Follow-up Comment #4, bug #21583 (project freeciv):

> can't reproduce on Linux
No surprise -- I think this stuff tends to be horribly OS-dependent.

This is one reason why changing anything in this area scares me, especially
just before a release.

Another similar source of fun is whether listening on a given port on IPv6
also exclusively binds on IPv4. We had lots of fun with this around bug
#15559. (In that case at least it was also configurable, with different Linux
distros choosing different defaults at different times; and *BSDs were
different again.)

A conversation with a friend makes me wonder whether our use of SO_REUSEADDR
might be implicated. This article
(which I've only skimmed so far) looks interesting. (But this could be a
complete red herring.)

>> Quitting the client then killed the manually started server 
>> (unsurprisingly). 
> How that? The server should wait for (re)connects, and does 
> so here. 
I think client_kill_server() is called, which sends "/quit" to what it thinks
is its spawned server. Since the client always ends up with "hack" access to a
server on the same machine (indeed, this can't be disabled: bug #20556) this
succeeds. (Or at least I'm guessing that's what happened, hence my lack of


Reply to this item at:


  Message sent via/by Gna!

Freeciv-dev mailing list

Reply via email to