[Freeciv-Dev] Re: new connect dialog (ver 1) (PR#1911)

[Mike Kaufman <kaufman@xxxxxxxxxxxxxxxxxxxxxx>]

On Wed, Aug 14, 2002 at 07:35:10PM +0200, Christian Knoke wrote:
> On Mon, Aug 12, 2002 at 11:45:45PM -0700, Mike Kaufman wrote:
> 
> > I need testers in a big way.
> 
> I only tried Gtk+ 1.2
> 
> I like it, it works really stable and reasonable.
> 
> And now the back side ...
> 
> > Specific code additions (outside the gui):
> > 
> > improved game loading with server started
> >  o the server won't just quit anymore when a bogus savegame is loaded.
> >  o the attributes bug isn't fixed yet :(
> 
> Don't know what that is.

this was "CMA crash after game load". turns out this problem is avoided
with the patch since the rename command is used.

> 
> > new server command line option
> >  o you can now set the access level of the first connection.
> 
> I wonder if this shouldn't be restricted. Cmdlevel hack only
> accessible if client is local (on the same machine)?

this is pointless since if you're starting the server, the console has
hack level access anyway.

> 
> >  o this is needed to give 'hack' to the player without stdout parsing
> >    madness.
> 
> Very good.
> 
> > new packet that sends various player info to the client on request.
> >   o only sends name, username, nation, is_alive, and, ai.control
> >   o need to have hack to have server pay attention to your request
> 
> Why?

this is probably a holdover from my old patches when we had non-established 
connections to the server sending this packet. It would have allowed a DOS 
attack. It can probably be moved back to info. after all, we're actually
sending less packet traffic than a list command. (all this crap was
needed when we had to load off the commandline)

> 
> >   o needed to be able to select players out of a savegame.
> > 
> > new server command: rename
> >   o was not implemented.
> >   o now renames the _username_ of a connection.
> >   o should go away again when auth is implemented.
> >   o needed to select players out of a savegame.
> > 
> > add some data for the chatline help
> >   o explain sending commands with chatline.
> > 
> > 
> > BUGS:
> >   generally: the attributes thing (which is not really a connectdlg bug)
> >   GTK: none that I'm aware of.
> 
> Well ...
> 
> - first a problem which arised with the latest cvs bugs. When the server
>   crashes, you never know what's happening, no output, no log. There
>   should be a way to pass a server log option to civserver.

this definitely is a problem. A solution would be to comment out:

fclose(stdout);
fclose(stderr); 

in connectdlg_common.c:117

this will spam the terminal you launched your client from with server
messages.

> 
> - There often remain stale civserver processes, e.g. if you click
>   "previous".

all of these should die after ultimate quit. do they?

> 
> - The AI level option is i18n'd, which doesn't work since civserver
>   doesn't understand "schwer" or "leicht".

this is fixable.

> 
> - It's very easy to cheat, just take a dead barbar, and you'll see the 
>   whole map.

hey, you've got the savegame on your own machine. it's just as easy to
load it into civworld and peek. what we might need is a ternary
differentiator: human/ai/barbarian


> 
> - probably a typo in server/civserver.c:113
> 
>       else if (strcasecmp(option, "info") == 0)
>         srvarg.first_level = ALLOW_NONE;

looks like it.

> 
> - Terminal is corrupted after exit.

damn it, when? I thought that I'd fixed all of those cases. This is
probably related to an already submitted bug with the atexit() proposed
fix. Does it always corrupt? For me, it usually only corrupts on a
server crash...

> 
> 
> Now the cosmetic things:
> 
> - Rename "Start" multi-player/network game --> "Join"

ok

> 
> - Rename "Save game..." --> "Save game as ..."

ok

> 
> - Rename "Disconnect" and "Quit" -> "Quit game" and "Quit Freeciv" ?

hmm. I might have to add cases for multiplayer/not

> 
> - The column with "human" is too short. "menschlich" doesn't fit.

ok

> 
> - Is it possible to have the file dialog always add an ".sav" ending
>   if not present? Maybe even filter "*.sav(.gz)?" ?

yes, and maybe (the xaw client does this, gtk may be more of a bitch
about it)

> I feel this patch is a good step to CSI. The next step would be
> a comfortable GUI for option selecting.

I really don't like this idea. It's a massive amount of code for very
little discernible benefit.

-mike