Re: unistd.h: daemon

[Bruno Haible]

Hi Simon,

> Glibc has a 'daemon' function, which is from BSD, however it is not
> documented in the glibc manual [1] and it is not in POSIX.
> [1] http://sourceware.org/bugzilla/show_bug.cgi?id=12251

There are many more functions in glibc that are not documented in texinfo
format. But it's documented in /usr/include/unistd.h; that's sufficient
to rely on it.

> Mike noticed that InetUtils is using it:
> 
> http://thread.gmane.org/gmane.comp.gnu.inetutils.bugs/3040/focus=3041
> 
> This begs the question of whether using the 'daemon' interface is a good
> idea or not.

It appears to be a Unix classic. The portmapper, sshd, cvsd use it, certainly
among others. I see nothing wrong with it. There's even a package that
wraps a shell command 'daemonize' around it:
  git://github.com/bmc/daemonize.git

> I see two approaches: 
> 
> 1) Regard the 'daemon' interface as broken, and let InetUtils call a
> different function, e.g., 'daemonize', and implement it in InetUtils
> (which already contains an implementation, so that is easy).
> 
> 2) Consider the 'daemon' interface as working, and provide a gnulib
> module for it so that it can be used interoperable.

What other indications do you have that the interface is "broken"?

The only problem I can see is that the typical implementations call
fork() once or twice - and therefore there is a portability problem
with respect to native Windows platforms (mingw).

I think it's a good candidate for a gnulib module nevertheless.

Bruno