tests: three more GIT_CONFIG_COUNT appends worked out by hand escape #818's scan, and paste()'s git submodule update --init still clones a local path with no insteadOf rewrite #835

Goal: every test that gives git a setting through GIT_CONFIG_COUNT builds it with appended_to_command_scope, a hand-worked append is something a test refuses, and a pasted git submodule update --init reads a just-pushed remote the way every other clone of the suite does, or the issue says why it need not.

Filed by lane openRepoShape-5, from the checklist candidates of #758's PR (#818, writer w758: "Left out, and checklist candidates", both bullets). #758 itself is not reopened; these are what its scan and its paste_env left. Tests only. The first and third lines are read from source; none is measured, because no test fails today.

Lands after: lane 5's #758 (PR #818, fix/758-git-command-scope, head b6e15591), stacked on #662 (PR #748, fix/662-insteadof-helper, 8368d1f8). #818 adds the scan this issue's second line is about and the per-call paste_env; appended_to_command_scope is on main (tests/conftest.py:204, from #713). path:line is at main 6626a3c5 for the three sites, and at #818's head b6e15591 for what #818 adds.

Lines

• Three more appends are worked out by hand beside appended_to_command_scope. w758's candidate 2; each is in a file tests/test_doctor_legs_members_next.py and 24 other files: tests append what they tell git to the GIT_CONFIG_COUNT entries the environment holds, so a pasted line keeps the session's maintenance settings and the Windows job's two (#758) #818's base did not hold, which is why tests: PASTE_ENV in tests/test_doctor_legs_members_next.py is an import-time snapshot that overwrites GIT_CONFIG_COUNT, seven tests hand-build the same append beside appended_to_command_scope, and six git clone --no-local calls build clone_just_pushed's options by hand #758's second line (seven tests) and tests/test_doctor_legs_members_next.py and 24 other files: tests append what they tell git to the GIT_CONFIG_COUNT entries the environment holds, so a pasted line keeps the session's maintenance settings and the Windows job's two (#758) #818 (25 files) do not carry it. Each takes an environment and returns it with one more GIT_CONFIG_* entry, counted from the environment's own count, so each is right today:
  • on_disk (tests/test_family_head_lost_gitmodules.py:94-100): index = int(env.get("GIT_CONFIG_COUNT", "0")), then core.ignorecase;
  • with_config (tests/test_family_bootstrap_gate_order.py:132-137, from bootstrap.py and validate-family.py: a member no credential would check out fails the family bootstrap, every row's gitlink is asked before any checkout, and --no-members prints the case-variant move (#648) #682): the same arithmetic for an arbitrary key and value;
  • typed_with_local_remotes (tests/test_family_add_spellings_remedies.py:132-141, from family.py add: a members/ renamed with git mv is renamed back in two steps through a free name, after what the disk lacks under it is put back, and a printed --sparse says the git it needs, so no remedy fails as typed (#685) #767): count = int(os.environ.get("GIT_CONFIG_COUNT", "0")), then three monkeypatch.setenv calls for protocol.file.allow.
    Direction: one call to appended_to_command_scope(env, [(key, value)]) each, as tests/test_doctor_legs_members_next.py and 24 other files: tests append what they tell git to the GIT_CONFIG_COUNT entries the environment holds, so a pasted line keeps the session's maintenance settings and the Windows job's two (#758) #818 does for the seven. typed_with_local_remotes holds a monkeypatch.context(), so it sets each key the helper returns.
• tests/test_doctor_legs_members_next.py and 24 other files: tests append what they tell git to the GIT_CONFIG_COUNT entries the environment holds, so a pasted line keeps the session's maintenance settings and the Windows job's two (#758) #818's scan does not see an append worked out by hand, by its own design. counts_set_to_a_number (tests/test_no_auto_maintenance.py:372-415 at b6e15591) finds a GIT_CONFIG_COUNT set to a number a test writes out, and test_the_scan_leaves_a_count_that_is_worked_out (:434) pins that {"GIT_CONFIG_COUNT": str(held + 1)} and int(env.get("GIT_CONFIG_COUNT", "0")) are left alone, so the three sites above pass it, and so will the next test that works out its own count. Direction: decide whether it should flag a GIT_CONFIG_COUNT value built by arithmetic (a str(...) call or a +) outside tests/conftest.py. If so, OWN_COUNTS (:338) names the tests that build one on purpose (tests/test_no_auto_maintenance.py, tests/test_adopt_git_reads.py and tests/test_clone_just_pushed.py, which see that the code under test keeps the entries they hold), and the worked-out case moves from "left" to "found". If not, say why on the issue and leave the three sites as a sweep with no guard.
• A pasted git submodule update --init still clones a local path with no insteadOf rewrite (w758's candidate 1, which is rv662's "build it from read_through_git"; read from source). tests/test_doctor_legs_members_next.py and 24 other files: tests append what they tell git to the GIT_CONFIG_COUNT entries the environment holds, so a pasted line keeps the session's maintenance settings and the Windows job's two (#758) #818's paste_env (tests/test_doctor_legs_members_next.py:62-82) gives a pasted line the session's maintenance settings and protocol.file.allow, which is what tests: PASTE_ENV in tests/test_doctor_legs_members_next.py is an import-time snapshot that overwrites GIT_CONFIG_COUNT, seven tests hand-build the same append beside appended_to_command_scope, and six git clone --no-local calls build clone_just_pushed's options by hand #758 asks. paste(row, cwd) (:109-117) takes no remotes, so no url.<file URI>.insteadOf rewrite reaches UPDATE = "submodule update --init --recursive --checkout" (:88) or any other NEXT it pastes. That is the class tests/conftest.py: one helper builds the url.<base>.insteadOf rewrite that clone_just_pushed and read_through_git each build, and the plain-path git submodule update --init sites of #554's review use it (#440) #662's second line names for this file's own submodule update calls (:631, :987, :1054 at its filing, moved onto submodule_update_through_git by tests/conftest.py: one helper builds the url.<file URI>.insteadOf rewrite that clone_just_pushed, read_through_git and the adopt tests' pasted each built, and the empty-mount submodule update --init sites use it (#662) #748), and a row's pasted NEXT is not among them. tests/conftest.py: one helper builds the url.<base>.insteadOf rewrite that clone_just_pushed and read_through_git each build, and the plain-path git submodule update --init sites of #554's review use it (#440) #662 itself says that after tests: the holder fixtures copy the family fixture with shutil.copytree while git's background gc or maintenance can still be packing it, so a test errors at setup when a loose object vanishes mid-copy (#440's race, through a file copy) #645 automatic maintenance is off in every repository the suite makes, so the rewrite is a second guard here. Direction: decide first whether paste still wants it. If so, paste(row, cwd, remotes=None) appends insteadof_entries(remotes) (tests/conftest.py:107 at tests/conftest.py: one helper builds the url.<file URI>.insteadOf rewrite that clone_just_pushed, read_through_git and the adopt tests' pasted each built, and the empty-mount submodule update --init sites use it (#662) #748's head 8368d1f8) through appended_to_command_scope, and the callers that paste a submodule update hand it the session's remotes. Prove it on a pasted submodule update --init of an empty mount: GIT_TRACE shows upload-pack and pack-objects, and the clone's origin keeps the plain path, as tests/test_doctor_legs_members_next.py and 24 other files: tests append what they tell git to the GIT_CONFIG_COUNT entries the environment holds, so a pasted line keeps the session's maintenance settings and the Windows job's two (#758) #818's body measured for the helper's own clone.

