status/commit: do not ignore staged submodules

This series fixes the problem that ignored but staged submodules do not show up in status and commit. Even though we do change default behavior here, I believe this is the Right Thing to do (and remember all interested parties in the discussion that raised this issue agreed on that [1]).

[1] http://article.gmane.org/gmane.comp.version-control.git/238173

Jens Lehmann (2):
status/commit: show staged submodules regardless of ignore config commit -m: commit staged submodules regardless of ignore config

 Documentation/config.txt     |  8 +++--
 Documentation/gitmodules.txt |  4 ++-
 builtin/commit.c             | 18 +++++++++--
 t/t7508-status.sh            | 74 ++++++++++++++++++++++++++++++++++++++++++--
 wt-status.c                  | 12 ++++++-
 5 files changed, 108 insertions(+), 8 deletions(-)

1.9.1.476.g510abc7

Currently setting submodule.<name>.ignore and/or diff.ignoreSubmodules to "all" suppresses all output of submodule changes for the diff family, status and commit. For status and commit this is really confusing, as it even when the user chooses to record a new commit for an ignored submodule by adding it manually this change won't show up under the to-be-committed changes. To add insult to injury, a later "git commit" will error out with "nothing to commit" when only ignored submodules are staged.

Fix that by making wt_status always print staged submodule changes, no matter what ignore settings are configured. The only exception is when the user explicitly uses the "--ignore-submodules=all" command line option, in that case the submodule output is still suppressed. This also makes "git commit" work again when only modifications of ignored submodules are staged, as that command uses the "commitable" member of the wt_status struct to determine if staged changes are present. But this only happens when the commit command uses the wt_status* functions to produce status output for human consumption (when forking an editor or with --dry-run), in all other cases (e.g. when run in a script with '-m') another code path is taken which uses index_differs_from() to determine if any changes are staged which still ignores submodules according to their configuration. This will be fixed in a follow-up commit.

Change t7508 to reflect this new behavior and add three new tests to show that a single staged submodule configured to be ignored will be committed when the status output is generated and won't be if not. Also update the documentation of the ignore config options accordingly.

---
 Documentation/config.txt     |  8 +++--
 Documentation/gitmodules.txt |  4 ++-
 t/t7508-status.sh            | 74 ++++++++++++++++++++++++++++++++++++++++++--
 wt-status.c                  | 12 ++++++-
 4 files changed, 92 insertions(+), 6 deletions(-)

index 84c7e3f..171a98e 100644
@@ -2289,7 +2289,9 @@ status.submodulesummary::
 	--summary-limit option of linkgit:git-submodule[1]). Please note
 	that the summary output command will be suppressed for all
 	submodules when `diff.ignoreSubmodules` is set to 'all' or only
-	for those submodules where `submodule.<name>.ignore=all`. To
+	for those submodules where `submodule.<name>.ignore=all`. The only
+	exception to that rule is that status and commit will show staged
+	submodule changes. To
 	also view the summary for ignored submodules you can either use
 	the --ignore-submodules=dirty command line option or the 'git
 	submodule summary' command, which shows a similar output but does

@@ -2320,7 +2322,9 @@ submodule.<name>.fetchRecurseSubmodules::
 submodule.<name>.ignore::
 	Defines under what circumstances "git status" and the diff family show
 	a submodule as modified. When set to "all", it will never be considered
-	modified, "dirty" will ignore all changes to the submodules work tree and
+	modified (but it will nonetheless show up in the output of status and
+	commit when it has been staged), "dirty" will ignore all changes
+	to the submodules work tree and
 	takes only differences between the HEAD of the submodule and the commit
 	recorded in the superproject into account. "untracked" will additionally
 	let submodules with modified tracked files in their work tree show up.

index 347a9f7..f6c0dfd 100644
@@ -67,7 +67,9 @@ submodule.<name>.fetchRecurseSubmodules::
 submodule.<name>.ignore::
 	Defines under what circumstances "git status" and the diff family show
 	a submodule as modified. When set to "all", it will never be considered
-	modified, "dirty" will ignore all changes to the submodules work tree and
+	modified (but will nonetheless show up in the output of status and
+	commit when it has been staged), "dirty" will ignore all changes
+	to the submodules work tree and
 	takes only differences between the HEAD of the submodule and the commit
 	recorded in the superproject into account. "untracked" will additionally
 	let submodules with modified tracked files in their work tree show up.

