Fix/cmp n flag bug

[gtchankvetadze]

cmp: fix -n flag incorrectly enabling verbose output

Bug: When using cmp -n NUM without -l, the program would incorrectly continue comparing after finding the first difference, printing multiple 'differ' lines instead of just one.

Root Cause: The bug was in line 144: if (opt) checked if ANY option was set (including -n), causing it to continue. It should only continue for -l (verbose) mode: if (opt & CMP_OPT_l).

Test Case:

echo -n AAAAA > file1
echo -n BBBBB > file2
# Before: cmp -n 3 file1 file2 → 3 lines of output
# After:  cmp -n 3 file1 file2 → 1 line of output
# Both:   cmp -l -n 3 file1 file2 → 3 lines (correct)

[robang74]

This repository is rarely updated as you can see by 100+ commits into your PR, sent the patch to: busybox@busybox.net

By the way, I am using my git-function project to keep update my fork of busybox:

+:git-shell:master:busybox> lsrmt
busybox.net git://busybox.net/busybox.git (fetch)
busybox.net git://busybox.net/busybox.git (push)
origin https://github.com/robang74/busybox.git (fetch)
origin https://github.com/robang74/busybox.git (push)

While git-functions project stays here:

• https://github.com/robang74/git-functions (*)

I will appreciate testers and contributors to the (*) gitshell project. I am very satisfied by using it.

[ningmingxiao]

Hello, I want to let sh process reap zombie processes （generate by orphan process） when it is run as init process in a container.

reproduce it
[root@host ~]# docker run -d busybox sh -c "sleep 1000"
4b333cedcff68d47dbb846b0004c5a163a1cd210b8199e22ee56b80e9700432f
[root@host ~]# docker exec -it 4b sh
/ #
/ # timeout 1 sleep 2
Terminated
/ # timeout 1 sleep 2
Terminated
/ # ps -ef
PID USER TIME COMMAND
1 root 0:00 sleep 1000
7 root 0:00 sh
15 root 0:00 [timeout]
18 root 0:00 [timeout]
19 root 0:00 ps -ef

many reap zombie processes

my  patch is 
cat 0001-feature-let-sh-process-reap-zombie-processes-when-it.patch
From 5dd9be23de0303959aa47000d2afc8d7ae248526 Mon Sep 17 00:00:00 2001
From: ningmingxiao <ning.mingxiao@zte.com.cn>
Date: Mon, 20 Apr 2026 09:25:43 +0800
Subject: [PATCH] feature: let sh process reap zombie processes when it's init
 process

---
 shell/ash.c | 19 ++++++++++++++++++-
 1 file changed, 18 insertions(+), 1 deletion(-)

diff --git a/shell/ash.c b/shell/ash.c
index fb887f31b..ed8cd67f6 100644
--- a/shell/ash.c
+++ b/shell/ash.c
@@ -10900,6 +10900,13 @@ bltincmd(int argc UNUSED_PARAM, char **argv UNUSED_PARAM)
         * as POSIX mandates */
        return back_exitstatus;
 }
+
+static void sigchld_handler(int signum) {
+  pid_t pid;
+  while ((pid = waitpid(-1, NULL, WNOHANG)) > 0) {
+  }
+}
+
 static int
 evalcommand(union node *cmd, int flags)
 {
@@ -11154,7 +11161,17 @@ evalcommand(union node *cmd, int flags)
                 * in a script or a subshell does not need forking,
                 * we can just exec it.
                 */
-               if (!(flags & EV_EXIT) || may_have_traps) {
+               if (!(flags & EV_EXIT) || may_have_traps||(getpid() == 1)) {
+                       if (getpid() == 1) {
+                               pid_t pid;
+                               struct sigaction sa;
+                               sa.sa_handler = sigchld_handler;
+                               sa.sa_flags = SA_NOCLDSTOP | SA_RESTART;
+                               sigemptyset(&sa.sa_mask);
+                               if (sigaction(SIGCHLD, &sa, NULL) < 0) {
+                                       return -1;
+                               }
+                       }
                        /* No, forking off a child is necessary */
                        INTOFF;
                        get_tty_state();
-- 
2.27.0

@robang74 I send email to busybox@busybox.net but nobody reply

[robang74]

@robang74 I send email to busybox@busybox.net but nobody reply

The getpid() is a function it cost byte in "text" section and time in "running", and you perform two times instead of storing it into a variable. CPU caching hasn't made for support poor coding.

Moreover, the subject says: "let sh process reap zombie processes when it's init process" and pid == 1 is about init process, so what are you trying to do with in that code to init?

Zombi process are there for a reason, and the reason isn't you write a patch to get read of them. It is because you do not use the container properly. At the 99.9%, an potentially you might have catch a 1-over-a-million bug into cointainer technology and also in that case isn't busybox to fix.

[ningmingxiao]

It always happen.
It's because new kernel version (orphan process must be adopted by process 1 in a container ) see https://github.com/containerd/containerd/blob/main/docs/runtime-v2.md#containerd-shim-runc-v2-as-sub-reaper for docker we have to set docker run --init to use a init process to reap zombie processes (but I don't want to add a external binary into the container) the init process in a container sometimes will reap zombie processes for their child processes but amost never wait grandchild processes exited.