issues/701
     "write_json_fail",  # miniwdl (and toil) bug, unserializable json is serialized, see https://github.com/chanzuckerberg/miniwdl/issues/702
     "read_object_task",  # object not supported
     "read_objects_task",  # object not supported
     "write_object_task",  # object not supported
     "write_objects_task",  # object not supported
-    "test_transpose",  # miniwdl bug, see https://github.com/chanzuckerberg/miniwdl/issues/699
-    "test_as_map_fail",  # miniwdl bug, evalerror, see https://github.com/chanzuckerberg/miniwdl/issues/700
-    "test_collect_by_key",  # same as test_as_map_
 ]
 
 WDL_12_UNIT_TESTS_UNSUPPORTED_BY_TOIL = WDL_11_UNIT_TESTS_UNSUPPORTED_BY_TOIL + [
-    "relative_paths_context",  # Toil can't yet resolve File coercion at task scope relative to task file.
+    "relative_paths_context",  # Workflow succeeds but output content does not match expected ("Expected and result do not match!")
     "file_directory_equality",  # String to Directory coercion not yet implemented.
-    "single_return_code_task",  # MiniWDL 1.13.1 only knows returnCodes and not return_codes.
-    "all_return_codes_task",  # MiniWDL 1.13.1 only knows returnCodes and not return_codes.
-    "test_runtime_info_task",  # MiniWDL 1.13.1 can't yet expose the task global.
-    "placeholder_none",  # 'outputs' section expected 1 results (['placeholder_none.s']), got 0 instead ([]) with exit code 1
-    "person_struct_task",  # Doesn't work as written in the spec; see https://github.com/openwdl/wdl/issues/739
-    "import_structs",  # Feature not yet implemented?
-    "environment_variable_should_echo",  # Ln 14 Col 45: Unexpected token STRING1_FRAGMENT
-    "outputs_task",  # 'outputs' section expected 2 results (['outputs.threshold', 'outputs.two_csvs']), got 3 instead (['outputs.two_csvs', 'outputs.csvs', 'outputs.threshold']) with exit code 0
-    "glob_task",  # 'outputs' section expected 1 results (['glob.last_file_contents']), got 2 instead (['glob.last_file_contents', 'glob.outfiles']) with exit code 0
-    "test_hints_task",  # Test is written as if the file has 3 lines, but it really has 2. See https://github.com/openwdl/wdl/issues/741
-    "input_hint_task",  # Missing outputs in test definition: https://github.com/openwdl/wdl/issues/740
-    "test_allow_nested_inputs",  # Ln 27 Col 3: Unexpected token HINTS
-    "multi_nested_inputs",  # Ln 8 Col 9: Unexpected token STRING1_FRAGMENT
-    "allow_nested",  # Ln 32 Col 9: Unexpected token STRING1_FRAGMENT
-    "test_find_task",  # Ln 9 Col 22: No such function: find
-    "test_matches_task",  # Ln 7 Col 29: No such function: matches
-    "change_extension_task",  # 'outputs' section expected 2 results (['change_extension.data', 'change_extension.index']), got 3 instead (['change_extension.index', 'change_extension.data', 'change_extension.data_file']) with exit code 0
-    "join_paths_task",  # Ln 14 Col 15: No such function: join_paths
-    "gen_files_task",  # 'outputs' section expected 1 results (['gen_files.glob_len']), got 2 instead (['gen_files.glob_len', 'gen_files.files']) with exit code 0
-    "file_sizes_task",  # WDL.Error.StaticTypeMismatch: Expected File? instead of Map[String,Pair[Int,File?]] 
+    "single_return_code_task",  # Task correctly exits 1 as declared via requirements.return_codes, but Toil/miniwdl still requires exit code 0 - return_codes override is not honored
+    "all_return_codes_task",  # Task correctly exits 42 as declared via requirements.return_codes: "*", but Toil/miniwdl still requires exit code 0 - return_codes override is not honored
+    "test_runtime_info_task",  # WDL.Error.EvalError from a KeyError; task runtime global still not fully exposed
+    "placeholder_none",  # WDL.Error.EvalError: select_first() given empty or all-null array; prevent this or append a default value
+    "environment_variable_should_echo",  # Now parses; Expected and result do not match!
+    "multi_nested_inputs",  # Workflow did not fail! Fails for toil-wdl and miniwdl
+    "join_paths_task",  # PermissionError: [Errno 13] Permission denied: '/usr/bin/sudo'
+    "file_sizes_task",  # WDL.Error.InputError: cannot coerce Map[String,Pair[Int,File?]] to Array[File?]
     "read_tsv_task",  # Ln 21 Col 5: Unknown type Object
-    "write_tsv_task",  # Ln 28 Col 16: write_tsv expects 1 argument(s)
-    "test_contains",  # Ln 25 Col 22: No such function: contains
-    "chunk_array",  # Ln 8 Col 17: No such function: chunk
-    "test_select_first",  # Ln 14 Col 17: select_first expects 1 argument(s)
-    "test_keys",  # Ln 32 Col 36: Expected Map[Any,Any] instead of Name
-    "test_contains_key",  # Ln 18 Col 20: No such function: contains_key
-    "test_values",  # Ln 28 Col 20: No such function: values
-    "test_length",  # length() isn't implemented for maps and strings yet
 ]
 
