{"thread":{"id":"65917","subject":"[PATCH v6 0/2] includeIf: add \"worktree\" condition for matching working tree path","startedAt":"2026-07-03T03:13:47Z","lastAt":"2026-07-07T19:23:52Z","messageCount":8,"participants":["Chen Linxuan via B4 Relay","Junio C Hamano","Patrick Steinhardt","Chen Linxuan"],"isPatch":true,"patchVersion":6,"patchTotal":2},"messages":[{"id":"547022","messageId":"20260703-includeif-worktree-v6-0-a13893ad9a7f@black-desk.cn","threadId":"65917","inReplyTo":null,"subject":"[PATCH v6 0/2] includeIf: add \"worktree\" condition for matching working tree path","fromName":"Chen Linxuan via B4 Relay","fromEmail":"devnull+me.black-desk.cn@kernel.org","sentAt":"2026-07-03T03:13:16Z","receivedAt":"2026-07-03T03:13:47Z","isPatch":true,"body":"The `includeIf` mechanism already supports matching on the `.git`\ndirectory path (`gitdir`) and the currently checked out branch\n(`onbranch`).  But in multi-worktree setups the `.git` directory of a\nlinked worktree points into the main repository's `.git/worktrees/`\narea, which makes `gitdir` patterns cumbersome when one wants to\ninclude config based on the working tree's checkout path instead.\n\nIntroduce two new condition keywords:\n\n  - `worktree:<pattern>` matches the realpath of the current worktree's\n    working directory against a glob pattern.\n  - `worktree/i:<pattern>` is the case-insensitive variant.\n\nSupported pattern features: glob wildcards, `**/` and `/**`, `~`\nexpansion, `./` relative paths, and trailing-`/` prefix matching.\nThe condition never matches in a bare repository.\n\nSigned-off-by: Chen Linxuan <me@black-desk.cn>\n---\nChanges in v6:\n- Rebase onto current `master` at Git 2.55.\n- Add an in-code comment explaining why the non-repository worktree\n  tests use the loose `**.path` pattern (suggested by Junio C Hamano).\n- Link to v5: https://lore.kernel.org/r/20260525-includeif-worktree-v5-0-1efe525d025a@black-desk.cn\n\nChanges in v5:\n- Fix Windows CI failure: use `**` glob pattern instead of `/` in the\n  \"worktree without repository\" tests, since `/` as a path pattern is\n  Unix-specific and does not match Windows paths.\n  Github CI pass: https://github.com/black-desk/git/actions/runs/26380466288\n- Add a test verifying case-sensitive matching by default, with the\n  `!CASE_INSENSITIVE_FS` prerequisite (suggested by Patrick Steinhardt).\n- Link to v4: https://lore.kernel.org/r/20260513-includeif-worktree-v4-0-f8e6212d1fba@black-desk.cn\n\nChanges in v4:\n- Deduplicate the worktree pattern documentation by referencing the\n  gitdir syntax instead of repeating the full pattern description\n  (suggested by Patrick Steinhardt).\n- Add documentation comparing includeIf \"worktree:\" with\n  extensions.worktreeConfig, including a concrete use case example\n  (suggested by Phillip Wood, Junio C Hamano).\n- Add a test verifying that the worktree condition does not match\n  during early config reading (suggested by Patrick Steinhardt).\n- Add tests for the non-repository (nongit) scenario (suggested by\n  Patrick Steinhardt).\n- Add a test for the case-insensitive \"worktree/i\" variant\n- Link to v3: https://lore.kernel.org/r/20260403-includeif-worktree-v3-0-109ce5782b03@black-desk.cn\n\nChanges in v3:\n- Apply Junio's suggestion.\n- Link to v2: https://lore.kernel.org/r/20260402-includeif-worktree-v2-0-36e339b898d7@black-desk.cn\n\nChanges in v2:\n\n- Add missing signed-off-by lines.\n- Link to v1: https://lore.kernel.org/r/20260401-includeif-worktree-v1-0-906db69f2c79@black-desk.cn\n\n---\nChen Linxuan (2):\n      config: refactor include_by_gitdir() into include_by_path()\n      config: add \"worktree\" and \"worktree/i\" includeIf conditions\n\n Documentation/config.adoc |  48 +++++++++++++++++\n config.c                  |  25 +++++----\n t/t1305-config-include.sh | 128 ++++++++++++++++++++++++++++++++++++++++++++++\n 3 files changed, 190 insertions(+), 11 deletions(-)\n---\nbase-commit: e9019fcafe0040228b8631c30f97ae1adb61bcdc\nchange-id: 20260401-includeif-worktree-fcb64950dfba\n\nBest regards,\n-- \nChen Linxuan <me@black-desk.cn>\n\n\n"},{"id":"547023","messageId":"20260703-includeif-worktree-v6-1-a13893ad9a7f@black-desk.cn","threadId":"65917","inReplyTo":"20260703-includeif-worktree-v6-0-a13893ad9a7f@black-desk.cn","subject":"[PATCH v6 1/2] config: refactor include_by_gitdir() into include_by_path()","fromName":"Chen Linxuan via B4 Relay","fromEmail":"devnull+me.black-desk.cn@kernel.org","sentAt":"2026-07-03T03:13:17Z","receivedAt":"2026-07-03T03:13:47Z","isPatch":true,"body":"From: Chen Linxuan <me@black-desk.cn>\n\nThe include_by_gitdir() function matches the realpath of a given\npath against a glob pattern, but its interface is tightly coupled to\nthe gitdir condition: it takes a struct config_options *opts and\nextracts opts->git_dir internally.