[Mono-dev] System.Unix.Native.Syscall.fork is commented out

[Jonathan Pryor]

On Mon, 2008-11-03 at 14:49 -0800, Sixes wrote:
> Jonathan Pryor wrote:
> > On Wed, 2008-09-03 at 10:33 +0200, Robert Jordan wrote:
> > 
> > Alas, I didn't get any further details, but that was enough for me to
> > remove fork() from Syscall: if it can't ever be reliably called, then it
> > shouldn't be callable.
>
> If you don't want to put System.Unix.Native.Syscall.fork() back in, how
> about a simple, static System.Unix.Native.Daemon class? It could be
> documented as only usable as the first call in Main().

That still wouldn't work, as *before* Main() is invoked there are
already multiple threads running.  You might not have the unknown lock
state issue, but you will have the runtime assuming that threads exist
which won't exist in the child process.  Oops.

So a "better" approach would be to look for the -d flag, and if present
detach from the controlling terminal (forget what it is), respawn the
current process via System.Diagnostics.Process w/o the -d flag, then
exit.  This sill create a child process that isn't attached to a
terminal, thus doing (with more work) what daemon(3) would do...

 - Jon