index ec7344e..86fec89 100644
@@ -519,9 +519,19 @@ static void wt_status_collect_changes_index(struct wt_status *s)
 	opt.def = s->is_initial ? EMPTY_TREE_SHA1_HEX : s->reference;
 	setup_revisions(0, NULL, &rev, &opt);
 
+	DIFF_OPT_SET(&rev.diffopt, OVERRIDE_SUBMODULE_CONFIG);
 	if (s->ignore_submodule_arg) {
-		DIFF_OPT_SET(&rev.diffopt, OVERRIDE_SUBMODULE_CONFIG);
 		handle_ignore_submodules_arg(&rev.diffopt, s->ignore_submodule_arg);
+	} else {
+		/*
+		 * Unless the user did explicitly request a submodule ignore
+		 * mode by passing a command line option we do not ignore any
+		 * changed submodule SHA-1s when comparing index and HEAD, no
+		 * matter what is configured. Otherwise the user won't be
+		 * shown any submodules she manually added (and which are
+		 * staged to be committed), which would be really confusing.
+		 */
+		handle_ignore_submodules_arg(&rev.diffopt, "dirty");
 	}
 
 	rev.diffopt.output_format |= DIFF_FORMAT_CALLBACK;

1.9.1.476.g510abc7

The previous commit fixed the problem that the staged but that ignored submodules did not show up in the status output of the commit command and weren't committed afterwards either. But when commit doesn't generate the status output (e.g. when used in a script with '-m') the ignored submodule will still not be committed. This is because in that case a different code path is taken which calls index_differs_from() instead of calling the wt_status functions.

Fix that by calling index_differs_from() from builtin/commit.c with a diff_options argument value that tells it not ignore any submodule changes unless the '--ignore-submodules' option is used. Even though this option isn't yet implemented for cmd_commit() but only for cmd_status() this prepares cmd_commit() to correctly handle the '--ignore-submodules' option later. As status and commit share the same ignore_submodule_arg variable this makes the code more robust against accidental breakage and documents how to correctly call index_differs_from().

Change the expected result of the test documenting this problem from failure to success.

---
 builtin/commit.c  | 18 ++++++++++++++++--
 t/t7508-status.sh |  2 +-
 2 files changed, 17 insertions(+), 3 deletions(-)

index d9550c5..a456a60 100644
@@ -833,8 +833,22 @@ static int prepare_to_commit(const char *index_file, const char *prefix,
 
 		if (get_sha1(parent, sha1))
 			commitable = !!active_nr;
-		else
-			commitable = index_differs_from(parent, 0);
+		else {
+			/*
+			 * Unless the user did explicitly request a submodule
+			 * ignore mode by passing a command line option we do
+			 * not ignore any changed submodule SHA-1s when
+			 * comparing index and parent, no matter what is
+			 * configured. Otherwise we won't commit any
+			 * submodules which were manually staged, which would
+			 * be really confusing.
+			 */
+			int diff_flags = DIFF_OPT_OVERRIDE_SUBMODULE_CONFIG;
+			if (ignore_submodule_arg &&
+			    !strcmp(ignore_submodule_arg, "all"))
+				diff_flags |= DIFF_OPT_IGNORE_SUBMODULES;
+			commitable = index_differs_from(parent, diff_flags);
+		}
 	}
 	strbuf_release(&committer_ident);
 

index e6483fc..d480069 100755
@@ -1523,7 +1523,7 @@ EOF
 	test_i18ngrep "^M. sm" output
 '
 
-test_expect_failure 'git commit -m will commit a staged but ignored submodule' '
+test_expect_success 'git commit -m will commit a staged but ignored submodule' '
 	git commit -uno -m message &&
 	git status -s --ignore-submodules=dirty >output &&
 	 test_i18ngrep ! "^M. sm" output &&

1.9.1.476.g510abc7