Skip to contentSkip to conversation

git --paginate: do not commit pager choice too early

PATCH In review Jonathan Nieder 6 participants 42 messages Jun 15, 2016

lore
Share & tools
lore
#0 lore raw

cover git --paginate: do not commit pager choice too early

Here’s a fix from the famous setup cleanup series[1].

The problem it fixes is somewhat obscure: ‘git -p foo’ is not paying attention to the repository-local ‘[core] pager’ configuration when run from a subdirectory. But it is a real bug, and the fix gives an example of how to deal with repository setup and should be safe.

Patch is against master. There is a small semantic conflict with jn/grep-open: SIMPLEPAGER should be changed to SIMPLEPAGERTTY in the prerequisites for the test_default_pager function. Please let me know if I should push a merge commit to help resolve that.

Thoughts?
Jonathan Nieder (3):
t7006 (pager): introduce helper for parameterized tests
t7006: test pager configuration for several git commands
tests: local config file should be honored from subdirs of toplevel

Nguyễn Thái Ngọc Duy (1):
git --paginate: do not commit pager choice too early

 git.c            |    2 +-
 t/t7006-pager.sh |  234 +++++++++++++++++++++++++++++++++++++++++------------
 2 files changed, 182 insertions(+), 54 deletions(-)

[1] http://thread.gmane.org/gmane.comp.version-control.git/144778

#1 lore raw

PATCH 1/4 t7006 (pager): introduce helper for parameterized tests

The current tests test pager configuration for ‘git log’, but other commands use a different setup procedure and should therefore be tested separately. Add a helper to make this easier.

This patch introduces the helper and changes some existing tests to use it. The only functional change should be the introduction of ‘git log - ’ to a few test descriptions.

---
 t/t7006-pager.sh |   72 ++++++++++++++++++++++++++++++++++++++++-------------
 1 files changed, 54 insertions(+), 18 deletions(-)
t/t7006-pager.sh +54 -18
index 9a83241..b117ebb 100755
@@ -172,58 +172,94 @@ then
 	test_set_prereq SIMPLEPAGER
 fi
 
-test_expect_success SIMPLEPAGER 'default pager is used by default' '
+# Use this helper to make it easy for the caller of your
+# terminal-using function to specify whether it should fail.
+# If you write
+#
+#	your_test() {
+#		parse_args "$@"
+#
+#		$test_expectation "$cmd - behaves well" "
+#			...
+#			$full_command &&
+#			...
+#		"
+#	}
+#
+# then your test can be used like this:
+#
+#	your_test expect_(success|failure) [test_must_fail] 'git foo'
+#
+parse_args() {
+	test_expectation="test_$1"
+	shift
+	if test "$1" = test_must_fail
+	then
+		full_command="test_must_fail test_terminal "
+		shift
+	else
+		full_command="test_terminal "
+	fi
+	cmd=$1
+	full_command="$full_command $1"
+}
+
+parse_args expect_success 'git log'
+$test_expectation SIMPLEPAGER "$cmd - default pager is used by default" "
 	unset PAGER GIT_PAGER;
 	test_might_fail git config --unset core.pager &&
 	rm -f default_pager_used ||
 	cleanup_fail &&
 
-	cat >$less <<-\EOF &&
+	cat >\$less <<-\EOF &&
 	#!/bin/sh
 	wc >default_pager_used
 	EOF
-	chmod +x $less &&
+	chmod +x \$less &&
 	(
-		PATH=.:$PATH &&
+		PATH=.:\$PATH &&
 		export PATH &&
-		test_terminal git log
+		$full_command
 	) &&
 	test -e default_pager_used
-'
+"
 
-test_expect_success TTY 'PAGER overrides default pager' '
+parse_args expect_success 'git log'
+$test_expectation TTY "$cmd - PAGER overrides default pager" "
 	unset GIT_PAGER;
 	test_might_fail git config --unset core.pager &&
 	rm -f PAGER_used ||
 	cleanup_fail &&
 
-	PAGER="wc >PAGER_used" &&
+	PAGER='wc >PAGER_used' &&
 	export PAGER &&
-	test_terminal git log &&
+	$full_command &&
 	test -e PAGER_used
-'
+"
 
-test_expect_success TTY 'core.pager overrides PAGER' '
+parse_args expect_success 'git log'
+$test_expectation TTY "$cmd - core.pager overrides PAGER" "
 	unset GIT_PAGER;
 	rm -f core.pager_used ||
 	cleanup_fail &&
 
 	PAGER=wc &&
 	export PAGER &&
-	git config core.pager "wc >core.pager_used" &&
-	test_terminal git log &&
+	git config core.pager 'wc >core.pager_used' &&
+	$full_command &&
 	test -e core.pager_used
-'
+"
 
-test_expect_success TTY 'GIT_PAGER overrides core.pager' '
+parse_args expect_success 'git log'
+$test_expectation TTY "$cmd - GIT_PAGER overrides core.pager" "
 	rm -f GIT_PAGER_used ||
 	cleanup_fail &&
 
 	git config core.pager wc &&
-	GIT_PAGER="wc >GIT_PAGER_used" &&
+	GIT_PAGER='wc >GIT_PAGER_used' &&
 	export GIT_PAGER &&
-	test_terminal git log &&
+	$full_command &&
 	test -e GIT_PAGER_used
-'
+"
 
 test_done
Signature · 1 line
1.7.1.579.ge2549
#2 lore raw

PATCH 2/4 t7006: test pager configuration for several git commands

In reply to Jonathan Nieder

Test choice of pager at several stages of repository setup. This provides some (admittedly uninteresting) examples to keep in mind when considering changes to the setup procedure.

---
The patch from ‘diff -w’ should be easier to review. I’ll send it as a reply.

Thanks to both Junio and Hannes for improvements.

 t/t7006-pager.sh |  146 +++++++++++++++++++++++++++++++++---------------------
 1 files changed, 89 insertions(+), 57 deletions(-)
t/t7006-pager.sh +89 -57
index b117ebb..4420f91 100755
@@ -204,62 +204,94 @@ parse_args() {
 	full_command="$full_command $1"
 }
 
-parse_args expect_success 'git log'
-$test_expectation SIMPLEPAGER "$cmd - default pager is used by default" "
-	unset PAGER GIT_PAGER;
-	test_might_fail git config --unset core.pager &&
-	rm -f default_pager_used ||
-	cleanup_fail &&
-
-	cat >\$less <<-\EOF &&
-	#!/bin/sh
-	wc >default_pager_used
-	EOF
-	chmod +x \$less &&
-	(
-		PATH=.:\$PATH &&
-		export PATH &&
-		$full_command
-	) &&
-	test -e default_pager_used
-"
-
-parse_args expect_success 'git log'
-$test_expectation TTY "$cmd - PAGER overrides default pager" "
-	unset GIT_PAGER;
-	test_might_fail git config --unset core.pager &&
-	rm -f PAGER_used ||
-	cleanup_fail &&
-
-	PAGER='wc >PAGER_used' &&
-	export PAGER &&
-	$full_command &&
-	test -e PAGER_used
-"
-
-parse_args expect_success 'git log'
-$test_expectation TTY "$cmd - core.pager overrides PAGER" "
-	unset GIT_PAGER;
-	rm -f core.pager_used ||
-	cleanup_fail &&
-
-	PAGER=wc &&
-	export PAGER &&
-	git config core.pager 'wc >core.pager_used' &&
-	$full_command &&
-	test -e core.pager_used
-"
-
-parse_args expect_success 'git log'
-$test_expectation TTY "$cmd - GIT_PAGER overrides core.pager" "
-	rm -f GIT_PAGER_used ||
-	cleanup_fail &&
-
-	git config core.pager wc &&
-	GIT_PAGER='wc >GIT_PAGER_used' &&
-	export GIT_PAGER &&
-	$full_command &&
-	test -e GIT_PAGER_used
-"
+test_default_pager() {
+	parse_args "$@"
+
+	$test_expectation SIMPLEPAGER "$cmd - default pager is used by default" "
+		unset PAGER GIT_PAGER;
+		test_might_fail git config --unset core.pager &&
+		rm -f default_pager_used ||
+		cleanup_fail &&
+
+		cat >\$less <<-\EOF &&
+		#!/bin/sh
+		wc >default_pager_used
+		EOF
+		chmod +x \$less &&
+		(
+			PATH=.:\$PATH &&
+			export PATH &&
+			$full_command
+		) &&
+		test -e default_pager_used
+	"
+}
+
+test_PAGER_overrides() {
+	parse_args "$@"
+
+	$test_expectation TTY "$cmd - PAGER overrides default pager" "
+		unset GIT_PAGER;
+		test_might_fail git config --unset core.pager &&
+		rm -f PAGER_used ||
+		cleanup_fail &&
+
+		PAGER='wc >PAGER_used' &&
+		export PAGER &&
+		$full_command &&
+		test -e PAGER_used
+	"
+}
+
+test_core_pager_overrides() {
+	parse_args "$@"
+
+	$test_expectation TTY "$cmd - core.pager overrides PAGER" "
+		unset GIT_PAGER;
+		rm -f core.pager_used ||
+		cleanup_fail &&
+
+		PAGER=wc &&
+		export PAGER &&
+		git config core.pager 'wc >core.pager_used' &&
+		$full_command &&
+		test -e core.pager_used
+	"
+}
+
+test_GIT_PAGER_overrides() {
+	parse_args "$@"
+
+	$test_expectation TTY "$cmd - GIT_PAGER overrides core.pager" "
+		rm -f GIT_PAGER_used ||
+		cleanup_fail &&
+
+		git config core.pager wc &&
+		GIT_PAGER='wc >GIT_PAGER_used' &&
+		export GIT_PAGER &&
+		$full_command &&
+		test -e GIT_PAGER_used
+	"
+}
+
+test_default_pager        expect_success 'git log'
+test_PAGER_overrides      expect_success 'git log'
+test_core_pager_overrides expect_success 'git log'
+test_GIT_PAGER_overrides  expect_success 'git log'
+
+test_default_pager        expect_success 'git -p log'
+test_PAGER_overrides      expect_success 'git -p log'
+test_core_pager_overrides expect_success 'git -p log'
+test_GIT_PAGER_overrides  expect_success 'git -p log'
+
+test_default_pager        expect_success test_must_fail 'git -p'
+test_PAGER_overrides      expect_success test_must_fail 'git -p'
+test_core_pager_overrides expect_success test_must_fail 'git -p'
+test_GIT_PAGER_overrides  expect_success test_must_fail 'git -p'
+
+test_default_pager        expect_success test_must_fail 'git -p nonsense'
+test_PAGER_overrides      expect_success test_must_fail 'git -p nonsense'
+test_core_pager_overrides expect_success test_must_fail 'git -p nonsense'
+test_GIT_PAGER_overrides  expect_success test_must_fail 'git -p nonsense'
 
 test_done
Signature · 1 line
1.7.1.579.ge2549
#3 lore raw

Re: [PATCH 2/4] t7006: test pager configuration for several git commands

Quoted 1 line · Jonathan Nieder

Test choice of pager at several stages of repository setup.

