Starting lynx (from ports, version 2.8.9) always starts with two processes: one "normal" working process and one child zombie process.
Rebuilding didn't help.
Uname -a: FreeBSD yokozuna 11.2-STABLE FreeBSD 11.2-STABLE #0 r335773: Fri Jun 29 15:55:48 CEST 2018 root@yokozuna:/usr/obj/usr/src/sys/YOKOZUNA amd64
I can't reproduce this. Can you please provide more details?
I don't know if I can tell much more. These are the lynx processes in htop after start:
66549 marco 20 0 23364 8884 S 0.0 0.2 0:03.57 │ │ ├─ xterm -geometry 110x40 -title lynx
66550 marco 20 0 16180 10480 S 0.0 0.3 0:00.77 │ │ │ └─ lynx
67187 marco 21 0 0 0 Z 0.0 0.0 0:00.01 │ │ │ └─ lynx
I guess you're starting lynx through xterm: xterm -e lynx ... , so this is not lynx but xterm's bug. I still can't reproduce this zombie lynx, though.
I'm not sure. It also happens when starting lynx in urxvt and it didn't happen in the previous version 220.127.116.11.
But it isn't a big problem either, lynx runs fine.
Hmm. I've compared code between 2-8-8rel2 and 2.8.9rel1. There are some changes in WWW/Library/Implementation/HTTCP.c (the only place where lynx uses fork). Can you please:
1) Start xterm/rxvt, run lynx in it manually and see if you still can reproduce this zombie issue.
2) Remove --enable-nsl-fork from CONFIGURE_ARGS in port's Makefile, rebuild port and try to reproduce lynx's zombie.
Yes removing the "--enable-nsl-fork" in the Makefile did the trick. Zombie is gone.
(In reply to Marco Beishuizen from comment #6)
Based on this result, do we still need to keep this bug open?
Not sure. The solution is more of a workaround. I can make a patch to make "--enable-nsl-fork" optional?
It's easy to add such option, but I don't see much reason to do that. The zombie should disappear after exiting parent lynx, so this situation is harmless.