\n\nRefactor it into a more generic include_by_path() helper that takes\na const char *path parameter directly, and update the gitdir and\ngitdir/i callers to pass opts->git_dir explicitly.  No behavior\nchange, just preparing for the addition of a new worktree condition\nthat will reuse the same path-matching logic with a different path.\n\nSigned-off-by: Chen Linxuan <me@black-desk.cn>\n---\n config.c | 19 ++++++++-----------\n 1 file changed, 8 insertions(+), 11 deletions(-)\n\ndiff --git a/config.c b/config.c\nindex 6a0de86e3ae9..00eeeea370c9 100644\n--- a/config.c\n+++ b/config.c\n@@ -235,23 +235,20 @@ static int prepare_include_condition_pattern(const struct key_value_info *kvi,\n \treturn 0;\n }\n \n-static int include_by_gitdir(const struct key_value_info *kvi,\n-\t\t\t     const struct config_options *opts,\n-\t\t\t     const char *cond, size_t cond_len, int icase)\n+static int include_by_path(const struct key_value_info *kvi,\n+\t\t\t   const char *path,\n+\t\t\t   const char *cond, size_t cond_len, int icase)\n {\n \tstruct strbuf text = STRBUF_INIT;\n \tstruct strbuf pattern = STRBUF_INIT;\n \tsize_t prefix;\n \tint ret = 0;\n-\tconst char *git_dir;\n \tint already_tried_absolute = 0;\n \n-\tif (opts->git_dir)\n-\t\tgit_dir = opts->git_dir;\n-\telse\n+\tif (!path)\n \t\tgoto done;\n \n-\tstrbuf_realpath(&text, git_dir, 1);\n+\tstrbuf_realpath(&text, path, 1);\n \tstrbuf_add(&pattern, cond, cond_len);\n \tret = prepare_include_condition_pattern(kvi, &pattern, &prefix);\n \tif (ret < 0)\n@@ -284,7 +281,7 @@ static int include_by_gitdir(const struct key_value_info *kvi,\n \t\t * which'll do the right thing\n \t\t */\n \t\tstrbuf_reset(&text);\n-\t\tstrbuf_add_absolute_path(&text, git_dir);\n+\t\tstrbuf_add_absolute_path(&text, path);\n \t\talready_tried_absolute = 1;\n \t\tgoto again;\n \t}\n@@ -400,9 +397,9 @@ static int include_condition_is_true(const struct key_value_info *kvi,\n \tconst struct config_options *opts = inc->opts;\n \n \tif (skip_prefix_mem(cond, cond_len, \"gitdir:\", &cond, &cond_len))\n-\t\treturn include_by_gitdir(kvi, opts, cond, cond_len, 0);\n+\t\treturn include_by_path(kvi, opts->git_dir, cond, cond_len, 0);\n \telse if (skip_prefix_mem(cond, cond_len, \"gitdir/i:\", &cond, &cond_len))\n-\t\treturn include_by_gitdir(kvi, opts, cond, cond_len, 1);\n+\t\treturn include_by_path(kvi, opts->git_dir, cond, cond_len, 1);\n \telse if (skip_prefix_mem(cond, cond_len, \"onbranch:\", &cond, &cond_len))\n \t\treturn include_by_branch(inc, cond, cond_len);\n \telse if (skip_prefix_mem(cond, cond_len, \"hasconfig:remote.*.url:\", &cond,\n\n-- \n2.53.0\n\n\n"},{"id":"547024","messageId":"20260703-includeif-worktree-v6-2-a13893ad9a7f@black-desk.cn","threadId":"65917","inReplyTo":"20260703-includeif-worktree-v6-0-a13893ad9a7f@black-desk.cn","subject":"[PATCH v6 2/2] config: add \"worktree\" and \"worktree/i\" includeIf conditions","fromName":"Chen Linxuan via B4 Relay","fromEmail":"devnull+me.black-desk.cn@kernel.org","sentAt":"2026-07-03T03:13:18Z","receivedAt":"2026-07-03T03:13:47Z","isPatch":true,"body":"From: Chen Linxuan <me@black-desk.cn>\n\nThe includeIf mechanism already supports matching on the .git\ndirectory path (gitdir) and the currently checked out branch\n(onbranch).  But in multi-worktree setups the .git directory of a\nlinked worktree points into the main repository's .git/worktrees/\narea, which makes gitdir patterns cumbersome when one wants to\ninclude config based on the working tree's checkout path instead.\n\nIntroduce two new condition keywords:\n\n  - worktree:<pattern> matches the realpath of the current worktree's\n    working directory (i.e. repo_get_work_tree()) against a glob\n    pattern.  This is the path returned by git rev-parse\n    --show-toplevel.\n\n  - worktree/i:<pattern> is the case-insensitive variant.\n\nThe implementation reuses the include_by_path() helper introduced in\nthe previous commit, passing the worktree path in place of the\ngitdir.  