Here it is again, without whitespace changes. :)

 t/t7006-pager.sh |   40 ++++++++++++++++++++++++++++++++++++----
 1 files changed, 36 insertions(+), 4 deletions(-)
t/t7006-pager.sh +36 -4
index b117ebb..4420f91 100755
@@ -204,7 +204,9 @@ parse_args() {
 	full_command="$full_command $1"
 }
 
-parse_args expect_success 'git log'
+test_default_pager() {
+	parse_args "$@"
+
 $test_expectation SIMPLEPAGER "$cmd - default pager is used by default" "
 	unset PAGER GIT_PAGER;
 	test_might_fail git config --unset core.pager &&
@@ -223,8 +225,11 @@ $test_expectation SIMPLEPAGER "$cmd - default pager is used by default" "
 	) &&
 	test -e default_pager_used
 "
+}
+
+test_PAGER_overrides() {
+	parse_args "$@"
 
-parse_args expect_success 'git log'
 $test_expectation TTY "$cmd - PAGER overrides default pager" "
 	unset GIT_PAGER;
 	test_might_fail git config --unset core.pager &&
@@ -236,8 +241,11 @@ $test_expectation TTY "$cmd - PAGER overrides default pager" "
 	$full_command &&
 	test -e PAGER_used
 "
+}
+
+test_core_pager_overrides() {
+	parse_args "$@"
 
-parse_args expect_success 'git log'
 $test_expectation TTY "$cmd - core.pager overrides PAGER" "
 	unset GIT_PAGER;
 	rm -f core.pager_used ||
@@ -249,8 +257,11 @@ $test_expectation TTY "$cmd - core.pager overrides PAGER" "
 	$full_command &&
 	test -e core.pager_used
 "
+}
+
+test_GIT_PAGER_overrides() {
+	parse_args "$@"
 
-parse_args expect_success 'git log'
 $test_expectation TTY "$cmd - GIT_PAGER overrides core.pager" "
 	rm -f GIT_PAGER_used ||
 	cleanup_fail &&
@@ -261,5 +272,26 @@ $test_expectation TTY "$cmd - GIT_PAGER overrides core.pager" "
 	$full_command &&
 	test -e GIT_PAGER_used
 "
+}
+
+test_default_pager        expect_success 'git log'
+test_PAGER_overrides      expect_success 'git log'
+test_core_pager_overrides expect_success 'git log'
+test_GIT_PAGER_overrides  expect_success 'git log'
+
+test_default_pager        expect_success 'git -p log'
+test_PAGER_overrides      expect_success 'git -p log'
+test_core_pager_overrides expect_success 'git -p log'
+test_GIT_PAGER_overrides  expect_success 'git -p log'
+
+test_default_pager        expect_success test_must_fail 'git -p'
+test_PAGER_overrides      expect_success test_must_fail 'git -p'
+test_core_pager_overrides expect_success test_must_fail 'git -p'
+test_GIT_PAGER_overrides  expect_success test_must_fail 'git -p'
+
+test_default_pager        expect_success test_must_fail 'git -p nonsense'
+test_PAGER_overrides      expect_success test_must_fail 'git -p nonsense'
+test_core_pager_overrides expect_success test_must_fail 'git -p nonsense'
+test_GIT_PAGER_overrides  expect_success test_must_fail 'git -p nonsense'
 
 test_done
Signature · 1 line
1.7.1.579.ge2549
#4 lore raw

PATCH 3/4 tests: local config file should be honored from subdirs of toplevel

In reply to Jonathan Nieder

When git is passed the --paginate option, starting up a pager requires deciding what pager to start, which requires access to the core.pager configuration. If --paginate is handled before searching for the git dir, this configuration will be missed.

In other words, with --paginate and only with --paginate, any repository-local core.pager setting is being ignored [*].

[*] unless the git directory is ./.git or GIT_DIR or GIT_CONFIG was set explicitly.

Add a test to demonstrate this counterintuitive behavior. Noticed while reading over a patch by Duy that fixes it.

---
 t/t7006-pager.sh |   26 ++++++++++++++++++++++++++
 1 files changed, 26 insertions(+), 0 deletions(-)
t/t7006-pager.sh +26 -0
index 4420f91..2b106be 100755
@@ -259,6 +259,28 @@ test_core_pager_overrides() {
 	"
 }
 
+test_core_pager_subdir() {
+	parse_args "$@"
+
+	$test_expectation TTY "$cmd - core.pager from subdirectory" "
+		unset GIT_PAGER;
+		rm -f core.pager_used &&
+		rm -fr sub ||
+		cleanup_fail &&
+
+		PAGER=wc &&
+		stampname=\$(pwd)/core.pager_used &&
+		export PAGER stampname &&
+		git config core.pager 'wc >\"\$stampname\"' &&
+		mkdir sub &&
+		(
+			cd sub &&
+			$full_command
+		) &&
+		test -e core.pager_used
+	"
+}
+
 test_GIT_PAGER_overrides() {
 	parse_args "$@"
 
@@ -277,21 +299,25 @@ test_GIT_PAGER_overrides() {
 test_default_pager        expect_success 'git log'
 test_PAGER_overrides      expect_success 'git log'
 test_core_pager_overrides expect_success 'git log'
+test_core_pager_subdir    expect_success 'git log'
 test_GIT_PAGER_overrides  expect_success 'git log'
 
 test_default_pager        expect_success 'git -p log'
 test_PAGER_overrides      expect_success 'git -p log'
 test_core_pager_overrides expect_success 'git -p log'
+test_core_pager_subdir    expect_failure 'git -p log'
 test_GIT_PAGER_overrides  expect_success 'git -p log'
 
 test_default_pager        expect_success test_must_fail 'git -p'
 test_PAGER_overrides      expect_success test_must_fail 'git -p'
 test_core_pager_overrides expect_success test_must_fail 'git -p'
+test_core_pager_subdir    expect_failure test_must_fail 'git -p'
 test_GIT_PAGER_overrides  expect_success test_must_fail 'git -p'
 
 test_default_pager        expect_success test_must_fail 'git -p nonsense'
 test_PAGER_overrides      expect_success test_must_fail 'git -p nonsense'
 test_core_pager_overrides expect_success test_must_fail 'git -p nonsense'
+test_core_pager_subdir    expect_failure test_must_fail 'git -p nonsense'
 test_GIT_PAGER_overrides  expect_success test_must_fail 'git -p nonsense'
 
 test_done
Signature · 1 line
1.7.1.579.ge2549
#5 lore raw

PATCH 4/4 git --paginate: do not commit pager choice too early

In reply to Jonathan Nieder

From: Nguyễn Thái Ngọc Duy <[email protected]>

When git is passed the --paginate option, starting up a pager requires deciding what pager to start, which requires access to the core.pager configuration.

At the relevant moment, the repository has not been searched for yet. Attempting to access the configuration at this point results in git_dir being set to .git [*], which is almost certainly not what was wanted. In particular, when run from a subdirectory of the toplevel, git --paginate does not respect the core.pager setting from the current repository.

[*] unless GIT_DIR or GIT_CONFIG is set

So delay the pager startup when possible:

1. run_argv() already commits pager choice inside run_builtin() if a command is found. For commands that use RUN_SETUP, waiting until then fixes the problem described above: once git knows where to look, it happily respects the core.pager setting.

2. list_common_cmds_help() prints out 29 lines and exits. This can benefit from pagination, so we need to commit the pager choice before writing this output.

Luckily ‘git’ without subcommand has no other reason to access a repository, so it would be intuitive to ignore repository-local configuration in this case. Simpler for now to choose a pager using the funny code that notices a repository that happens to be at .git. That this accesses a repository when it is very convenient to is a bug but not an important one.

3. help_unknown_cmd() prints out a few lines to stderr. It is not important to paginate this, so don’t.

---
Thanks for reading.

 git.c            |    2 +-
 t/t7006-pager.sh |   58 ++++++++++++++++++++++++++++++++++++++++++-----------
 2 files changed, 47 insertions(+), 13 deletions(-)
git.c +1 -1
index 99f0363..25e7693 100644
@@ -511,12 +511,12 @@ int main(int argc, const char **argv)
 	argv++;
 	argc--;
 	handle_options(&argv, &argc, NULL);
-	commit_pager_choice();
 	if (argc > 0) {
 		if (!prefixcmp(argv[0], "--"))
 			argv[0] += 2;
 	} else {
 		/* The user didn't specify a command; give them help */
+		commit_pager_choice();
 		printf("usage: %s\n\n", git_usage_string);
 		list_common_cmds_help();
 		printf("\n%s\n", git_more_info_string);
t/t7006-pager.sh +46 -12

101 lines not on this page · raw message

Signature · 1 line
1.7.1.579.ge2549
#6 lore raw
In reply to Jonathan Nieder
Quoted 11 lines · Jonathan Nieder

Here’s a fix from the famous setup cleanup series[1].

The problem it fixes is somewhat obscure: ‘git -p foo’ is not paying attention to the repository-local ‘[core] pager’ configuration when run from a subdirectory. But it is a real bug, and the fix gives an example of how to deal with repository setup and should be safe.

Patch is against master. There is a small semantic conflict with jn/grep-open: SIMPLEPAGER should be changed to SIMPLEPAGERTTY in the prerequisites for the test_default_pager function. Please let me know if I should push a merge commit to help resolve that.

I just read through the patches, and it seems like a reasonable fix. So in that sense, ACK.

But reading the message for patch 4/4, I can't help but wonder if the right way forward is for the git wrapper to _always_ find the repository as the very first thing. I.e., make finding the repository a non-destructive thing (especially don't end in a different directory), and just do it before we run any other code.

For a few commands like "init" or "clone", we would end up just ignoring this information, but the worst case should be a little bit of wasted effort.

And then all these silly setup and timing issues could just go away, as we would know that GIT_DIR and GIT_WORK_TREE would be set properly.

-Peff

#7 lore raw
Quoted 3 lines · Jeff King

But reading the message for patch 4/4, I can't help but wonder if the right way forward is for the git wrapper to _always_ find the repository as the very first thing.

[...]

Quoted 2 lines

the worst case should be a little bit of
wasted effort.

This is a very useful thought. Git without the chdir() to toplevel would be pleasanter, I think. (Yes, I realize you weren’t necessarily suggesting suppressing a chdir_to_toplevel() in place of setup_git_repository() and friends.)

Regarding the repository search: automounters can cause pain if DISCOVERY_ACROSS_FILESYSTEM is set and GIT_CEILING_DIRECTORIES is not set appropriately. An important script that does

  tmpdir=$(mktemp -d)
  cd "$tmpdir"
  ...

  git diff --no-index a b
  ...

could hang indefinitely if some nut has created a named pipe at /tmp/.git. I haven’t come up with any non far-fetched reason to mind the repository search, though.

