On Fri, 12 Dec 2025, Stephen Borrill wrote:
I've been testing copyparty on NetBSD 10.1
https://github.com/9001/copyparty
It runs just fine, but ignores signals such as SIGINT, so you can't quit
without kill -9.
The same versions of copyparty and python work as expected on FreeBSD 14.3
and OpenBSD 7.8, so there's something a bit funky about NetBSD's signal
handling here.
Hmm. In FreeBSD, OpenBSD & Linux, pthread_sigmask(3) seems to apply per-thread,
but on NetBSD, the whole process gets blocked--not just the thread which called
pthread_sigmask():
```
qemu$ uname -a
NetBSD qemu.local 10.1_STABLE NetBSD 10.1_STABLE (GENERIC) #0: Sat Nov 29
16:02:26 UTC 2025
[email protected]:/usr/src/sys/arch/i386/compile/GENERIC i386
qemu$ ps -p $(pgrep python3.12) -o sigcatch,sigignore,sigmask
CAUGHT IGNORED BLOCKED
20004002 99489000 20004002
qemu$
```
20004002 => USR1, INT & TERM blocked. You can still SIGHUP the python process.
Try this program on NetBSD and the other OSes to see what I think's going on:
```
$ cat pthread_sigmask.c
/**
* On NetBSD, pthread_sigmask() in a thread blocks process too.
* on Linux, it only blocks the calling thread.
*/
#include <err.h>
#include <pthread.h>
#include <signal.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
static void *
sig_thread(void *arg)
{
sigset_t set;
int rc, sig;
/*
* Block SIGINT in thread--main thread should _not_ be affected.
*/
sigemptyset(&set);
sigaddset(&set, SIGINT);
if ((rc = pthread_sigmask(SIG_BLOCK, &set, NULL)) != 0)
errc(EXIT_FAILURE, rc, "pthread_sigmask");
for (;;) {
if ((rc = sigwait(&set, &sig)) != 0)
errc(EXIT_FAILURE, rc, "sigwait");
printf("Thread got signal %d\n", sig);
}
}
int
main(void)
{
pthread_t thread;
int rc;
if ((rc = pthread_create(&thread, NULL, &sig_thread, NULL)) != 0)
errc(EXIT_FAILURE, rc, "pthread_create");
printf("Main waiting for interrupt\n");
pause();
exit(EXIT_SUCCESS);
}
$ cc -pthread -o pt pthread_sigmask.c # -lbsd (Linux)
```
-RVP