The condition never matches in bare repositories (where\nthere is no worktree) or during early config reading (where no\nrepository is available).\n\nAdd documentation describing the new conditions, including a comparison\nwith extensions.worktreeConfig.  Add tests covering bare repositories,\nmultiple worktrees, symlinked worktree paths, case-sensitive and\ncase-insensitive matching, early config reading, and non-repository\nscenarios.\n\nSigned-off-by: Chen Linxuan <me@black-desk.cn>\n---\n Documentation/config.adoc |  48 +++++++++++++++++\n config.c                  |   6 +++\n t/t1305-config-include.sh | 128 ++++++++++++++++++++++++++++++++++++++++++++++\n 3 files changed, 182 insertions(+)\n\ndiff --git a/Documentation/config.adoc b/Documentation/config.adoc\nindex 15b1a4d59347..c153da986e4a 100644\n--- a/Documentation/config.adoc\n+++ b/Documentation/config.adoc\n@@ -146,6 +146,46 @@ refer to linkgit:gitignore[5] for details. For convenience:\n \tThis is the same as `gitdir` except that matching is done\n \tcase-insensitively (e.g. on case-insensitive file systems)\n \n+`worktree`::\n+\tThe data that follows the keyword `worktree` and a colon is used as a\n+\tglob pattern. If the working directory of the current worktree matches\n+\tthe pattern, the include condition is met.\n++\n+The worktree location is the path where files are checked out (as returned\n+by `git rev-parse --show-toplevel`). This is different from `gitdir`, which\n+matches the `.git` directory path. In a linked worktree, the worktree path\n+is the directory where that worktree's files are located, not the main\n+repository's `.git` directory.\n++\n+The pattern uses the same glob syntax as `gitdir` (including `~/`, `./`,\n+`**/`, and trailing-`/` prefix matching). This condition will never match\n+in a bare repository (which has no worktree).\n++\n+This is useful when you want to apply configuration based on where the\n+working tree is located on the filesystem. For example, a contributor who\n+works on the same project both personally and as an employee can use\n+different `user.name` and `user.email` values depending on which directory\n+the worktree is checked out under:\n++\n+----\n+[includeIf \"worktree:/home/user/work/\"]\n+    path = ~/.config/git/work.inc\n+[includeIf \"worktree:/home/user/personal/\"]\n+    path = ~/.config/git/personal.inc\n+----\n++\n+While `extensions.worktreeConfig` (see linkgit:git-worktree[1]) also supports\n+per-worktree configuration, it stores the config inside each repository's\n+`.git/config.worktree` file and requires running `git config --worktree`\n+inside each worktree individually. In contrast, `includeIf \"worktree:...\"`\n+can be set once in a global or system-level configuration file (e.g.\n+`~/.config/git/config`) and applies to all repositories at once based on\n+their worktree location.\n+\n+`worktree/i`::\n+\tThis is the same as `worktree` except that matching is done\n+\tcase-insensitively (e.g. on case-insensitive file systems)\n+\n `onbranch`::\n \tThe data that follows the keyword `onbranch` and a colon is taken to be a\n \tpattern with standard globbing wildcards and two additional\n@@ -244,6 +284,14 @@ Example\n [includeIf \"gitdir:~/to/group/\"]\n \tpath = /path/to/foo.inc\n \n+; include if the worktree is at /path/to/project-build\n+[includeIf \"worktree:/path/to/project-build\"]\n+\tpath = build-config.inc\n+\n+; include for all worktrees inside /path/to/group\n+[includeIf \"worktree:/path/to/group/\"]\n+\tpath = group-config.inc\n+\n ; relative paths are always relative to the including\n ; file (if the condition is true); their location is not\n ; affected by the condition\ndiff --git a/config.c b/config.c\nindex 00eeeea370c9..9d6d7872d76c 100644\n--- a/config.c\n+++ b/config.c\n@@ -400,6 +400,12 @@ static int include_condition_is_true(const struct key_value_info *kvi,\n \t\treturn include_by_path(kvi, opts->git_dir, cond, cond_len, 0);\n \telse if (skip_prefix_mem(cond, cond_len, \"gitdir/i:\", &cond, &cond_len))\n \t\treturn include_by_path(kvi, opts->git_dir, cond, cond_len, 1);\n+\telse if (skip_prefix_mem(cond, cond_len, \"worktree:\", &cond, &cond_len))\n+\t\treturn include_by_path(kvi, inc->repo ? repo_get_work_tree(inc->repo) : NULL,\n+\t\t\t\t       cond, cond_len, 0);\n+\telse if (skip_prefix_mem(cond, cond_len, \"worktree/i:\", &cond, &cond_len))\n+\t\treturn include_by_path(kvi, inc->repo ? repo_get_work_tree(inc->repo) : NULL,\n+\t\t\t\t       cond, cond_len, 1);\n \telse if (skip_prefix_mem(cond, cond_len, \"onbranch:\", &cond, &cond_len))\n \t\treturn include_by_branch(inc, cond, cond_len);\n \telse if (skip_prefix_mem(cond, cond_len, \"hasconfig:remote.