crash in afp_flush() on macOS when log file is not defined

[rdmark]

As reported by user Slimes on the TinkerDifferent web forums, there is a recurring crash when you run netatalk with debug logging enabled but no log file defined (i.e. logging to syslog, which macOS doesn't support.)

The pattern was:

crashing was every 30 to 60 seconds after connecting from os9.2 quicksilver chooser.

[rdmark]

An example of a working afp.conf configuration

;
; Netatalk 4.x configuration file
;

[Global]
; Global server settings
log level = default:debug
log file = /opt/homebrew/var/netatalk/netatalk4.log
uam list = uams_clrtxt.so uams_guest.so

[Homes]
;basedir regex = /Users << this line seems to make no difference.

[AFP Work]
path = /Volumes/Work
volume name = AFP Work

; [my backup]
; path = /path/to/backup
; time machine = yes
; volume name = My Backup Volume

[rdmark]

Callstack

0 libsystem_kernel.dylib 0x19688a5b0 __pthread_kill + 8
1 libsystem_pthread.dylib 0x1968c4888 pthread_kill + 296
2 libsystem_c.dylib 0x1967ca808 abort + 124
3 libatalk.19.dylib 0x10023744c sig_fault + 288
4 libsystem_platform.dylib 0x1968ce744 _sigtramp + 56
5 afpd 0x100181d84 afp_flush + 28
6 afpd 0x100181d84 afp_flush + 28
7 afpd 0x10016b3b8 afp_over_dsi + 1632
8 afpd 0x1001890cc main + 2316
9 dyld 0x196505d54 start + 7184