Will think about this.

#8 lore raw
Quoted 11 lines · Jonathan Nieder
Quoted 3 lines

But reading the message for patch 4/4, I can't help but wonder if the right way forward is for the git wrapper to _always_ find the repository as the very first thing.

[...]

Quoted 2 lines

the worst case should be a little bit of
wasted effort.

This is a very useful thought. Git without the chdir() to toplevel would be pleasanter, I think. (Yes, I realize you weren’t necessarily suggesting suppressing a chdir_to_toplevel() in place of setup_git_repository() and friends.)

I had just meant that we would not do the chdir() initially, but would do so right before running the actual command which wanted repository setup (and commands like init which do not do that setup would never chdir to the toplevel). But we must always run at least aliases and external sub-commands from the toplevel to keep backwards compatibility.

Quoted 11 lines

Regarding the repository search: automounters can cause pain if DISCOVERY_ACROSS_FILESYSTEM is set and GIT_CEILING_DIRECTORIES is not set appropriately. An important script that does

  tmpdir=$(mktemp -d)
  cd "$tmpdir"
  ...

  git diff --no-index a b
  ...
Quoted 5 lines

could hang indefinitely if some nut has created a named pipe at /tmp/.git. I haven’t come up with any non far-fetched reason to mind the repository search, though.

Will think about this.

Yeah, I considered that, too. But I doubt it is a big problem. If it were, then just running "git log" when you are not in a repository would be similarly painful. The people who have setups like that are already having to deal with GIT_CEILING_DIRECTORIES to make things less painful (though I imagine the new cross-filesystem discovery code will prevent most of the pain without them having to actually do anything).

So yes, we are adding the extra lookup for commands like "git clone", but I suspect in practice nobody will care. If it is a big deal, we can do something like:

if (!strcmp(cmd, "clone") || !strcmp(cmd, "init"))

     ... don't do setup ...

which I admit is a terrible hack, but it is an optimization, not a correctness thing. So we are not shutting out external user-defined commands that might not care about the repo. We are just making our common built-in ones a bit more efficient.

-Peff

#9 lore raw

2010/6/28 Jeff King <[email protected]>:

Quoted 14 lines · Jonathan Nieder
Quoted 4 lines
Quoted 4 lines

But reading the message for patch 4/4, I can't help but wonder if the right way forward is for the git wrapper to _always_ find the repository as the very first thing.

Quoted 8 lines

[...]

Quoted 2 lines

the worst case should be a little bit of
wasted effort.

This is a very useful thought.  Git without the chdir() to toplevel would be pleasanter, I think.  (Yes, I realize you weren’t necessarily suggesting suppressing a chdir_to_toplevel() in place of setup_git_repository() and friends.)

Quoted 5 lines

I had just meant that we would not do the chdir() initially, but would do so right before running the actual command which wanted repository setup (and commands like init which do not do that setup would never chdir to the toplevel). But we must always run at least aliases and external sub-commands from the toplevel to keep backwards compatibility.

I think I'm missing something here. Looking up aliases means repository search must be done in git wrapper anyway. Even if we don't have to do repo search in git wrapper, the very first thing a sub command does is likely git_config(), which will need repo again. As long as repo search does not have any side effects, everything should be fine, IMO.

Quoted 7 lines

So yes, we are adding the extra lookup for commands like "git clone", but I suspect in practice nobody will care. If it is a big deal, we can do something like:

 if (!strcmp(cmd, "clone") || !strcmp(cmd, "init")) ... don't do setup ...

If git wrapper does not need setup (i.e. aliases) at all, that should happen without hacks. "clone" and "init" do not have RUN_SETUP (nor RUN_SETUP_GENTLY in my series), so there should not be any setup before cmd_{init,clone} is run.

PS. I haven't forgotten my cleanup series. Day job has been taking too much of my energy unfortunately.

Signature · 1 line
Duy
#10 lore raw
In reply to Jonathan Nieder
Quoted 4 lines · Jonathan Nieder

Patch is against master. There is a small semantic conflict with jn/grep-open: SIMPLEPAGER should be changed to SIMPLEPAGERTTY in the prerequisites for the test_default_pager function. Please let me know if I should push a merge commit to help resolve that.

Thanks for advance warning; please double check the merge result in 'pu' when I push it out...

#11 lore raw
Quoted 9 lines · Junio C Hamano
Quoted 4 lines · Jonathan Nieder

Patch is against master. There is a small semantic conflict with jn/grep-open: SIMPLEPAGER should be changed to SIMPLEPAGERTTY in the prerequisites for the test_default_pager function. Please let me know if I should push a merge commit to help resolve that.

Thanks for advance warning; please double check the merge result in 'pu' when I push it out...

I hate an enumeration that pretends to be exhausitive but is not.

So delay the pager startup when possible:

1. run_argv() already commits pager choice inside run_builtin() if a command is found. For commands that use RUN_SETUP, waiting until then fixes the problem described above: once git knows where to look, it happily respects the core.pager setting.

... and for commands that do not use RUN_SETUP, what happens?

2. list_common_cmds_help() prints out 29 lines and exits....
3. help_unknown_cmd() prints out a few lines to stderr. It is not important to paginate this, so don’t.

Missing from the above enumeration are are external commands. They depend on commit_pager_choice() to be called before execv_dashed_external() gets called. For example, "git -p request-pull $args" no longer works with this patch.

Sigh..

#12 lore raw
Quoted 1 line · Junio C Hamano

... and for commands that do not use RUN_SETUP, what happens?

run_builtin() still commits pager choice. The bug is neither fixed nor made worse for them.

Quoted 4 lines

Missing from the above enumeration are are external commands. They depend on commit_pager_choice() to be called before execv_dashed_external() gets called. For example, "git -p request-pull $args" no longer works with this patch.

Something like the following may help.

This does not protect against calling setup_pager() more than once. Once the first pager has been set up, isatty(1) is false, preventing additional pagers from being spawned and competing with it.

As futureproofing against --paginate=always, setup_pager() should probably be taught to check pager_in_use(), but that should probably wait for a separate patch.

git.c +2 -0
index 25e7693..2dcd952 100644
@@ -162,6 +162,7 @@ static int handle_alias(int *argcp, const char ***argv)
 	int unused_nongit;
 
 	subdir = setup_git_directory_gently(&unused_nongit);
+	commit_pager_choice();
 
 	alias_command = (*argv)[0];
 	alias_string = alias_lookup(alias_command);
@@ -433,6 +434,7 @@ static void execv_dashed_external(const char **argv)
 	int status;
 
 	strbuf_addf(&cmd, "git-%s", argv[0]);
+	commit_pager_choice();
 
 	/*
 	 * argv[0] must be the git command, but the argv array
#13 lore raw

PATCH 1/1 git --paginate: paginate external commands again

In reply to Junio C Hamano

73e25e7c (git --paginate: do not commit pager choice too early, 2010-06-26) failed to take some cases into account.

1b. Builtins that do not use RUN_SETUP (like git config) do
not find GIT_DIR set correctly when the pager is launched from run_builtin(). So the core.pager configuration is not honored from subdirectories of the toplevel for them.

4a. External git commands (like git request-pull) relied on the
early pager launch to take care of handling the -p option. Ever since 73e25e7c, they do not honor the -p option at all.

4b. Commands invoked through ! aliases (like ls) were also relying
on the early pager launch.

Fix (4a) by launching the pager (if requested) before running such a “dashed external”. For simplicity, this still does not search for a .git directory before running the external command; when run from a subdirectory of the toplevel, therefore, the “[core] pager” configuration is still not honored.

Fix (4b) by launching pager if requested before carrying out such an alias. Actually doing this has no effect, since the pager (if any) would have already been launched in a failed attempt to try a dashed external first. The choice-of-pager-not-honored-from- subdirectory bug still applies here, too.

(1b) is not a regression. There is no need to fix it yet.

Noticed by Junio.

---

Quoted 1 line · Junio C Hamano

I hate an enumeration that pretends to be exhausitive but is not.

Sorry about that.

Here’s a patch that could be squashed in or applied on top.

 git.c            |    3 +++
 t/t7006-pager.sh |   47 +++++++++++++++++++++++++++++++++++------------
 2 files changed, 38 insertions(+), 12 deletions(-)
git.c +3 -0
index 25e7693..99517dd 100644
@@ -167,6 +167,7 @@ static int handle_alias(int *argcp, const char ***argv)
 	alias_string = alias_lookup(alias_command);
 	if (alias_string) {
 		if (alias_string[0] == '!') {
+			commit_pager_choice();
 			if (*argcp > 1) {
 				struct strbuf buf;
 
@@ -432,6 +433,8 @@ static void execv_dashed_external(const char **argv)
 	const char *tmp;
 	int status;
 
+	commit_pager_choice();
+
 	strbuf_addf(&cmd, "git-%s", argv[0]);
 
 	/*
t/t7006-pager.sh +35 -12

58 lines not on this page · raw message

Signature · 1 line
1.7.2.rc2.532.g65d9b.dirty
#14 lore raw
In reply to Junio C Hamano

2010/7/15 Junio C Hamano <[email protected]>:

Quoted 6 lines

Missing from the above enumeration are are external commands.  They depend on commit_pager_choice() to be called before execv_dashed_external() gets called.  For example, "git -p request-pull $args" no longer works with this patch.

Sigh..

I'm sorry. I missed those paths..

Signature · 1 line
Duy
#15 lore raw
In reply to Junio C Hamano
Quoted 8 lines · Junio C Hamano

So delay the pager startup when possible:

1. run_argv() already commits pager choice inside run_builtin() if a command is found. For commands that use RUN_SETUP, waiting until then fixes the problem described above: once git knows where to look, it happily respects the core.pager setting.

... and for commands that do not use RUN_SETUP, what happens?

Let’s help some of them out a little.

These patches teach a few more built-ins to do something like RUN_SETUP; more precisely, a RUN_SETUP_GENTLY facility is introduced to run setup_git_directory_gently() early just like NEEDS_PREFIX has always caused setup_git_directory() to be run early.

This series wouldn’t be possible without Duy’s recent efforts to roll out other fixes from the famous nd/setup topic --- thanks! The patches should be familiar[1] and I think they are better justified now.

The first 8 patches apply on top to jn/paginate-fix from pu. Between patch 8 and 9 I have a merge of jn/maint-setup-fix^, with the following explanation:

Merge branch 'jn/maint-setup-fix' (early part) into jn/paginate-fix

That topic includes a fix for “git index-pack” (which used to save the cwd before the repository search to work around a bug) which is needed before the repository search can be safely run earlier.

That merge involves a trivial conflict between adding the RUN_SETUP_GENTLY flag for this topic and removing USE_PAGER for grep -O. The conflict is only nominal --- the two changes do not interfere with each other.