*.url:\", &cond,\ndiff --git a/t/t1305-config-include.sh b/t/t1305-config-include.sh\nindex f3892578e4ff..4e840dfdb35b 100755\n--- a/t/t1305-config-include.sh\n+++ b/t/t1305-config-include.sh\n@@ -396,4 +396,132 @@ test_expect_success 'onbranch without repository but explicit nonexistent Git di\n \ttest_must_fail nongit git --git-dir=nonexistent config get foo.bar\n '\n \n+# worktree: conditional include tests\n+\n+test_expect_success 'conditional include, worktree bare repo' '\n+\tgit init --bare wt-bare &&\n+\t(\n+\t\tcd wt-bare &&\n+\t\techo \"[includeIf \\\"worktree:/\\\"]path=bar-bare\" >>config &&\n+\t\techo \"[test]wtbare=1\" >bar-bare &&\n+\t\ttest_must_fail git config test.wtbare\n+\t)\n+'\n+\n+test_expect_success 'conditional include, worktree multiple worktrees' '\n+\tgit init wt-multi &&\n+\t(\n+\t\tcd wt-multi &&\n+\t\ttest_commit initial &&\n+\t\tgit worktree add -b linked-branch ../wt-linked HEAD &&\n+\t\tgit worktree add -b prefix-branch ../wt-prefix/linked HEAD\n+\t) &&\n+\twt_main=\"$(cd wt-multi && pwd)\" &&\n+\twt_linked=\"$(cd wt-linked && pwd)\" &&\n+\twt_prefix_parent=\"$(cd wt-prefix && pwd)\" &&\n+\tcat >>wt-multi/.git/config <<-EOF &&\n+\t[includeIf \"worktree:$wt_main\"]\n+\t\tpath = main-config\n+\t[includeIf \"worktree:$wt_linked\"]\n+\t\tpath = linked-config\n+\t[includeIf \"worktree:$wt_prefix_parent/\"]\n+\t\tpath = prefix-config\n+\tEOF\n+\techo \"[test]mainvar=main\" >wt-multi/.git/main-config &&\n+\techo \"[test]linkedvar=linked\" >wt-multi/.git/linked-config &&\n+\techo \"[test]prefixvar=prefix\" >wt-multi/.git/prefix-config &&\n+\techo main >expect &&\n+\tgit -C wt-multi config test.mainvar >actual &&\n+\ttest_cmp expect actual &&\n+\ttest_must_fail git -C wt-multi config test.linkedvar &&\n+\ttest_must_fail git -C wt-multi config test.prefixvar &&\n+\techo linked >expect &&\n+\tgit -C wt-linked config test.linkedvar >actual &&\n+\ttest_cmp expect actual &&\n+\ttest_must_fail git -C wt-linked config test.mainvar &&\n+\ttest_must_fail git -C wt-linked config test.prefixvar &&\n+\techo prefix >expect &&\n+\tgit -C wt-prefix/linked config test.prefixvar >actual &&\n+\ttest_cmp expect actual &&\n+\ttest_must_fail git -C wt-prefix/linked config test.mainvar &&\n+\ttest_must_fail git -C wt-prefix/linked config test.linkedvar\n+'\n+\n+test_expect_success SYMLINKS 'conditional include, worktree resolves symlinks' '\n+\tmkdir real-wt &&\n+\tln -s real-wt link-wt &&\n+\tgit init link-wt/repo &&\n+\t(\n+\t\tcd link-wt/repo &&\n+\t\t# repo->worktree resolves symlinks, so use real path in pattern\n+\t\techo \"[includeIf \\\"worktree:**/real-wt/repo\\\"]path=bar-link\" >>.git/config &&\n+\t\techo \"[test]wtlink=2\" >.git/bar-link &&\n+\t\techo 2 >expect &&\n+\t\tgit config test.wtlink >actual &&\n+\t\ttest_cmp expect actual\n+\t)\n+'\n+\n+test_expect_success !CASE_INSENSITIVE_FS 'conditional include, worktree, case sensitive' '\n+\tgit init wt-case &&\n+\t(\n+\t\tcd wt-case &&\n+\t\ttest_commit initial &&\n+\t\twt_path=\"$(pwd)\" &&\n+\t\twt_upper=$(echo \"$wt_path\" | tr a-z A-Z) &&\n+\t\techo \"[includeIf \\\"worktree:$wt_upper\\\"]path=case-inc\" >>.git/config &&\n+\t\techo \"[test]wtcase=1\" >.git/case-inc &&\n+\t\ttest_must_fail git config test.wtcase\n+\t)\n+'\n+\n+test_expect_success 'conditional include, worktree, icase' '\n+\tgit init wt-icase &&\n+\t(\n+\t\tcd wt-icase &&\n+\t\ttest_commit initial &&\n+\t\twt_path=\"$(pwd)\" &&\n+\t\twt_upper=$(echo \"$wt_path\" | tr a-z A-Z) &&\n+\t\techo \"[includeIf \\\"worktree/i:$wt_upper\\\"]path=icase-inc\" >>.git/config &&\n+\t\techo \"[test]wticase=1\" >.git/icase-inc &&\n+\t\techo 1 >expect &&\n+\t\tgit config test.wticase >actual &&\n+\t\ttest_cmp expect actual\n+\t)\n+'\n+\n+# The \"worktree\" condition cannot match during early config reading\n+# because the repository object is not yet fully initialized and\n+# repo_get_work_tree() returns NULL.