 

src/toil/wdl/wdltoil.py

@@ -858,6 +858,7 @@ def fill_execution_cache(
     output_bindings: WDLBindings,
     file_store: AbstractFileStore,
     wdl_options: WDLContext,
+    input_bindings: WDLBindings,
     miniwdl_logger: logging.Logger | None = None,
     miniwdl_config: WDL.runtime.config.Loader | None = None,
 ) -> WDLBindings:
@@ -946,7 +947,12 @@ def assign_shared_fs_path(inode: AnyINode) -> AnyINode:
     )
 
     # Save the bindings to the cache, representing all files with their shared filesystem paths.
-    miniwdl_cache.put(cache_key, view_shared_fs_paths(output_bindings))
+    miniwdl_cache.put(
+        cache_key,
+        view_shared_fs_paths(output_bindings),
+        inputs=view_shared_fs_paths(input_bindings),
+        add_paths=WDL.runtime.cache.CallCacheAddPaths(),
+    )
     logger.debug("Saved result to cache under %s", cache_key)
 
     # Keep using the transformed bindings so that later tasks use
@@ -2415,6 +2421,8 @@ def wrapper(v: WDL.Value.Base) -> WDL.Value.File:
                     WDL.Env.Bindings(
                         WDL.Env.Binding("file", WDL.Value.File(exported_path))
                     ),
+                    inputs=file_input_bindings,
+                    add_paths=WDL.runtime.cache.CallCacheAddPaths(),
                 )
 
                 # Apply the shared filesystem path to the virtualized file
@@ -4531,6 +4539,7 @@ def get_path_in_container(inode: AnyINode) -> AnyINode | None:
                 output_bindings,
                 file_store,
                 self._wdl_options,
+                unwrap(self._task_internal_bindings),
                 miniwdl_logger=miniwdl_logger,
                 miniwdl_config=miniwdl_config,
             )
@@ -5632,6 +5641,7 @@ def run(self, file_store: AbstractFileStore) -> Promised[WDLBindings]:
             self._enclosing_bindings,
             wdl_options=self._wdl_options,
             cache_key=cache_key,
+            input_bindings=bindings,
             local=True,
         )
         sink.addFollowOn(outputs_job)
@@ -5654,6 +5664,7 @@ def __init__(
         enclosing_bindings: WDLBindings,
         wdl_options: WDLContext,
         cache_key: str | None = None,
+        input_bindings: Promised[WDLBindings] | None = None,
         **kwargs: Any,
     ):
         """
@@ -5667,13 +5678,17 @@ def __init__(
         :param cache_key: If set and storing into the call cache is on, will
                cache the workflow execution result under the given key in a
                MiniWDL-compatible way.
+
+        :param input_bindings: The workflow call's input bindings, matching
+               what cache_key was derived from. Required if cache_key is set.
         """
         super().__init__(wdl_options=wdl_options, **kwargs)
 
         self._bindings = bindings
         self._enclosing_bindings = enclosing_bindings
         self._workflow = workflow
         self._cache_key = cache_key
+        self._input_bindings = input_bindings
 
     @report_wdl_errors("evaluate outputs")
     def run(self, file_store: AbstractFileStore) -> WDLBindings:
@@ -5757,8 +5772,13 @@ def run(self, file_store: AbstractFileStore) -> WDLBindings:
         output_bindings = virtualize_inodes(output_bindings, standard_library)
 
         if self._cache_key is not None:
+            assert self._input_bindings is not None
             output_bindings = fill_execution_cache(
-                self._cache_key, output_bindings, file_store, self._wdl_options
+                self._cache_key,
+                output_bindings,
+                file_store,
+                self._wdl_options,
+                unwrap(self._input_bindings),
             )
 
         # Let Files that are not output or available outside the call go out of