One final note: this series is deliberately very conservative, in that it does not touch commands like diff (which has a --no-index mode suppressing the repository search). Such commands are neither helped nor hindered by this series, and I think that during this transition time we should be opportunistic rather than rigid: the rule is to run the repository search as soon as can be easily justified but no sooner.

Thoughts welcome, as always.

Nguyễn Thái Ngọc Duy (12):
git wrapper: introduce startup_info struct
setup: remember whether repository was found
git wrapper: allow setup_git_directory_gently() be called earlier
shortlog: run setup_git_directory_gently() sooner
grep: run setup_git_directory_gently() sooner
apply: run setup_git_directory_gently() sooner
bundle: run setup_git_directory_gently() sooner
config: run setup_git_directory_gently() sooner
index-pack: run setup_git_directory_gently() sooner
ls-remote: run setup_git_directory_gently() sooner
var: run setup_git_directory_gently() sooner
merge-file: run setup_git_directory_gently() sooner

[1] http://thread.gmane.org/gmane.comp.version-control.git/144000/focus=144110

#16 lore raw

PATCH 1/12 git wrapper: introduce startup_info struct

From: Nguyễn Thái Ngọc Duy <[email protected]>

The startup_info struct will collect information managed by the git setup code, such as the prefix for relative paths passed on the command line (i.e., path to the starting cwd from the toplevel of the work tree) and whether a git repository has been found.

In other words, startup_info is intended to be a collection of global variables with results that were previously returned from setup functions. This state is global anyway (since the cwd is), even if it is not currently tracked that way. Letting these values persist means there is more flexibility in deciding when to run setup.

For now, the struct is empty.

---
Changes from last round:
- commit message
- remove unnecessary memset (since statics are already automatically zero-initialized)

 cache.h       |    5 +++++
 environment.c |    1 +
 git.c         |    3 +++
 3 files changed, 9 insertions(+), 0 deletions(-)
cache.h +5 -0
index c9fa3df..0822c3b 100644
@@ -1097,6 +1097,11 @@ void overlay_tree_on_cache(const char *tree_name, const char *prefix);
 char *alias_lookup(const char *alias);
 int split_cmdline(char *cmdline, const char ***argv);
 
+/* git.c */
+struct startup_info {
+};
+extern struct startup_info *startup_info;
+
 /* builtin/merge.c */
 int checkout_fast_forward(const unsigned char *from, const unsigned char *to);
 
environment.c +1 -0
index 83d38d3..eeb2687 100644
@@ -53,6 +53,7 @@ enum object_creation_mode object_creation_mode = OBJECT_CREATION_MODE;
 char *notes_ref_name;
 int grafts_replace_parents = 1;
 int core_apply_sparse_checkout;
+struct startup_info *startup_info;
 
 /* Parallel index stat data preload? */
 int core_preload_index = 0;
git.c +3 -0
index 99517dd..8a8d2b2 100644
@@ -14,6 +14,7 @@ const char git_usage_string[] =
 const char git_more_info_string[] =
 	"See 'git help COMMAND' for more information on a specific command.";
 
+static struct startup_info git_startup_info;
 static int use_pager = -1;
 struct pager_config {
 	const char *cmd;
@@ -489,6 +490,8 @@ int main(int argc, const char **argv)
 {
 	const char *cmd;
 
+	startup_info = &git_startup_info;
+
 	cmd = git_extract_argv0_path(argv[0]);
 	if (!cmd)
 		cmd = "git-help";
Signature · 1 line
1.7.2.1.544.ga752d.dirty
#17 lore raw

PATCH 2/12 setup: remember whether repository was found

In reply to Jonathan Nieder

From: Nguyễn Thái Ngọc Duy <[email protected]>

As v1.7.2~16^2 (git --paginate: paginate external commands again, 2010-07-14) explains, builtins (like git config) that do not use RUN_SETUP are not finding GIT_DIR set correctly when it is time to launch the pager from run_builtin(). If they were to search for a repository sooner, then the outcome of such early repository accesses would be more predictable and reliable.

The cmd_*() functions learn whether a repository was found through the *nongit_ok return value from setup_git_directory_gently(). If run_builtin() is to take care of the repository search itself, that datum needs to be retrievable from somewhere else. Use the startup_info struct for this.

As a bonus, this information becomes available to functions such as git_config() which might want to avoid trying to access a repository when none is present.

---
Tweaked the log message a bit.

 cache.h |    1 +
 setup.c |   12 +++++++++++-
 2 files changed, 12 insertions(+), 1 deletions(-)
cache.h +1 -0
index 0822c3b..f9e82a8 100644
@@ -1099,6 +1099,7 @@ int split_cmdline(char *cmdline, const char ***argv);
 
 /* git.c */
 struct startup_info {
+	int have_repository;
 };
 extern struct startup_info *startup_info;
 
setup.c +11 -1
index 7e04602..1e32fb7 100644
@@ -315,7 +315,7 @@ const char *read_gitfile_gently(const char *path)
  * We cannot decide in this function whether we are in the work tree or
  * not, since the config can only be read _after_ this function was called.
  */
-const char *setup_git_directory_gently(int *nongit_ok)
+static const char *setup_git_directory_gently_1(int *nongit_ok)
 {
 	const char *work_tree_env = getenv(GIT_WORK_TREE_ENVIRONMENT);
 	const char *env_ceiling_dirs = getenv(CEILING_DIRECTORIES_ENVIRONMENT);
@@ -470,6 +470,16 @@ const char *setup_git_directory_gently(int *nongit_ok)
 	return cwd + offset;
 }
 
+const char *setup_git_directory_gently(int *nongit_ok)
+{
+	const char *prefix;
+
+	prefix = setup_git_directory_gently_1(nongit_ok);
+	if (startup_info)
+		startup_info->have_repository = !nongit_ok || !*nongit_ok;
+	return prefix;
+}
+
 int git_config_perm(const char *var, const char *value)
 {
 	int i;
Signature · 1 line
1.7.2.1.544.ga752d.dirty
#18 lore raw

PATCH 3/12 git wrapper: allow setup_git_directory_gently() be called earlier

In reply to Jonathan Nieder

From: Nguyễn Thái Ngọc Duy <[email protected]>

In the spirit of v1.4.2-rc3~34^2^2 (Call setup_git_directory() much earlier, 2006-07-28), let run_builtin() take care of searching for a repository for built-ins that want to make use of one if present.

So now you can mark your command with RUN_SETUP_GENTLY and use

	nongit = !startup_info->have_repository;

in place of

	prefix = setup_git_directory_gently(&nongit);

and everything will be the same, except the repository is discovered a little sooner.

As v1.7.2~16^2 (2010-07-14) explains, this should allow more commands to robustly use features like "git --paginate" that look at local configuration before the command is actually run.

This patch sets up the infrastructure. Later patches will teach particular commands to use it.

---
- reordered setup option bits (RUN_SETUP_GENTLY lives next to RUN_SETUP now)

- squashed in "builtin: check pager.<cmd> configuration if RUN_SETUP_GENTLY is used"

 git.c |   13 +++++++++----
 1 files changed, 9 insertions(+), 4 deletions(-)
git.c +9 -4
index 8a8d2b2..c46c996 100644
@@ -230,13 +230,14 @@ static int handle_alias(int *argcp, const char ***argv)
 
 const char git_version_string[] = GIT_VERSION;
 
-#define RUN_SETUP	(1<<0)
-#define USE_PAGER	(1<<1)
+#define RUN_SETUP		(1<<0)
+#define RUN_SETUP_GENTLY	(1<<1)
+#define USE_PAGER		(1<<2)
 /*
  * require working tree to be present -- anything uses this needs
  * RUN_SETUP for reading from the configuration file.
  */
-#define NEED_WORK_TREE	(1<<2)
+#define NEED_WORK_TREE		(1<<3)
 
 struct cmd_struct {
 	const char *cmd;
@@ -255,8 +256,12 @@ static int run_builtin(struct cmd_struct *p, int argc, const char **argv)
 	if (!help) {
 		if (p->option & RUN_SETUP)
 			prefix = setup_git_directory();
+		if (p->option & RUN_SETUP_GENTLY) {
+			int nongit_ok;
+			prefix = setup_git_directory_gently(&nongit_ok);
+		}
 
-		if (use_pager == -1 && p->option & RUN_SETUP)
+		if (use_pager == -1 && p->option & (RUN_SETUP | RUN_SETUP_GENTLY))
 			use_pager = check_pager_config(p->cmd);
 		if (use_pager == -1 && p->option & USE_PAGER)
 			use_pager = 1;
Signature · 1 line
1.7.2.1.544.ga752d.dirty
#19 lore raw

PATCH 4/12 shortlog: run setup_git_directory_gently() sooner

In reply to Jonathan Nieder

From: Nguyễn Thái Ngọc Duy <[email protected]>

shortlog already runs a repository search unconditionally; running such a search earlier is not very risky.

Without this change, the “[pager] shortlog” configuration is not respected at all: “git shortlog” unconditionally paginates.

The tests are a bit slow. Running the full battery like this for all built-in commands would be counterproductive; the intent is rather to test shortlog as a representative example command using ..._gently().

---
Clarified commit message; simplified diff by keeping the nongit variable; added tests.

 builtin/shortlog.c |    3 +--
 git.c              |    2 +-
 t/t7006-pager.sh   |    9 +++++++++
 3 files changed, 11 insertions(+), 3 deletions(-)
builtin/shortlog.c +1 -2
index 5089502..36cb24b 100644
@@ -249,7 +249,7 @@ int cmd_shortlog(int argc, const char **argv, const char *prefix)
 {
 	static struct shortlog log;
 	static struct rev_info rev;
-	int nongit;
+	int nongit = !startup_info->have_repository;
 
 	static const struct option options[] = {
 		OPT_BOOLEAN('n', "numbered", &log.sort_by_number,
@@ -265,7 +265,6 @@ int cmd_shortlog(int argc, const char **argv, const char *prefix)
 
 	struct parse_opt_ctx_t ctx;
 
-	prefix = setup_git_directory_gently(&nongit);
 	git_config(git_default_config, NULL);
 	shortlog_init(&log);
 	init_revisions(&rev, prefix);
git.c +1 -1
index c46c996..b821058 100644
@@ -384,7 +384,7 @@ static void handle_internal_command(int argc, const char **argv)
 		{ "revert", cmd_revert, RUN_SETUP | NEED_WORK_TREE },
 		{ "rm", cmd_rm, RUN_SETUP },
 		{ "send-pack", cmd_send_pack, RUN_SETUP },
-		{ "shortlog", cmd_shortlog, USE_PAGER },
+		{ "shortlog", cmd_shortlog, RUN_SETUP_GENTLY | USE_PAGER },
 		{ "show-branch", cmd_show_branch, RUN_SETUP },
 		{ "show", cmd_show, RUN_SETUP | USE_PAGER },
 		{ "status", cmd_status, RUN_SETUP | NEED_WORK_TREE },
t/t7006-pager.sh +9 -0
index 9215c2f..084dfdb 100755
@@ -377,4 +377,13 @@ test_GIT_PAGER_overrides  expect_success test_must_fail 'git -p'
 
 test_doesnt_paginate      expect_failure test_must_fail 'git -p nonsense'
 
+test_pager_choices                       'git shortlog'
+test_expect_success 'setup: configure shortlog not to paginate' '
+	git config pager.shortlog false
+'
+test_doesnt_paginate      expect_success 'git shortlog'
+test_no_local_config_subdir expect_success 'git shortlog'
+test_default_pager        expect_success 'git -p shortlog'
+test_core_pager_subdir    expect_success 'git -p shortlog'
+
 test_done
Signature · 1 line
1.7.2.1.544.ga752d.dirty
#20 lore raw

PATCH 5/12 grep: run setup_git_directory_gently() sooner

In reply to Jonathan Nieder

From: Nguyễn Thái Ngọc Duy <[email protected]>

git grep already runs a repository search unconditionally, even when the --no-index option is supplied; running such a search earlier is not very risky.

