Skip to contentSkip to conversation

submodule status: propagate SIGPIPE

PATCH Superseded Phillip Wood via GitGitGadget 3 participants 6 messages Sep 20, 2024 – Sep 23, 2024

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 submodule status: propagate SIGPIPE

From: Phillip Wood <[email protected]>

It has been reported than running

     git submodule status --recurse | grep -q ^+

results in an unexpected error message

fatal: failed to recurse into submodule $submodule

When "git submodule--helper" recurses into a submodule it creates a child process. If that process fails then the error message above is displayed by the parent. In the case above the child is killed by SIGPIPE as "grep -q" exits as soon as it sees the first match. Fix this by propagating SIGPIPE so that it is visible to the process running git. We could propagate other signals but I'm not sure there is much value in doing that. In the common case of the user pressing Ctrl-C or Ctrl-\ then SIGINT or SIGQUIT will be sent to the foreground process group and so the parent process will receive the same signal as the child.

---
submodule status: propagate SIGPIPE

Note that I haven't checked if any other submodule subcommands are affected by this - I'll leave that to someone more familiar with the code.

Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-1799%2Fphillipwood%2Fsubmodule-status-sigpipe-v1
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-1799/phillipwood/submodule-status-sigpipe-v1
Pull-Request: https://github.com/gitgitgadget/git/pull/1799

 builtin/submodule--helper.c | 7 ++++++-
 t/t7422-submodule-output.sh | 7 +++++++
 2 files changed, 13 insertions(+), 1 deletion(-)
builtin/submodule--helper.c +6 -1
index a46ffd49b34..a8e497ef3c6 100644
@@ -30,6 +30,7 @@
 #include "advice.h"
 #include "branch.h"
 #include "list-objects-filter-options.h"
+#include <signal.h>
 
 #define OPT_QUIET (1 << 0)
 #define OPT_CACHED (1 << 1)
@@ -695,6 +696,7 @@ static void status_submodule(const char *path, const struct object_id *ce_oid,
 
 	if (flags & OPT_RECURSIVE) {
 		struct child_process cpr = CHILD_PROCESS_INIT;
+		int res;
 
 		cpr.git_cmd = 1;
 		cpr.dir = path;
@@ -710,7 +712,10 @@ static void status_submodule(const char *path, const struct object_id *ce_oid,
 		if (flags & OPT_QUIET)
 			strvec_push(&cpr.args, "--quiet");
 
-		if (run_command(&cpr))
+		res = run_command(&cpr);
+		if (res == SIGPIPE + 128)
+			raise(SIGPIPE);
+		else if (res)
 			die(_("failed to recurse into submodule '%s'"), path);
 	}
 
t/t7422-submodule-output.sh +7 -0
index ab946ec9405..c1686d6bb5f 100755
@@ -167,4 +167,11 @@ do
 	'
 done
 
+test_expect_success !MINGW 'git submodule status --recursive propagates SIGPIPE' '
+	{ git submodule status --recursive 2>err; echo $?>status; } |
+		grep -q X/S &&
+	test_must_be_empty err &&
+	test_match_signal 13 "$(cat status)"
+'
+
 test_done

base-commit: ed155187b429a2a6b6475efe1767053df37ccfe1

Signature · 1 line
gitgitgadget
#1 lore raw
Quoted 10 lines · Phillip Wood via GitGitGadget

From: Phillip Wood <[email protected]>

It has been reported than running

     git submodule status --recurse | grep -q ^+

results in an unexpected error message

fatal: failed to recurse into submodule $submodule
...

builtin/submodule--helper.c static void status_submodule(const char *path, const struct object_id *ce_oid, · line 717

Quoted 4 lines
builtin/submodule--helper.c +3 -1
-		if (run_command(&cpr))
+		res = run_command(&cpr);
+		if (res == SIGPIPE + 128)
+			raise(SIGPIPE);

OK, that is straight-forward. This makes sure that we do the same thing we would do if we, not our child, got a SIGPIPE.

Quoted 9 lines
builtin/submodule--helper.c +1 -0
+		else if (res)
 			die(_("failed to recurse into submodule '%s'"), path);
 	}
t/t7422-submodule-output.sh +0 -0
index ab946ec9405..c1686d6bb5f 100755

t/t7422-submodule-output.sh do · line 174

Quoted 8 lines
t/t7422-submodule-output.sh +5 -0
@@ -167,4 +167,11 @@ do
 	'
 done
 
+test_expect_success !MINGW 'git submodule status --recursive propagates SIGPIPE' '
+	{ git submodule status --recursive 2>err; echo $?>status; } |
+		grep -q X/S &&
+	test_must_be_empty err &&
+	test_match_signal 13 "$(cat status)"

I am not a huge fun of assuming SIGPIPE is 13 everywhere, but at least we can tweak test_match_signal when we find oddball systems, so ... OK.

In practice, we only use 13 and 15 with test_match_signal, so we could have a new "test-tool signal-name" that maps textual signal names to the number the platform gives to them for the platform on which the tests are running, if it ever turns out to be a problem.

Looking good.

Will queue. Thanks.

#2 lore raw
In reply to Phillip Wood via GitGitGadget

builtin/submodule--helper.c line 33

Quoted 9 lines · Phillip Wood via GitGitGadget
builtin/submodule--helper.c +1 -0
index a46ffd49b34..a8e497ef3c6 100644
@@ -30,6 +30,7 @@
 #include "advice.h"
 #include "branch.h"
 #include "list-objects-filter-options.h"
+#include <signal.h>

Do we really need this?