\n+test_expect_success 'conditional include, worktree does not match in early config' '\n+\tgit init wt-early &&\n+\t(\n+\t\tcd wt-early &&\n+\t\ttest_commit initial &&\n+\t\twt_path=\"$(pwd)\" &&\n+\t\techo \"[includeIf \\\"worktree:$wt_path\\\"]path=early-inc\" >>.git/config &&\n+\t\techo \"[test]wtearly=1\" >.git/early-inc &&\n+\t\ttest-tool config read_early_config test.wtearly >actual &&\n+\t\ttest_must_be_empty actual\n+\t)\n+'\n+\n+# Use a loose pattern so the \"present in non-worktree cases\" check works\n+# for Unix-style absolute paths and Windows paths like D:/a/git/...\n+test_expect_success 'conditional include, worktree without repository' '\n+\ttest_when_finished \"rm -f .gitconfig config.inc\" &&\n+\tgit config set -f .gitconfig \"includeIf.worktree:**.path\" config.inc &&\n+\tgit config set -f config.inc foo.bar baz &&\n+\tgit config get foo.bar &&\n+\ttest_must_fail nongit git config get foo.bar\n+'\n+\n+test_expect_success 'conditional include, worktree without repository but explicit nonexistent Git directory' '\n+\ttest_when_finished \"rm -f .gitconfig config.inc\" &&\n+\tgit config set -f .gitconfig \"includeIf.worktree:**.path\" config.inc &&\n+\tgit config set -f config.inc foo.bar baz &&\n+\tgit config get foo.bar &&\n+\ttest_must_fail nongit git --git-dir=nonexistent config get foo.bar\n+'\n+\n test_done\n\n-- \n2.53.0\n\n\n"},{"id":"547050","messageId":"xmqq1pdkjvi5.fsf@gitster.g","threadId":"65917","inReplyTo":"20260703-includeif-worktree-v6-2-a13893ad9a7f@black-desk.cn","subject":"Re: [PATCH v6 2/2] config: add \"worktree\" and \"worktree/i\" includeIf conditions","fromName":"Junio C Hamano","fromEmail":"gitster@pobox.com","sentAt":"2026-07-03T09:02:58Z","receivedAt":"2026-07-03T09:03:02Z","isPatch":true,"body":"Chen Linxuan via B4 Relay <devnull+me.black-desk.cn@kernel.org>\nwrites:\n\n> +`worktree`::\n> +\tThe data that follows the keyword `worktree` and a colon is used as a\n> +\tglob pattern. If the working directory of the current worktree matches\n> +\tthe pattern, the include condition is met.\n> ...\n> +can be set once in a global or system-level configuration file (e.g.\n> +`~/.config/git/config`) and applies to all repositories at once based on\n> +their worktree location.\n> +\n> +`worktree/i`::\n> +\tThis is the same as `worktree` except that matching is done\n> +\tcase-insensitively (e.g. on case-insensitive file systems)\n> +\n\nOK.  I briefly wondered if\n\n        `worktree`::\n        `worktree/i`::\n                What follows the keyword `worktree` (or `worktree/i`) and a\n                colon is used as a glob pattern. If the working directory of\n                the current worktree matches (with `/i` the match is made\n                case-insensitively) the pattern, ...\n\nis easier for those who looks up 'worktree' to notice (without\nhaving to scroll too far to look at the other entry) that there is a\ncase insensitive option available.  As the construct used in this\npatch mimicks how `gitdir` and `gitdir/i` are described, however, I\nthink such a change is better done as a separate topic, long after\nthis patch lands and graduates to the master, to clean up both\n`gitdir` and `worktree` in the same commit.  So let's leave it out\nof this topic.\n\n>  \telse if (skip_prefix_mem(cond, cond_len, \"gitdir/i:\", &cond, &cond_len))\n>  \t\treturn include_by_path(kvi, opts->git_dir, cond, cond_len, 1);\n> +\telse if (skip_prefix_mem(cond, cond_len, \"worktree:\", &cond, &cond_len))\n> +\t\treturn include_by_path(kvi, inc->repo ? repo_get_work_tree(inc->repo) : NULL,\n> +\t\t\t\t       cond, cond_len, 0);\n> +\telse if (skip_prefix_mem(cond, cond_len, \"worktree/i:\", &cond, &cond_len))\n> +\t\treturn include_by_path(kvi, inc->repo ? repo_get_work_tree(inc->repo) : NULL,\n> +\t\t\t\t       cond, cond_len, 1);\n\nFairly straight-forward.\n\n> +# Use a loose pattern so the \"present in non-worktree cases\" check works\n> +# for Unix-style absolute paths and Windows paths like D:/a/git/...\n> +test_expect_success 'conditional include, worktree without repository' '\n> +\ttest_when_finished \"rm -f .gitconfig config.inc\" &&\n> +\tgit config set -f .gitconfig \"includeIf.worktree:**.path\" config.inc &&\n> +\tgit config set -f config.inc foo.bar baz &&\n> +\tgit config get foo.bar &&\n> +\ttest_must_fail nongit git config get foo.bar\n> +'\n\nThis looks much easier to understand than the previous round.\n\nThanks.