Just like with shortlog, without this change, the “[pager] grep” configuration is not respected at all.

---
Aside from rewriting the commit message and adding tests, this drops the interesting

- /* die the same way as if we did it at the beginning */
- setup_git_directory();
+ die("No git repository found");

hunk. That change might be a good idea but it does not fit the theme of this chapter.

 builtin/grep.c   |    6 ++----
 git.c            |    2 +-
 t/t7006-pager.sh |   13 +++++++++++++
 3 files changed, 16 insertions(+), 5 deletions(-)
builtin/grep.c +2 -4
index d0a73da..cd44926 100644
@@ -791,7 +791,7 @@ int cmd_grep(int argc, const char **argv, const char *prefix)
 	const char **paths = NULL;
 	int i;
 	int dummy;
-	int nongit = 0, use_index = 1;
+	int use_index = 1;
 	struct option options[] = {
 		OPT_BOOLEAN(0, "cached", &cached,
 			"search in index instead of in the work tree"),
@@ -879,8 +879,6 @@ int cmd_grep(int argc, const char **argv, const char *prefix)
 		OPT_END()
 	};
 
-	prefix = setup_git_directory_gently(&nongit);
-
 	/*
 	 * 'git grep -h', unlike 'git grep -h <pattern>', is a request
 	 * to show usage information and exit.
@@ -925,7 +923,7 @@ int cmd_grep(int argc, const char **argv, const char *prefix)
 			     PARSE_OPT_STOP_AT_NON_OPTION |
 			     PARSE_OPT_NO_INTERNAL_HELP);
 
-	if (use_index && nongit)
+	if (use_index && !startup_info->have_repository)
 		/* die the same way as if we did it at the beginning */
 		setup_git_directory();
 
git.c +1 -1
index b821058..0240179 100644
@@ -336,7 +336,7 @@ static void handle_internal_command(int argc, const char **argv)
 		{ "fsck-objects", cmd_fsck, RUN_SETUP },
 		{ "gc", cmd_gc, RUN_SETUP },
 		{ "get-tar-commit-id", cmd_get_tar_commit_id },
-		{ "grep", cmd_grep, USE_PAGER },
+		{ "grep", cmd_grep, RUN_SETUP_GENTLY | USE_PAGER },
 		{ "hash-object", cmd_hash_object },
 		{ "help", cmd_help },
 		{ "index-pack", cmd_index_pack },
t/t7006-pager.sh +13 -0
index 084dfdb..fd7f77b 100755
@@ -105,6 +105,19 @@ test_expect_success TTY 'no pager with --no-pager' '
 	! test -e paginated.out
 '
 
+test_expect_success TTY 'configuration can disable pager' '
+	rm -f paginated.out &&
+	test_might_fail git config --unset pager.grep &&
+	test_terminal git grep initial &&
+	test -e paginated.out &&
+
+	rm -f paginated.out &&
+	git config pager.grep false &&
+	test_when_finished "git config --unset pager.grep" &&
+	test_terminal git grep initial &&
+	! test -e paginated.out
+'
+
 # A colored commit log will begin with an appropriate ANSI escape
 # for the first color; the text "commit" comes later.
 colorful() {
Signature · 1 line
1.7.2.1.544.ga752d.dirty
#21 lore raw

PATCH 6/12 apply: run setup_git_directory_gently() sooner

In reply to Jonathan Nieder

From: Nguyễn Thái Ngọc Duy <[email protected]>

As v1.7.2~16^2 (2010-07-14) explains, without this change, “git --paginate apply” can ignore the repository-local “[core] pager” configuration.

---
New explanation and test; simplified patch by keeping the is_not_gitdir local.

 builtin/apply.c  |    3 +--
 git.c            |    2 +-
 t/t7006-pager.sh |    3 +++
 3 files changed, 5 insertions(+), 3 deletions(-)
builtin/apply.c +1 -2
index 562e534..08f3586 100644
@@ -3611,7 +3611,7 @@ int cmd_apply(int argc, const char **argv, const char *unused_prefix)
 {
 	int i;
 	int errs = 0;
-	int is_not_gitdir;
+	int is_not_gitdir = !startup_info->have_repository;
 	int binary;
 	int force_apply = 0;
 
@@ -3684,7 +3684,6 @@ int cmd_apply(int argc, const char **argv, const char *unused_prefix)
 		OPT_END()
 	};
 
-	prefix = setup_git_directory_gently(&is_not_gitdir);
 	prefix_length = prefix ? strlen(prefix) : 0;
 	git_config(git_apply_config, NULL);
 	if (apply_default_whitespace)
git.c +1 -1
index 0240179..82b390b 100644
@@ -301,7 +301,7 @@ static void handle_internal_command(int argc, const char **argv)
 		{ "add", cmd_add, RUN_SETUP | NEED_WORK_TREE },
 		{ "stage", cmd_add, RUN_SETUP | NEED_WORK_TREE },
 		{ "annotate", cmd_annotate, RUN_SETUP },
-		{ "apply", cmd_apply },
+		{ "apply", cmd_apply, RUN_SETUP_GENTLY },
 		{ "archive", cmd_archive },
 		{ "bisect--helper", cmd_bisect__helper, RUN_SETUP | NEED_WORK_TREE },
 		{ "blame", cmd_blame, RUN_SETUP },
t/t7006-pager.sh +3 -0
index fd7f77b..96fbb0f 100755
@@ -399,4 +399,7 @@ test_no_local_config_subdir expect_success 'git shortlog'
 test_default_pager        expect_success 'git -p shortlog'
 test_core_pager_subdir    expect_success 'git -p shortlog'
 
+test_core_pager_subdir    expect_success test_must_fail \
+                                         'git -p apply </dev/null'
+
 test_done
Signature · 1 line
1.7.2.1.544.ga752d.dirty
#22 lore raw

PATCH v2 6/12 apply: run setup_git_directory_gently() sooner

As v1.7.2~16^2 (2010-07-14) explains, without this change, “git --paginate apply” can ignore the repository-local “[core] pager” configuration.

---
This is a replacement for 4ce097c (apply: run setup_git_directory_gently() sooner) in pu, branch jn/paginate-fix, which fixes prefix loss in 4ce097c.

All tests seem to run ok for me.

 builtin/apply.c  |    6 +++---
 git.c            |    2 +-
 t/t7006-pager.sh |    3 +++
 3 files changed, 7 insertions(+), 4 deletions(-)
builtin/apply.c +3 -3
index 12ef9ea..f005ba1 100644
@@ -3607,11 +3607,11 @@ static int option_parse_directory(const struct option *opt,
 	return 0;
 }
 
-int cmd_apply(int argc, const char **argv, const char *unused_prefix)
+int cmd_apply(int argc, const char **argv, const char *prefix_)
 {
 	int i;
 	int errs = 0;
-	int is_not_gitdir;
+	int is_not_gitdir = !startup_info->have_repository;
 	int binary;
 	int force_apply = 0;
 
@@ -3684,7 +3684,7 @@ int cmd_apply(int argc, const char **argv, const char *unused_prefix)
 		OPT_END()
 	};
 
-	prefix = setup_git_directory_gently(&is_not_gitdir);
+	prefix = prefix_;
 	prefix_length = prefix ? strlen(prefix) : 0;
 	git_config(git_apply_config, NULL);
 	if (apply_default_whitespace)
git.c +1 -1
index 5a47bfb..38dbe70 100644
@@ -301,7 +301,7 @@ static void handle_internal_command(int argc, const char **argv)
 		{ "add", cmd_add, RUN_SETUP | NEED_WORK_TREE },
 		{ "stage", cmd_add, RUN_SETUP | NEED_WORK_TREE },
 		{ "annotate", cmd_annotate, RUN_SETUP },
-		{ "apply", cmd_apply },
+		{ "apply", cmd_apply, RUN_SETUP_GENTLY },
 		{ "archive", cmd_archive },
 		{ "bisect--helper", cmd_bisect__helper, RUN_SETUP | NEED_WORK_TREE },
 		{ "blame", cmd_blame, RUN_SETUP },
t/t7006-pager.sh +3 -0
index bb95335..6c86d70 100755
@@ -391,4 +391,7 @@ test_no_local_config_subdir expect_success 'git shortlog'
 test_default_pager        expect_success 'git -p shortlog'
 test_core_pager_subdir    expect_success 'git -p shortlog'
 
+test_core_pager_subdir    expect_success test_must_fail \
+					 'git -p apply </dev/null'
+
 test_done
Signature · 1 line
1.7.1.rc1.69.g24c2f7
#23 lore raw

Re: [PATCH 06/12 v2] apply: run setup_git_directory_gently() sooner

Quoted 4 lines · Nguyễn Thái Ngọc Duy

As v1.7.2~16^2 (2010-07-14) explains, without this change, “git --paginate apply” can ignore the repository-local “[core] pager” configuration.

Quoted 8 lines

---
This is a replacement for 4ce097c (apply: run setup_git_directory_gently() sooner) in pu, branch jn/paginate-fix, which fixes prefix loss in 4ce097c.

Thanks for a quick fix.

#24 lore raw

Re: [PATCH 06/12] apply: run setup_git_directory_gently() sooner

In reply to Jonathan Nieder

2010/8/6 Jonathan Nieder <[email protected]>:

Quoted 5 lines

From: Nguyễn Thái Ngọc Duy <[email protected]>

As v1.7.2~16^2 (2010-07-14) explains, without this change, “git --paginate apply” can ignore the repository-local “[core] pager” configuration.

Applying this patch broke the following tests:

    ./t4119-apply-config.sh ./t4111-apply-subdir.sh

./t4131-apply-fake-ancestor.sh

I didn't look into why, they're breaking in pu now at 6ea3604. I didn't look into why, see the smoke report at
http://smoke.git.nix.is/app/projects/report_details/33

#25 lore raw

Re: [PATCH 06/12] apply: run setup_git_directory_gently() sooner

Quoted 7 lines · Ævar Arnfjörð Bjarmason

2010/8/6 Jonathan Nieder <[email protected]>:

Quoted 5 lines

From: Nguyễn Thái Ngọc Duy <[email protected]>

As v1.7.2~16^2 (2010-07-14) explains, without this change, “git --paginate apply” can ignore the repository-local “[core] pager” configuration.