As with any other Git built-in that relies on git-compat-util.h to handle such system-dependencies, direct inclusion of system headers like this is highly questionable.

#3 lore raw

Hi Junio

Quoted 14 lines · Junio C Hamano

builtin/submodule--helper.c line 33

Quoted 9 lines · Phillip Wood via GitGitGadget
builtin/submodule--helper.c +1 -0
index a46ffd49b34..a8e497ef3c6 100644
@@ -30,6 +30,7 @@
  #include "advice.h"
  #include "branch.h"
  #include "list-objects-filter-options.h"
+#include <signal.h>

Do we really need this?

Quoted 3 lines

As with any other Git built-in that relies on git-compat-util.h to handle such system-dependencies, direct inclusion of system headers like this is highly questionable.

Good point - I really need to figure out how to stop emacs' lsp mode automatically adding includes. I removed its "helpful" addition of <csignal> but forgot to remove <signal.h> as well.

Thanks

Phillip

#4 lore raw

PATCH v2 1/1 submodule status: propagate SIGPIPE

In reply to Phillip Wood via GitGitGadget

From: Phillip Wood <[email protected]>

It has been reported than running

     git submodule status --recurse | grep -q ^+

results in an unexpected error message

fatal: failed to recurse into submodule $submodule

When "git submodule--helper" recurses into a submodule it creates a child process. If that process fails then the error message above is displayed by the parent. In the case above the child is killed by SIGPIPE as "grep -q" exits as soon as it sees the first match. Fix this by propagating SIGPIPE so that it is visible to the process running git. We could propagate other signals but I'm not sure there is much value in doing that. In the common case of the user pressing Ctrl-C or Ctrl-\ then SIGINT or SIGQUIT will be sent to the foreground process group and so the parent process will receive the same signal as the child.

---
submodule status: propagate SIGPIPE

Note that I haven't checked if any other submodule subcommands are affected by this - I'll leave that to someone more familiar with the code.

Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-1799%2Fphillipwood%2Fsubmodule-status-sigpipe-v2
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-1799/phillipwood/submodule-status-sigpipe-v2
Pull-Request: https://github.com/gitgitgadget/git/pull/1799

Range-diff vs v1:

1: 64233d5ee0a ! 1: 169e2c06f1a submodule status: propagate SIGPIPE
@@ Commit message

## builtin/submodule--helper.c ##
-@@
- #include "advice.h"
- #include "branch.h"
- #include "list-objects-filter-options.h"
-+#include <signal.h>
-
- #define OPT_QUIET (1 << 0)
- #define OPT_CACHED (1 << 1)
@@ builtin/submodule--helper.c: static void status_submodule(const char *path, const struct object_id *ce_oid,

if (flags & OPT_RECURSIVE) {

 builtin/submodule--helper.c | 6 +++++-
 t/t7422-submodule-output.sh | 7 +++++++
 2 files changed, 12 insertions(+), 1 deletion(-)
builtin/submodule--helper.c +5 -1
index a46ffd49b34..4daca494b80 100644
@@ -695,6 +695,7 @@ static void status_submodule(const char *path, const struct object_id *ce_oid,
 
 	if (flags & OPT_RECURSIVE) {
 		struct child_process cpr = CHILD_PROCESS_INIT;
+		int res;
 
 		cpr.git_cmd = 1;
 		cpr.dir = path;
@@ -710,7 +711,10 @@ static void status_submodule(const char *path, const struct object_id *ce_oid,
 		if (flags & OPT_QUIET)
 			strvec_push(&cpr.args, "--quiet");
 
-		if (run_command(&cpr))
+		res = run_command(&cpr);
+		if (res == SIGPIPE + 128)
+			raise(SIGPIPE);
+		else if (res)
 			die(_("failed to recurse into submodule '%s'"), path);
 	}
 
t/t7422-submodule-output.sh +7 -0
index ab946ec9405..c1686d6bb5f 100755
@@ -167,4 +167,11 @@ do
 	'
 done
 
+test_expect_success !MINGW 'git submodule status --recursive propagates SIGPIPE' '
+	{ git submodule status --recursive 2>err; echo $?>status; } |
+		grep -q X/S &&
+	test_must_be_empty err &&
+	test_match_signal 13 "$(cat status)"
+'
+
 test_done

base-commit: ed155187b429a2a6b6475efe1767053df37ccfe1

Signature · 1 line
gitgitgadget
#5 lore raw
Quoted 21 lines · Phillip Wood via GitGitGadget

From: Phillip Wood <[email protected]>

It has been reported than running

     git submodule status --recurse | grep -q ^+

results in an unexpected error message

fatal: failed to recurse into submodule $submodule

When "git submodule--helper" recurses into a submodule it creates a child process. If that process fails then the error message above is displayed by the parent. In the case above the child is killed by SIGPIPE as "grep -q" exits as soon as it sees the first match. Fix this by propagating SIGPIPE so that it is visible to the process running git. We could propagate other signals but I'm not sure there is much value in doing that. In the common case of the user pressing Ctrl-C or Ctrl-\ then SIGINT or SIGQUIT will be sent to the foreground process group and so the parent process will receive the same signal as the child.

Quoted 8 lines

---
submodule status: propagate SIGPIPE

Note that I haven't checked if any other submodule subcommands are affected by this - I'll leave that to someone more familiar with the code.

Thanks. Queued.