\n"},{"id":"547068","messageId":"akeW4yFC8uuu2o8a@pks.im","threadId":"65917","inReplyTo":"20260703-includeif-worktree-v6-2-a13893ad9a7f@black-desk.cn","subject":"Re: [PATCH v6 2/2] config: add \"worktree\" and \"worktree/i\" includeIf conditions","fromName":"Patrick Steinhardt","fromEmail":"ps@pks.im","sentAt":"2026-07-03T11:02:59Z","receivedAt":"2026-07-03T11:03:06Z","isPatch":true,"body":"On Fri, Jul 03, 2026 at 11:13:18AM +0800, Chen Linxuan via B4 Relay wrote:\n> diff --git a/t/t1305-config-include.sh b/t/t1305-config-include.sh\n> index f3892578e4ff..4e840dfdb35b 100755\n> --- a/t/t1305-config-include.sh\n> +++ b/t/t1305-config-include.sh\n> @@ -396,4 +396,132 @@ test_expect_success 'onbranch without repository but explicit nonexistent Git di\n[snip]\n> +test_expect_success SYMLINKS 'conditional include, worktree resolves symlinks' '\n> +\tmkdir real-wt &&\n> +\tln -s real-wt link-wt &&\n> +\tgit init link-wt/repo &&\n> +\t(\n> +\t\tcd link-wt/repo &&\n> +\t\t# repo->worktree resolves symlinks, so use real path in pattern\n> +\t\techo \"[includeIf \\\"worktree:**/real-wt/repo\\\"]path=bar-link\" >>.git/config &&\n> +\t\techo \"[test]wtlink=2\" >.git/bar-link &&\n> +\t\techo 2 >expect &&\n> +\t\tgit config test.wtlink >actual &&\n> +\t\ttest_cmp expect actual\n> +\t)\n> +'\n\nOkay, this covers one scenario. But with \"gitdir:\" we're actually able\nto use both the symlinked and the real location:\n\n    test_expect_success SYMLINKS 'conditional include, worktree matching symlink' '\n    \tmkdir sym-real &&\n    \tln -s sym-real sym-link &&\n    \tgit init sym-link/repo &&\n    \t(\n    \t\tcd sym-link/repo &&\n    \t\tlink_path=\"$(pwd)\" &&\n    \t\treal_path=\"$(test-tool path-utils real_path \"$link_path\")\" &&\n    \t\tcat >>.git/config <<-EOF &&\n    \t\t[includeIf \"gitdir:$link_path/.git\"]\n    \t\t\tpath = gitdir-link\n    \t\t[includeIf \"gitdir:$real_path/.git\"]\n    \t\t\tpath = gitdir-real\n    \t\t[includeIf \"worktree:$link_path\"]\n    \t\t\tpath = worktree-link\n    \t\t[includeIf \"worktree:$real_path\"]\n    \t\t\tpath = worktree-real\n    \t\tEOF\n    \t\techo \"[test]gitdirlink=1\" >.git/gitdir-link &&\n    \t\techo \"[test]gitdirreal=1\" >.git/gitdir-real &&\n    \t\techo \"[test]worktreelink=1\" >.git/worktree-link &&\n    \t\techo \"[test]worktreereal=1\" >.git/worktree-real &&\n\n    \t\tgit config get test.gitdirlink &&\n    \t\tgit config get test.gitdirreal &&\n    \t\tgit config get test.worktreereal &&\n    \t\ttest_must_fail git config test.worktreelink\n    \t)\n    '\n\nThe last call to git-config(1) fails, which is inconsistent with how\nresolve the path for \"gitdir\".\n\nOther than that I didn't have anything to add, thanks!\n\nPatrick\n"},{"id":"547216","messageId":"CAC1kPDNBecLbmZwjfR5-CsNheF3rcbZ5=SQ+cwjzpFMjFr9KGQ@mail.gmail.com","threadId":"65917","inReplyTo":"akeW4yFC8uuu2o8a@pks.im","subject":"Re: [PATCH v6 2/2] config: add \"worktree\" and \"worktree/i\" includeIf conditions","fromName":"Chen Linxuan","fromEmail":"me@black-desk.cn","sentAt":"2026-07-06T12:18:39Z","receivedAt":"2026-07-06T12:18:54Z","isPatch":true,"body":"On Fri, Jul 3, 2026 at 7:03 PM Patrick Steinhardt <ps@pks.im> wrote:\n>\n> On Fri, Jul 03, 2026 at 11:13:18AM +0800, Chen Linxuan via B4 Relay wrote:\n> > diff --git a/t/t1305-config-include.sh b/t/t1305-config-include.sh\n> > index f3892578e4ff..4e840dfdb35b 100755\n> > --- a/t/t1305-config-include.sh\n> > +++ b/t/t1305-config-include.sh\n> > @@ -396,4 +396,132 @@ test_expect_success 'onbranch without repository but explicit nonexistent Git di\n> [snip]\n> > +test_expect_success SYMLINKS 'conditional include, worktree resolves symlinks' '\n> > +     mkdir real-wt &&\n> > +     ln -s real-wt link-wt &&\n> > +     git init link-wt/repo &&\n> > +     (\n> > +             cd link-wt/repo &&\n> > +             # repo->worktree resolves symlinks, so use real path in pattern\n> > +             echo \"[includeIf \\\"worktree:**/real-wt/repo\\\"]path=bar-link\" >>.git/config &&\n> > +             echo \"[test]wtlink=2\" >.git/bar-link &&\n> > +             echo 2 >expect &&\n> > +             git config test.wtlink >actual &&\n> > +             test_cmp expect actual\n> > +     )\n> > +'\n>\n> Okay, this covers one scenario. But with \"gitdir:\" we're actually able\n> to use both the symlinked and the real location:\n>\n>     test_expect_success SYMLINKS 'conditional include, worktree matching symlink' '\n>         mkdir sym-real &&\n>         ln -s sym-real sym-link &&\n>         git init sym-link/repo &&\n>         (\n>                 cd sym-link/repo &&\n>                 link_path=\"$(pwd)\" &&\n>                 