Quoted 8 lines

Applying this patch broke the following tests:

   ./t4119-apply-config.sh ./t4111-apply-subdir.sh ./t4131-apply-fake-ancestor.sh

I didn't look into why, they're breaking in pu now at 6ea3604. I didn't look into why, see the smoke report at
http://smoke.git.nix.is/app/projects/report_details/33

The patch loses prefix and git-apply couldn't find files on disk. This patch may fix it. I'm running tests now.

builtin/apply.c +0 -0
index 6c4b747..29f4194 100644
@@ -3779,7 +3779,7 @@ static int option_parse_directory(const struct

option *opt,
return 0;
}

-int cmd_apply(int argc, const char **argv, const char *unused_prefix) +int cmd_apply(int argc, const char **argv, const char *prefix_)
{
int i;
int errs = 0;
@@ -3856,6 +3856,7 @@ int cmd_apply(int argc, const char **argv, const char *unused_prefix)
OPT_END()
};

+ prefix = prefix_;
prefix_length = prefix ? strlen(prefix) : 0; git_config(git_apply_config, NULL); if (apply_default_whitespace)

Signature · 1 line
Duy
#26 lore raw

Re: [PATCH 06/12] apply: run setup_git_directory_gently() sooner

Quoted 2 lines · Nguyen Thai Ngoc Duy

The patch loses prefix and git-apply couldn't find files on disk. This patch may fix it. I'm running tests now.

Oh, good catch. Thanks!

Quoted 15 lines
builtin/apply.c +1 -1
index 6c4b747..29f4194 100644
@@ -3779,7 +3779,7 @@ static int option_parse_directory(const struct
 option *opt,
 	return 0;
 }
 
-int cmd_apply(int argc, const char **argv, const char *unused_prefix)
+int cmd_apply(int argc, const char **argv, const char *prefix_)
 {
 	int i;
builtin/apply.c +0 -0
 	int errs = 0;

builtin/apply.c int cmd_apply(int argc, const char **argv, const · line 3860

Quoted 5 lines
builtin/apply.c +1 -0
@@ -3856,6 +3856,7 @@ int cmd_apply(int argc, const char **argv, const
 char *unused_prefix)
 		OPT_END()
 	};
 
+	prefix = prefix_;

This is why we shouldn’t be using globals so cavalierly. :)

#27 lore raw

PATCH 7/12 bundle: run setup_git_directory_gently() sooner

In reply to Jonathan Nieder

From: Nguyễn Thái Ngọc Duy <[email protected]>

Without this change, “git -p bundle” does not always respect the repository-local “[core] pager” setting.

It is hard to notice because subcommands other than “git bundle unbundle” do not produce much output.

---
New tests and explanation.

The first new test perhaps deserves some explanation. Its only purpose is to let us use the

	test -e paginated.out || test -e subdir/paginated.out

hedge in other tests with good conscience.

 builtin/bundle.c |    6 ++----
 git.c            |    2 +-
 t/t7006-pager.sh |   33 +++++++++++++++++++++++++++++++++
 3 files changed, 36 insertions(+), 5 deletions(-)
builtin/bundle.c +2 -4
index 2006cc5..80649ba 100644
@@ -18,7 +18,6 @@ static const char builtin_bundle_usage[] =
 int cmd_bundle(int argc, const char **argv, const char *prefix)
 {
 	struct bundle_header header;
-	int nongit;
 	const char *cmd, *bundle_file;
 	int bundle_fd = -1;
 	char buffer[PATH_MAX];
@@ -31,7 +30,6 @@ int cmd_bundle(int argc, const char **argv, const char *prefix)
 	argc -= 2;
 	argv += 2;
 
-	prefix = setup_git_directory_gently(&nongit);
 	if (prefix && bundle_file[0] != '/') {
 		snprintf(buffer, sizeof(buffer), "%s/%s", prefix, bundle_file);
 		bundle_file = buffer;
@@ -54,11 +52,11 @@ int cmd_bundle(int argc, const char **argv, const char *prefix)
 		return !!list_bundle_refs(&header, argc, argv);
 	}
 	if (!strcmp(cmd, "create")) {
-		if (nongit)
+		if (!startup_info->have_repository)
 			die("Need a repository to create a bundle.");
 		return !!create_bundle(&header, bundle_file, argc, argv);
 	} else if (!strcmp(cmd, "unbundle")) {
-		if (nongit)
+		if (!startup_info->have_repository)
 			die("Need a repository to unbundle.");
 		return !!unbundle(&header, bundle_fd) ||
 			list_bundle_refs(&header, argc, argv);
git.c +1 -1
index 82b390b..94bb1e3 100644
@@ -306,7 +306,7 @@ static void handle_internal_command(int argc, const char **argv)
 		{ "bisect--helper", cmd_bisect__helper, RUN_SETUP | NEED_WORK_TREE },
 		{ "blame", cmd_blame, RUN_SETUP },
 		{ "branch", cmd_branch, RUN_SETUP },
-		{ "bundle", cmd_bundle },
+		{ "bundle", cmd_bundle, RUN_SETUP_GENTLY },
 		{ "cat-file", cmd_cat_file, RUN_SETUP },
 		{ "checkout", cmd_checkout, RUN_SETUP | NEED_WORK_TREE },
 		{ "checkout-index", cmd_checkout_index,
t/t7006-pager.sh +33 -0
index 96fbb0f..6680668 100755
@@ -57,6 +57,21 @@ test_expect_success TTY 'some commands use a pager' '
 	test -e paginated.out
 '
 
+test_expect_failure TTY 'pager runs from subdir' '
+	echo subdir/paginated.out >expected &&
+	mkdir -p subdir &&
+	rm -f paginated.out subdir/paginated.out &&
+	(
+		cd subdir &&
+		test_terminal git log
+	) &&
+	{
+		ls paginated.out subdir/paginated.out ||
+		:
+	} >actual &&
+	test_cmp expected actual
+'
+
 test_expect_success TTY 'some commands do not use a pager' '
 	rm -f paginated.out ||
 	cleanup_fail &&
@@ -118,6 +133,24 @@ test_expect_success TTY 'configuration can disable pager' '
 	! test -e paginated.out
 '
 
+test_expect_success 'configuration can enable pager (from subdir)' '
+	rm -f paginated.out &&
+	mkdir -p subdir &&
+	git config pager.bundle true &&
+	test_when_finished "git config --unset pager.bundle" &&
+
+	git bundle create test.bundle --all &&
+	rm -f paginated.out subdir/paginated.out &&
+	(
+		cd subdir &&
+		test_terminal git bundle unbundle ../test.bundle
+	) &&
+	{
+		test -e paginated.out ||
+		test -e subdir/paginated.out
+	}
+'
+
 # A colored commit log will begin with an appropriate ANSI escape
 # for the first color; the text "commit" comes later.
 colorful() {
Signature · 1 line
1.7.2.1.544.ga752d.dirty
#28 lore raw

Re: [PATCH 07/12] bundle: run setup_git_directory_gently() sooner

Quoted 4 lines · Jonathan Nieder
 builtin/bundle.c |    6 ++----
 git.c            |    2 +-
 t/t7006-pager.sh |   33 +++++++++++++++++++++++++++++++++
 3 files changed, 36 insertions(+), 5 deletions(-)

[...]

Quoted 11 lines
t/t7006-pager.sh +11 -0
+test_expect_success 'configuration can enable pager (from subdir)' '
+	rm -f paginated.out &&
+	mkdir -p subdir &&
+	git config pager.bundle true &&
+	test_when_finished "git config --unset pager.bundle" &&
+
+	git bundle create test.bundle --all &&
+	rm -f paginated.out subdir/paginated.out &&
+	(
+		cd subdir &&
+		test_terminal git bundle unbundle ../test.bundle

t/t7006-pager.sh test_expect_success TTY 'configuration can disable pager' ' · line 156

Quoted 10 lines
t/t7006-pager.sh +7 -0
+	) &&
+	{
+		test -e paginated.out ||
+		test -e subdir/paginated.out
+	}
+'
+
 # A colored commit log will begin with an appropriate ANSI escape
 # for the first color; the text "commit" comes later.
 colorful() {

On my valgrind test setup, this never worked (i.e., fails and bisects to this commit).

Oddly, I am seeing this error message from the second test (second in t7006, not in this patch):

expecting success:

          rm -f stdout_is_tty ||
          cleanup_fail &&

          if test -t 1
          then
                  >stdout_is_tty
          elif
                  test_have_prereq PERL &&
                  "$PERL_PATH" "$TEST_DIRECTORY"/t7006/test-terminal.perl \
                          sh -c "test -t 1"
          then
                  >test_terminal_works
          fi

Can't locate IO/Pty.pm in @INC (@INC contains: <snip>) at /local/home/trast/git/t/t7006/test-terminal.perl line 4. BEGIN failed--compilation aborted at /local/home/trast/git/t/t7006/test-terminal.perl line 4. ok 2 - set up terminal for tests

Which raises a few questions: Why was this never an issue before? Am I supposed to have IO::Pty with a perl install (it's a perl 5.8.8) or does the test need a prerequisite other than HAVE_PERL?4

Signature · 2 lines
Thomas Rast
trast@{inf,student}.ethz.ch
#29 lore raw

Re: [PATCH 07/12] bundle: run setup_git_directory_gently() sooner

Quoted 3 lines · Thomas Rast
Quoted 1 line · Jonathan Nieder

+test_expect_success 'configuration can enable pager (from subdir)' '

[...]

Quoted 1 line
Quoted 1 line
t/t7006-pager.sh +1 -0
+		test_terminal git bundle unbundle ../test.bundle

[...]

Quoted 2 lines

On my valgrind test setup, this never worked (i.e., fails and bisects to this commit).

Agh, sloppy me... Fixes below.

Quoted 2 lines

Oddly, I am seeing this error message from the second test (second in t7006, not in this patch):

[...]

Quoted 5 lines

Can't locate IO/Pty.pm in @INC (@INC contains: <snip>) at /local/home/trast/git/t/t7006/test-terminal.perl line 4. BEGIN failed--compilation aborted at /local/home/trast/git/t/t7006/test-terminal.perl line 4. ok 2 - set up terminal for tests

Which raises a few questions: Why was this never an issue before?

That’s expected behavior, marked with “ok”. :) That test is checking if IO::Pty is available and works; if not, the relevant tests should be skipped.

Quoted 1 line

does the test need a prerequisite other than HAVE_PERL?4

Exactly.

