Merge pull request datalad#7431 from adswa/git-version

Increase minimal Git version to 2.25

changelog.d/pr-7431.md

@@ -0,0 +1,3 @@
+### 🚀 Enhancements and New Features
+
+- Increase minimal Git version to 2.25.  Fixes [#7389](https://github.com/datalad/datalad/issues/7389) via [PR #7431](https://github.com/datalad/datalad/pull/7431) (by [@adswa](https://github.com/adswa))

datalad/core/local/tests/test_diff.py

@@ -373,12 +373,9 @@ def test_path_diff(_path=None, linkpath=None):
     eq_(plain_recursive, ds.diff(path=['.', '.'], recursive=True, annex='all',
                                  result_renderer='disabled'))
     # neither do nested paths
-    if not "2.24.0" <= ds.repo.git_version < "2.25.0":
-        # Release 2.24.0 contained a regression that was fixed with 072a231016
-        # (2019-12-10).
-        eq_(plain_recursive,
-            ds.diff(path=['.', 'subds_modified'], recursive=True, annex='all',
-                    result_renderer='disabled'))
+    eq_(plain_recursive,
+        ds.diff(path=['.', 'subds_modified'], recursive=True, annex='all',
+                result_renderer='disabled'))
     # when invoked in a subdir of a dataset it still reports on the full thing
     # just like `git status`, as long as there are no paths specified
     with chpwd(op.join(path, 'directory_untracked')):

datalad/core/local/tests/test_save.py

@@ -800,12 +800,9 @@ def test_surprise_subds(path=None):
     # If subrepo is an adjusted branch, it would have a commit, making most of
     # this test irrelevant because it is about the unborn branch edge case.
     adjusted = somerepo.is_managed_branch()
-    # This edge case goes away with Git v2.22.0.
-    fixed_git = somerepo.git_version >= '2.22.0'
-
     # save non-recursive
     res = ds.save(recursive=False, on_failure='ignore')
-    if not adjusted and fixed_git:
+    if not adjusted:
         # We get an appropriate error about no commit being checked out.
         assert_in_results(res, action='add_submodule', status='error')
 
@@ -814,30 +811,18 @@ def test_surprise_subds(path=None):
     assert_repo_status(subds.path, untracked=['subfile'])
     assert_repo_status(somerepo.path, untracked=['subfile'])
 
-    if adjusted or fixed_git:
-        if adjusted:
-            # adjusted branch: #datalad/3178 (that would have a commit)
-            modified = [subds.repo.pathobj, somerepo.pathobj]
-            untracked = []
-        else:
-            # Newer Git versions refuse to add a sub-repository with no commits
-            # checked out.
-            modified = [subds.repo.pathobj]
-            untracked = ['d1']
+    if adjusted:
+        # adjusted branch: #datalad/3178 (that would have a commit)
+        modified = [subds.repo.pathobj, somerepo.pathobj]
+        untracked = []
+    else:
+        # Newer Git versions refuse to add a sub-repository with no commits
+        # checked out.
+        modified = [subds.repo.pathobj]
+        untracked = ['d1']
         assert_repo_status(ds.path, modified=modified, untracked=untracked)
         assert_not_in(ds.repo.pathobj / 'd1' / 'subrepo' / 'subfile',
                       ds.repo.get_content_info())
-    else:
-        # however, while the subdataset is added (and reported as modified
-        # because it content is still untracked) the subrepo
-        # cannot be added (it has no commit)
-        # worse: its untracked file add been added to the superdataset
-        assert_repo_status(ds.path, modified=['d2/subds'])
-        assert_in(ds.repo.pathobj / 'd1' / 'subrepo' / 'subfile',
-                  ds.repo.get_content_info())
-    # with proper subdatasets, all evil is gone
-    assert_not_in(ds.repo.pathobj / 'd2' / 'subds' / 'subfile',
-                  ds.repo.get_content_info())
 
 
 @with_tree({"foo": ""})

datalad/support/gitrepo.py

@@ -388,25 +388,8 @@ def remap_filenames(out):
 
     return  _wrap_normalize_paths
 
-
-if "2.24.0" <= external_versions["cmd:git"] < "2.25.0":
-    # An unintentional change in Git 2.24.0 led to `ls-files -o` traversing
-    # into untracked submodules when multiple pathspecs are given, returning
-    # repositories that are deeper than the first level. This helper filters
-    # these deeper levels out so that save_() doesn't fail trying to add them.
-    #
-    # This regression fixed with upstream's 072a231016 (2019-12-10).
-    def _prune_deeper_repos(repos: list[Path]) -> list[Path]:
-        firstlevel_repos = []
-        prev = None
-        for repo in sorted(repos):
-            if not (prev and str(repo).startswith(prev)):
-                prev = str(repo)
-                firstlevel_repos.append(repo)
-        return firstlevel_repos
-else:
-    def _prune_deeper_repos(repos: list[Path]) -> list[Path]:
-        return repos
+def _prune_deeper_repos(repos: list[Path]) -> list[Path]:
+    return repos
 
 
 class GitProgress(WitlessProtocol):
@@ -834,7 +817,7 @@ class GitRepo(CoreGitRepo):
     # should do it once
     _config_checked = False
 
-    GIT_MIN_VERSION = "2.19.1"
+    GIT_MIN_VERSION = "2.25"
     git_version = None
 
     @classmethod
@@ -2949,14 +2932,6 @@ def _get_content_info_line_helper(self, ref: Optional[str], info: dict[Path, dic
             inf: dict[str, str | int | None] = {}
             props = props_re.match(line)
             if not props:
-                # Kludge: Filter out paths starting with .git/ to work around
-                # an `ls-files -o` bug that was fixed in Git 2.25.
-                #
-                # TODO: Drop this condition when GIT_MIN_VERSION is at least
-                # 2.25.
-                if line.startswith(".git/"):
-                    lgr.debug("Filtering out .git/ file: %s", line)
-                    continue
                 # not known to Git, but Git always reports POSIX
                 path = ut.PurePosixPath(line)
                 inf['gitshasum'] = None