real_path=\"$(test-tool path-utils real_path \"$link_path\")\" &&\n>                 cat >>.git/config <<-EOF &&\n>                 [includeIf \"gitdir:$link_path/.git\"]\n>                         path = gitdir-link\n>                 [includeIf \"gitdir:$real_path/.git\"]\n>                         path = gitdir-real\n>                 [includeIf \"worktree:$link_path\"]\n>                         path = worktree-link\n>                 [includeIf \"worktree:$real_path\"]\n>                         path = worktree-real\n>                 EOF\n>                 echo \"[test]gitdirlink=1\" >.git/gitdir-link &&\n>                 echo \"[test]gitdirreal=1\" >.git/gitdir-real &&\n>                 echo \"[test]worktreelink=1\" >.git/worktree-link &&\n>                 echo \"[test]worktreereal=1\" >.git/worktree-real &&\n>\n>                 git config get test.gitdirlink &&\n>                 git config get test.gitdirreal &&\n>                 git config get test.worktreereal &&\n>                 test_must_fail git config test.worktreelink\n>         )\n>     '\n>\n> The last call to git-config(1) fails, which is inconsistent with how\n> resolve the path for \"gitdir\".\n>\n\nI investigated the symlink mismatch.\n\n`gitdir:` works because `opts->git_dir` still preserves the discovered or\nuser-provided spelling, and `include_by_path()` matches both its realpath\nand its absolute non-realpath form.\n\n`worktree:` is different: `repo_get_work_tree()` returns\n`repo->worktree`, which is stored by `repo_set_worktree()` via\n`real_pathdup(path, 1)`. So the symlink spelling is already lost before\nwe evaluate includeIf conditions.\n\nChanging `repo->worktree` itself to preserve the original spelling looks\nrisky, because several users access `repo->worktree` directly, and setup\ncode appears to rely on it being canonical.\n\nMy current possible v7 approach is to keep `repo->worktree` canonical,\nbut store an additional absolute, normalized, non-realpath worktree path\nfor `includeIf.worktree`. For the ordinary discovered-repository case,\nthis has to be derived in `setup_discovered_git_dir()` from physical\n`cwd`, the worktree-root offset, and a validated `$PWD`, because\n`set_git_work_tree()` is otherwise only called with `\".\"`.\n\nThis makes your suggested test pass, but the plumbing is less trivial\nthan the original patch. Does this approach sound reasonable, or would\nyou prefer different semantics for symlinked worktree paths?\n\nChen Linxuan\n\n> Other than that I didn't have anything to add, thanks!\n>\n> Patrick\n>\n"},{"id":"547340","messageId":"ak0am-pEdtOvyBp4@pks.im","threadId":"65917","inReplyTo":"CAC1kPDNBecLbmZwjfR5-CsNheF3rcbZ5=SQ+cwjzpFMjFr9KGQ@mail.gmail.com","subject":"Re: [PATCH v6 2/2] config: add \"worktree\" and \"worktree/i\" includeIf conditions","fromName":"Patrick Steinhardt","fromEmail":"ps@pks.im","sentAt":"2026-07-07T15:26:19Z","receivedAt":"2026-07-07T15:26:26Z","isPatch":true,"body":"On Mon, Jul 06, 2026 at 08:18:39PM +0800, Chen Linxuan wrote:\n> On Fri, Jul 3, 2026 at 7:03 PM Patrick Steinhardt <ps@pks.im> wrote:\n> >\n> > On Fri, Jul 03, 2026 at 11:13:18AM +0800, Chen Linxuan via B4 Relay wrote:\n> > > diff --git a/t/t1305-config-include.sh b/t/t1305-config-include.sh\n> > > index f3892578e4ff..4e840dfdb35b 100755\n> > > --- a/t/t1305-config-include.sh\n> > > +++ b/t/t1305-config-include.sh\n> > > @@ -396,4 +396,132 @@ test_expect_success 'onbranch without repository but explicit nonexistent Git di\n> > [snip]\n> > > +test_expect_success SYMLINKS 'conditional include, worktree resolves symlinks' '\n> > > +     mkdir real-wt &&\n> > > +     ln -s real-wt link-wt &&\n> > > +     git init link-wt/repo &&\n> > > +     (\n> > > +             cd link-wt/repo &&\n> > > +             # repo->worktree resolves symlinks, so use real path in pattern\n> > > +             echo \"[includeIf \\\"worktree:**/real-wt/repo\\\"]path=bar-link\" >>.git/config &&\n> > > +             echo \"[test]wtlink=2\" >.git/bar-link &&\n> > > +             echo 2 >expect &&\n> > > +             git config test.wtlink >actual &&\n> > > +             test_cmp expect actual\n> > > +     )\n> > > +'\n> >\n> > Okay, this covers one scenario. But with \"gitdir:\" we're actually able\n> > to use both the symlinked and the real location:\n> >\n> >     test_expect_success SYMLINKS 'conditional include, worktree matching symlink' '\n> >         