Jonathan Nieder (2):
t7006 (pager): add missing TTY prerequisite
t7006 (pager): add missing TTY prerequisite

 t/t7006-pager.sh |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
#30 lore raw

PATCH 2/2 t7006 (pager): add missing TTY prerequisite

The config pagination test should not run if there is not a tty available to force pagination on.

---
This one fixes for “config: run setup_git_directory_gently() sooner”.

Thanks again for the report. I’ll think more about how to make problems like this easily reproducible (a test_mentions_prereq() function?).

 t/t7006-pager.sh |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
t/t7006-pager.sh +1 -1
index fcd846a..fb744e3 100755
@@ -134,7 +134,7 @@ test_expect_success TTY 'configuration can disable pager' '
 	! test -e paginated.out
 '
 
-test_expect_success 'git config uses a pager if configured to' '
+test_expect_success TTY 'git config uses a pager if configured to' '
 	rm -f paginated.out &&
 	git config pager.config true &&
 	test_when_finished "git config --unset pager.config" &&
Signature · 1 line
1.7.2.1.544.ga752d.dirty
#31 lore raw

Re: [PATCH 2/2] t7006 (pager): add missing TTY prerequisite

Quoted 8 lines · Jonathan Nieder

The config pagination test should not run if there is not a tty available to force pagination on.

---
This one fixes for “config: run setup_git_directory_gently() sooner”.

Somehow we didn't seem to have got 1/2, but here is what I'll queue.

By the way, could you stop using those "pretty quotes" and stick to the good old "ASCII 0x22" double-quotes? I find them ugly in the git log output.

-- >8 --
From: Jonathan Nieder <[email protected]>
Date: Mon, 16 Aug 2010 03:08:34 -0500
Subject: [PATCH] t7006 (pager): add missing TTY prerequisites

The "git bundle unbundle" and "git config" pagination tests are not supposed to run when stdout is not a terminal and IO::Pty not available to make one on the fly.

---
 t/t7006-pager.sh |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
t/t7006-pager.sh +2 -2
index 017565f..fb744e3 100755
@@ -134,7 +134,7 @@ test_expect_success TTY 'configuration can disable pager' '
 	! test -e paginated.out
 '
 
-test_expect_success 'git config uses a pager if configured to' '
+test_expect_success TTY 'git config uses a pager if configured to' '
 	rm -f paginated.out &&
 	git config pager.config true &&
 	test_when_finished "git config --unset pager.config" &&
@@ -142,7 +142,7 @@ test_expect_success 'git config uses a pager if configured to' '
 	test -e paginated.out
 '
 
-test_expect_success 'configuration can enable pager (from subdir)' '
+test_expect_success TTY 'configuration can enable pager (from subdir)' '
 	rm -f paginated.out &&
 	mkdir -p subdir &&
 	git config pager.bundle true &&
Signature · 1 line
1.7.2.1.224.g2f41ea
#32 lore raw

PATCH 8/12 config: run setup_git_directory_gently() sooner

In reply to Jonathan Nieder

From: Nguyễn Thái Ngọc Duy <[email protected]>

For the pager choice (and the choice to paginate) to reflect the current repository configuration, the repository needs to be located first.

---
Relative to the previous round, this simplifies the patch by retaining the nongit var; squashes in a test; and adds some words of explanation to the log message.

 builtin/config.c |    5 ++---
 git.c            |    4 ++--
 t/t7006-pager.sh |    8 ++++++++
 3 files changed, 12 insertions(+), 5 deletions(-)
builtin/config.c +2 -3
index f3d1660..3f8ef91 100644
@@ -330,11 +330,10 @@ static int get_colorbool(int print)
 		return get_colorbool_found ? 0 : 1;
 }
 
-int cmd_config(int argc, const char **argv, const char *unused_prefix)
+int cmd_config(int argc, const char **argv, const char *prefix)
 {
-	int nongit;
+	int nongit = !startup_info->have_repository;
 	char *value;
-	const char *prefix = setup_git_directory_gently(&nongit);
 
 	config_exclusive_filename = getenv(CONFIG_ENVIRONMENT);
 
git.c +2 -2
index 94bb1e3..f57003d 100644
@@ -319,7 +319,7 @@ static void handle_internal_command(int argc, const char **argv)
 		{ "clean", cmd_clean, RUN_SETUP | NEED_WORK_TREE },
 		{ "commit", cmd_commit, RUN_SETUP | NEED_WORK_TREE },
 		{ "commit-tree", cmd_commit_tree, RUN_SETUP },
-		{ "config", cmd_config },
+		{ "config", cmd_config, RUN_SETUP_GENTLY },
 		{ "count-objects", cmd_count_objects, RUN_SETUP },
 		{ "describe", cmd_describe, RUN_SETUP },
 		{ "diff", cmd_diff },
@@ -376,7 +376,7 @@ static void handle_internal_command(int argc, const char **argv)
 		{ "reflog", cmd_reflog, RUN_SETUP },
 		{ "remote", cmd_remote, RUN_SETUP },
 		{ "replace", cmd_replace, RUN_SETUP },
-		{ "repo-config", cmd_config },
+		{ "repo-config", cmd_config, RUN_SETUP_GENTLY },
 		{ "rerere", cmd_rerere, RUN_SETUP },
 		{ "reset", cmd_reset, RUN_SETUP },
 		{ "rev-list", cmd_rev_list, RUN_SETUP },
t/t7006-pager.sh +8 -0
index 6680668..dbbe8d9 100755
@@ -133,6 +133,14 @@ test_expect_success TTY 'configuration can disable pager' '
 	! test -e paginated.out
 '
 
+test_expect_success 'git config uses a pager if configured to' '
+	rm -f paginated.out &&
+	git config pager.config true &&
+	test_when_finished "git config --unset pager.config" &&
+	test_terminal git config --list &&
+	test -e paginated.out
+'
+
 test_expect_success 'configuration can enable pager (from subdir)' '
 	rm -f paginated.out &&
 	mkdir -p subdir &&
Signature · 1 line
1.7.2.1.544.ga752d.dirty
#33 lore raw

PATCH 9/12 index-pack: run setup_git_directory_gently() sooner

In reply to Jonathan Nieder

From: Nguyễn Thái Ngọc Duy <[email protected]>

index-pack already runs a repository search unconditionally; running such a search earlier is not risky and ensures GIT_DIR will be set correctly if the configuration needs to be accessed from run_builtin().

---
Requires the jn/maint-setup-fix topic to apply. See the cover letter "[PATCH jn/paginate-fix 0/12] Re: ..." for more on that.

Just like last round, except for the new log message.

 builtin/index-pack.c |    2 --
 git.c                |    2 +-
 2 files changed, 1 insertions(+), 3 deletions(-)
builtin/index-pack.c +0 -2
index 89a1f12..e852890 100644
@@ -880,12 +880,10 @@ int cmd_index_pack(int argc, const char **argv, const char *prefix)
 	char *index_name_buf = NULL, *keep_name_buf = NULL;
 	struct pack_idx_entry **idx_objects;
 	unsigned char pack_sha1[20];
-	int nongit;
 
 	if (argc == 2 && !strcmp(argv[1], "-h"))
 		usage(index_pack_usage);
 
-	prefix = setup_git_directory_gently(&nongit);
 	git_config(git_index_pack_config, NULL);
 	if (prefix && chdir(prefix))
 		die("Cannot come back to cwd");
git.c +1 -1
index e89c010..92ec5d3 100644
@@ -339,7 +339,7 @@ static void handle_internal_command(int argc, const char **argv)
 		{ "grep", cmd_grep, RUN_SETUP_GENTLY },
 		{ "hash-object", cmd_hash_object },
 		{ "help", cmd_help },
-		{ "index-pack", cmd_index_pack },
+		{ "index-pack", cmd_index_pack, RUN_SETUP_GENTLY },
 		{ "init", cmd_init_db },
 		{ "init-db", cmd_init_db },
 		{ "log", cmd_log, RUN_SETUP | USE_PAGER },
Signature · 1 line
1.7.2.1.544.ga752d.dirty
#34 lore raw

PATCH 10/12 ls-remote: run setup_git_directory_gently() sooner

In reply to Jonathan Nieder

From: Nguyễn Thái Ngọc Duy <[email protected]>

ls-remote already runs a repository search unconditionally to learn about remote nicknames and "[url] insteadof" shortcuts. Run that search a little sooner, and now one can try

	[pager]
		ls-remote

to automatically paginate ls-remote output, or use repository-local

	[core]
		pager = whatever

with "git --paginate ls-remote <url>".

---
As before, aside from the commit message.

 builtin/ls-remote.c |    3 ---
 git.c               |    4 ++--
 2 files changed, 2 insertions(+), 5 deletions(-)
builtin/ls-remote.c +0 -3
index 34480cf..97eed40 100644
@@ -32,7 +32,6 @@ int cmd_ls_remote(int argc, const char **argv, const char *prefix)
 {
 	int i;
 	const char *dest = NULL;
-	int nongit;
 	unsigned flags = 0;
 	int quiet = 0;
 	const char *uploadpack = NULL;
@@ -42,8 +41,6 @@ int cmd_ls_remote(int argc, const char **argv, const char *prefix)
 	struct transport *transport;
 	const struct ref *ref;
 
-	setup_git_directory_gently(&nongit);
-
 	for (i = 1; i < argc; i++) {
 		const char *arg = argv[i];
 
git.c +2 -2
index 92ec5d3..63630e7 100644
@@ -345,7 +345,7 @@ static void handle_internal_command(int argc, const char **argv)
 		{ "log", cmd_log, RUN_SETUP | USE_PAGER },
 		{ "ls-files", cmd_ls_files, RUN_SETUP },
 		{ "ls-tree", cmd_ls_tree, RUN_SETUP },
-		{ "ls-remote", cmd_ls_remote },
+		{ "ls-remote", cmd_ls_remote, RUN_SETUP_GENTLY },
 		{ "mailinfo", cmd_mailinfo },
 		{ "mailsplit", cmd_mailsplit },
 		{ "merge", cmd_merge, RUN_SETUP | NEED_WORK_TREE },
@@ -366,7 +366,7 @@ static void handle_internal_command(int argc, const char **argv)
 		{ "pack-objects", cmd_pack_objects, RUN_SETUP },
 		{ "pack-redundant", cmd_pack_redundant, RUN_SETUP },
 		{ "patch-id", cmd_patch_id },
-		{ "peek-remote", cmd_ls_remote },
+		{ "peek-remote", cmd_ls_remote, RUN_SETUP_GENTLY },
 		{ "pickaxe", cmd_blame, RUN_SETUP },
 		{ "prune", cmd_prune, RUN_SETUP },
 		{ "prune-packed", cmd_prune_packed, RUN_SETUP },
Signature · 1 line
1.7.2.1.544.ga752d.dirty
#35 lore raw

PATCH 11/12 var: run setup_git_directory_gently() sooner

In reply to Jonathan Nieder

From: Nguyễn Thái Ngọc Duy <[email protected]>

Part of a campaign to make repository-local configuration available early (simplifying the startup sequence for built-in commands).

