Abort the push operation (i.e. do not migrate the objects from temporary
to permanent storage) if the client has disconnected while the
pre-receive hook was running.

This reduces the risk of inconsistencies on network errors or if the
user hits ctrl-c while the pre-receive hook is running.

Send a keepalive packet (empty) on sideband 2 (the one to report
progress). If the client has exited the write() operation should fail
and the push will be aborted. This only works when sideband*
capabilities are advertised by the client.

Note: if the write() operation fails, receive-pack will likely be killed
via SIGPIPE and even so, since the client is likely gone already, the
error strings will go nowhere. I only added them for code consistency.

Signed-off-by: Robin Jarry <redacted>
---
v3 -> v4:
  - reworded the comment block s/ensure/notice/
  - used write_in_full() instead of write_or_die()
  - set error_string fields for code consistency

 builtin/receive-pack.c | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)

diff --git a/builtin/receive-pack.c b/builtin/receive-pack.c
index 9f4a0b816cf9..f8b9a9312733 100644
--- a/builtin/receive-pack.c
+++ b/builtin/receive-pack.c

@@ -1971,6 +1971,22 @@ static void execute_commands(struct command *commands,
 		return;
 	}
 
+	/*
+	 * Send a keepalive packet on sideband 2 (progress info) to notice
+	 * a client that has disconnected (e.g. killed with ^C) while
+	 * pre-receive was running.
+	 */
+	if (use_sideband) {
+		static const char buf[] = "0005\2";
+		if (write_in_full(1, buf, sizeof(buf) - 1) < 0) {
+			for (cmd = commands; cmd; cmd = cmd->next) {
+				if (!cmd->error_string)
+					cmd->error_string = "pusher went away";
+			}
+			return;
+		}
+	}
+
 	/*
 	 * Now we'll start writing out refs, which means the objects need
 	 * to be in their final positions so that other processes can see them.

-- 
2.35.0.4.g44a5d4affccf