Verify: tests/test_family_head_lost_gitmodules.py, tests/test_family_bootstrap_gate_order.py, tests/test_family_add_spellings_remedies.py, tests/test_no_auto_maintenance.py, tests/test_doctor_legs_members_next.py, tests/test_clone_just_pushed.py and tests/test_repo_hygiene.py pass on Linux. For the first line, the existing tests of each file hold the arithmetic's result; add one case that starts them under GIT_CONFIG_COUNT=2 (the Windows job's two entries) and asserts git config --show-origin --list through the changed helper lists both and the new key, as test_a_pasted_line_keeps_the_entries_the_environment_holds_at_the_call does for paste. For the second line, a mutant that puts one hand-worked append back, caught by the scan. For the third, the GIT_TRACE run above, if the line is taken.

Overlaps. #818 edits tests/test_doctor_legs_members_next.py and tests/test_no_auto_maintenance.py, so this lands after it; #748, #774, #775 and #820 edit tests/test_doctor_legs_members_next.py too, in other hunks. #820 and #822 edit tests/test_family_head_lost_gitmodules.py and tests/test_family_bootstrap_gate_order.py; #828 edits tests/test_family_add_spellings_remedies.py (new in #767, on main). tests/conftest.py is edited by #723, #746, #748, #806, #814 and #818. No pinned file changes.

Parent: #461.

Milestone: v1.1.1. Tests only, no test fails today, and each of the three sites appends correctly to the entries its caller holds.

🤖 Generated with Claude Code