---
snuck in a little vertical compression while at it ;-)

 builtin/var.c |    9 ++-------
 git.c         |    2 +-
 2 files changed, 3 insertions(+), 8 deletions(-)
builtin/var.c +2 -7
index 70fdb4d..0744bb8 100644
@@ -74,14 +74,9 @@ static int show_config(const char *var, const char *value, void *cb)
 
 int cmd_var(int argc, const char **argv, const char *prefix)
 {
-	const char *val;
-	int nongit;
-	if (argc != 2) {
+	const char *val = NULL;
+	if (argc != 2)
 		usage(var_usage);
-	}
-
-	setup_git_directory_gently(&nongit);
-	val = NULL;
 
 	if (strcmp(argv[1], "-l") == 0) {
 		git_config(show_config, NULL);
git.c +1 -1
index 63630e7..84bef76 100644
@@ -398,7 +398,7 @@ static void handle_internal_command(int argc, const char **argv)
 		{ "update-ref", cmd_update_ref, RUN_SETUP },
 		{ "update-server-info", cmd_update_server_info, RUN_SETUP },
 		{ "upload-archive", cmd_upload_archive },
-		{ "var", cmd_var },
+		{ "var", cmd_var, RUN_SETUP_GENTLY },
 		{ "verify-tag", cmd_verify_tag, RUN_SETUP },
 		{ "version", cmd_version },
 		{ "whatchanged", cmd_whatchanged, RUN_SETUP | USE_PAGER },
Signature · 1 line
1.7.2.1.544.ga752d.dirty
#36 lore raw

PATCH 12/12 merge-file: run setup_git_directory_gently() sooner

In reply to Jonathan Nieder

From: Nguyễn Thái Ngọc Duy <[email protected]>

Part of a campaign to make repository-local configuration available early (simplifying the startup sequence for built-in commands).

---
As before (except the commit message).

Well, that’s it. There’s also a separate fix for "check-ref-format --branch", which I’ll send under separate cover.

Still to be considered are various commands that currently do not do an unconditional repository search, especially low-level commands that are not about the current repository:
verify-pack, mailinfo, hash-object, archive --remote, check-ref-format, diff --no-index, help, and all the others. Let’s bite off one thing at a time. :)

Thanks for reading; I hope the patches were not too dull. Thoughts (especially improvements) welcome, as always.

 builtin/merge-file.c |    4 +---
 git.c                |    2 +-
 2 files changed, 2 insertions(+), 4 deletions(-)
builtin/merge-file.c +1 -3
index b8e9e5b..b6664d4 100644
@@ -28,7 +28,6 @@ int cmd_merge_file(int argc, const char **argv, const char *prefix)
 	xmparam_t xmp = {{0}};
 	int ret = 0, i = 0, to_stdout = 0;
 	int quiet = 0;
-	int nongit;
 	struct option options[] = {
 		OPT_BOOLEAN('p', "stdout", &to_stdout, "send results to standard output"),
 		OPT_SET_INT(0, "diff3", &xmp.style, "use a diff3 based merge", XDL_MERGE_DIFF3),
@@ -50,8 +49,7 @@ int cmd_merge_file(int argc, const char **argv, const char *prefix)
 	xmp.style = 0;
 	xmp.favor = 0;
 
-	prefix = setup_git_directory_gently(&nongit);
-	if (!nongit) {
+	if (startup_info->have_repository) {
 		/* Read the configuration file */
 		git_config(git_xmerge_config, NULL);
 		if (0 <= git_xmerge_style)
git.c +1 -1
index 84bef76..5433875 100644
@@ -350,7 +350,7 @@ static void handle_internal_command(int argc, const char **argv)
 		{ "mailsplit", cmd_mailsplit },
 		{ "merge", cmd_merge, RUN_SETUP | NEED_WORK_TREE },
 		{ "merge-base", cmd_merge_base, RUN_SETUP },
-		{ "merge-file", cmd_merge_file },
+		{ "merge-file", cmd_merge_file, RUN_SETUP_GENTLY },
 		{ "merge-index", cmd_merge_index, RUN_SETUP },
 		{ "merge-ours", cmd_merge_ours, RUN_SETUP },
 		{ "merge-recursive", cmd_merge_recursive, RUN_SETUP | NEED_WORK_TREE },
Signature · 1 line
1.7.2.1.544.ga752d.dirty
#37 lore raw

cover fix "check-ref-format --branch" from subdir of toplevel

In reply to Jonathan Nieder

Here’s a quick fix from the nd/setup series. Tested against master, though it probably should be against maint.

Jonathan Nieder (2):
check-ref-format: split off functions for subcommands check-ref-format --branch: run repository search

 builtin/check-ref-format.c  |   44 ++++++++++++++++++++++++++----------------
 t/t1402-check-ref-format.sh |   17 ++++++++++++++++
 2 files changed, 44 insertions(+), 17 deletions(-)
#38 lore raw

PATCH 1/2 check-ref-format: handle subcommands in separate functions

The code for each subcommand should be easier to read and manipulate this way.

---
 builtin/check-ref-format.c |   42 +++++++++++++++++++++++++-----------------
 1 files changed, 25 insertions(+), 17 deletions(-)
builtin/check-ref-format.c +25 -17
index b106c65..8707ee9 100644
@@ -33,28 +33,36 @@ static void collapse_slashes(char *dst, const char *src)
 	*dst = '\0';
 }
 
+static int check_ref_format_branch(const char *arg)
+{
+	struct strbuf sb = STRBUF_INIT;
+
+	if (strbuf_check_branch_ref(&sb, arg))
+		die("'%s' is not a valid branch name", arg);
+	printf("%s\n", sb.buf + 11);
+	return 0;
+}
+
+static int check_ref_format_print(const char *arg)
+{
+	char *refname = xmalloc(strlen(arg) + 1);
+
+	if (check_ref_format(arg))
+		return 1;
+	collapse_slashes(refname, arg);
+	printf("%s\n", refname);
+	return 0;
+}
+
 int cmd_check_ref_format(int argc, const char **argv, const char *prefix)
 {
 	if (argc == 2 && !strcmp(argv[1], "-h"))
 		usage(builtin_check_ref_format_usage);
 
-	if (argc == 3 && !strcmp(argv[1], "--branch")) {
-		struct strbuf sb = STRBUF_INIT;
-
-		if (strbuf_check_branch_ref(&sb, argv[2]))
-			die("'%s' is not a valid branch name", argv[2]);
-		printf("%s\n", sb.buf + 11);
-		exit(0);
-	}
-	if (argc == 3 && !strcmp(argv[1], "--print")) {
-		char *refname = xmalloc(strlen(argv[2]) + 1);
-
-		if (check_ref_format(argv[2]))
-			exit(1);
-		collapse_slashes(refname, argv[2]);
-		printf("%s\n", refname);
-		exit(0);
-	}
+	if (argc == 3 && !strcmp(argv[1], "--branch"))
+		return check_ref_format_branch(argv[2]);
+	if (argc == 3 && !strcmp(argv[1], "--print"))
+		return check_ref_format_print(argv[2]);
 	if (argc != 2)
 		usage(builtin_check_ref_format_usage);
 	return !!check_ref_format(argv[1]);
Signature · 1 line
1.7.2.1.544.ga752d.dirty
#39 lore raw

PATCH 2/2 Allow "check-ref-format --branch" from subdirectory

In reply to Jonathan Nieder

check-ref-format --branch requires access to the repository to resolve refs like @{-1}.

Noticed by Duy.

Cc Nguyễn Thái Ngọc Duy <[email protected]>

---
The original used the RUN_SETUP_GENTLY flag to run setup unconditionally. That might still be a good idea, but I am more comfortable running setup just for this subcommand for now so commands like "git check-ref-format refs/foo" are not affected.

 builtin/check-ref-format.c  |    2 ++
 t/t1402-check-ref-format.sh |   17 +++++++++++++++++
 2 files changed, 19 insertions(+), 0 deletions(-)
builtin/check-ref-format.c +2 -0
index 8707ee9..ae3f281 100644
@@ -36,7 +36,9 @@ static void collapse_slashes(char *dst, const char *src)
 static int check_ref_format_branch(const char *arg)
 {
 	struct strbuf sb = STRBUF_INIT;
+	int nongit;
 
+	setup_git_directory_gently(&nongit);
 	if (strbuf_check_branch_ref(&sb, arg))
 		die("'%s' is not a valid branch name", arg);
 	printf("%s\n", sb.buf + 11);
t/t1402-check-ref-format.sh +17 -0
index eb45afb..782e75d 100755
@@ -41,6 +41,23 @@ test_expect_success "check-ref-format --branch @{-1}" '
 	refname2=$(git check-ref-format --branch @{-2}) &&
 	test "$refname2" = master'
 
+test_expect_success 'check-ref-format --branch from subdir' '
+	mkdir subdir &&
+
+	T=$(git write-tree) &&
+	sha1=$(echo A | git commit-tree $T) &&
+	git update-ref refs/heads/master $sha1 &&
+	git update-ref refs/remotes/origin/master $sha1
+	git checkout master &&
+	git checkout origin/master &&
+	git checkout master &&
+	refname=$(
+		cd subdir &&
+		git check-ref-format --branch @{-1}
+	) &&
+	test "$refname" = "$sha1"
+'
+
 valid_ref_normalized() {
 	test_expect_success "ref name '$1' simplifies to '$2'" "
 		refname=\$(git check-ref-format --print '$1') &&
Signature · 1 line
1.7.2.1.544.ga752d.dirty
#40 lore raw

Re: [PATCH 2/2] Allow "check-ref-format --branch" from subdirectory

Quoted 2 lines · Jonathan Nieder

check-ref-format --branch requires access to the repository to resolve refs like @{-1}.

Thanks; the patch makes sense to me.

#41 lore raw
In reply to Jonathan Nieder

2010/8/6 Jonathan Nieder <[email protected]>:

Quoted 14 lines

Nguyễn Thái Ngọc Duy (12):
 git wrapper: introduce startup_info struct
 setup: remember whether repository was found
 git wrapper: allow setup_git_directory_gently() be called earlier
 shortlog: run setup_git_directory_gently() sooner
 grep: run setup_git_directory_gently() sooner
 apply: run setup_git_directory_gently() sooner
 bundle: run setup_git_directory_gently() sooner
 config: run setup_git_directory_gently() sooner
 index-pack: run setup_git_directory_gently() sooner
 ls-remote: run setup_git_directory_gently() sooner
 var: run setup_git_directory_gently() sooner
 merge-file: run setup_git_directory_gently() sooner

Quoted 2 lines

[1] http://thread.gmane.org/gmane.comp.version-control.git/144000/focus=144110

I was waiting for jn/maint-setup-fix to graduate before pushing out some more patches then I got side tracked by the subtree clone. Anyway, thanks!

Signature · 1 line
Duy