mkdir sym-real &&\n> >         ln -s sym-real sym-link &&\n> >         git init sym-link/repo &&\n> >         (\n> >                 cd sym-link/repo &&\n> >                 link_path=\"$(pwd)\" &&\n> >                 real_path=\"$(test-tool path-utils real_path \"$link_path\")\" &&\n> >                 cat >>.git/config <<-EOF &&\n> >                 [includeIf \"gitdir:$link_path/.git\"]\n> >                         path = gitdir-link\n> >                 [includeIf \"gitdir:$real_path/.git\"]\n> >                         path = gitdir-real\n> >                 [includeIf \"worktree:$link_path\"]\n> >                         path = worktree-link\n> >                 [includeIf \"worktree:$real_path\"]\n> >                         path = worktree-real\n> >                 EOF\n> >                 echo \"[test]gitdirlink=1\" >.git/gitdir-link &&\n> >                 echo \"[test]gitdirreal=1\" >.git/gitdir-real &&\n> >                 echo \"[test]worktreelink=1\" >.git/worktree-link &&\n> >                 echo \"[test]worktreereal=1\" >.git/worktree-real &&\n> >\n> >                 git config get test.gitdirlink &&\n> >                 git config get test.gitdirreal &&\n> >                 git config get test.worktreereal &&\n> >                 test_must_fail git config test.worktreelink\n> >         )\n> >     '\n> >\n> > The last call to git-config(1) fails, which is inconsistent with how\n> > resolve the path for \"gitdir\".\n> >\n> \n> I investigated the symlink mismatch.\n> \n> `gitdir:` works because `opts->git_dir` still preserves the discovered or\n> user-provided spelling, and `include_by_path()` matches both its realpath\n> and its absolute non-realpath form.\n> \n> `worktree:` is different: `repo_get_work_tree()` returns\n> `repo->worktree`, which is stored by `repo_set_worktree()` via\n> `real_pathdup(path, 1)`. So the symlink spelling is already lost before\n> we evaluate includeIf conditions.\n> \n> Changing `repo->worktree` itself to preserve the original spelling looks\n> risky, because several users access `repo->worktree` directly, and setup\n> code appears to rely on it being canonical.\n> \n> My current possible v7 approach is to keep `repo->worktree` canonical,\n> but store an additional absolute, normalized, non-realpath worktree path\n> for `includeIf.worktree`. For the ordinary discovered-repository case,\n> this has to be derived in `setup_discovered_git_dir()` from physical\n> `cwd`, the worktree-root offset, and a validated `$PWD`, because\n> `set_git_work_tree()` is otherwise only called with `\".\"`.\n> \n> This makes your suggested test pass, but the plumbing is less trivial\n> than the original patch. Does this approach sound reasonable, or would\n> you prefer different semantics for symlinked worktree paths?\n\nIt certainly sounds a bit ugly, but I'd rather have something that's\nugly than something that's inconsistent for our users *shrug*\n\nThanks!\n\nPatrick\n"},{"id":"547389","messageId":"xmqqfr1u38oq.fsf@gitster.g","threadId":"65917","inReplyTo":"ak0am-pEdtOvyBp4@pks.im","subject":"Re: [PATCH v6 2/2] config: add \"worktree\" and \"worktree/i\" includeIf conditions","fromName":"Junio C Hamano","fromEmail":"gitster@pobox.com","sentAt":"2026-07-07T19:23:49Z","receivedAt":"2026-07-07T19:23:52Z","isPatch":true,"body":"Patrick Steinhardt <ps@pks.im> writes:\n\n>> > The last call to git-config(1) fails, which is inconsistent with how\n>> > resolve the path for \"gitdir\".\n>> \n>> I investigated the symlink mismatch.\n>> ...\n>> My current possible v7 approach is to keep `repo->worktree` canonical,\n>> but store an additional absolute, normalized, non-realpath worktree path\n>> for `includeIf.worktree`. For the ordinary discovered-repository case,\n>> this has to be derived in `setup_discovered_git_dir()` from physical\n>> `cwd`, the worktree-root offset, and a validated `$PWD`, because\n>> `set_git_work_tree()` is otherwise only called with `\".\"`.\n>> \n>> This makes your suggested test pass, but the plumbing is less trivial\n>> than the original patch. Does this approach sound reasonable, or would\n>> you prefer different semantics for symlinked worktree paths?\n>\n> It certainly sounds a bit ugly, but I'd rather have something that's\n> ugly than something that's inconsistent for our users *shrug*\n\nOK, so I'd expect v7 to come and then we hopefully can declare\nvictory ;-).  Thanks, both.\n"}]}