On Wednesday 24 January 2007 18:55, Brett Nash wrote:
> > > To be frank - I found it a PITA.  Now I to be honest a lot of this info
> > > I have since found, but it is scattered - so it's not easy.  So at this
> > > point I'm going to be a real critical bastard.
<snip>
> Okay first server won't let me log in, second won't resolve... started
> trying to get a local server.
>
<snip>

>       Download tpserver-cpp
>       -> Configure... failed
>       apt-get lots more
>       -> Configure... failed
>       Download libtprl
>       -> Configure... failed
>       apt-get a little more
>       -> Configure && make  && sudo make install
>       -> Back to tpserver-cpp... configure && make && sudo make install

The README file in tpserver-cpp outlines most (if not all) the dependences of 
tpserver-cpp.

>       % hash -r # update shell paths
>       % tpserver-cpp
>       default-load-handler: cannot open input file:
> "/home/nash/work/thousandparsec/../modules/tpcl/mzscheme/designstruct.scm"
> (No such file or directory; errno=2) default-load-handler: cannot open
> input file:
> "/usr/local/share/designstruct.scm" (No such file or directory; errno=2)
> zsh: segmentation fault (core dumped)  tpserver-cpp

I have fixed this now.  I tend to use the guile tpcl/scheme implementation, 
which is a little nicer about the file location problem. 

> Urr...
>       Spot a suspicious ..
> % cd ~/work/thousandparsec/tpserver-cpp/tpserver
> % ./tpserver-cpp
>
> Nothing....

Yes, nothing.  If you had read the README file (or the updated getting started 
page) you would have seen that tpserver-cpp uses a config file.

I have just added a few defaults for when someone runs tpserver-cpp without 
any command line arguments. Still doesn't get it on the network though.

I am also providing two config files for the AI Competition in a few days 
time.

> In another shell...
> nmap localhost...
> Nope, no new ports open...

There is a good chance that nmap would not map the ports tpserver-cpp uses.

> No idea...
>
> Download stable release...
>
> Repeat process...
>
> Search website... work out reference to conf file is _really_ necessary
> and it is unable to find it's own in a sane way.

Well, it tries /etc/tpserver.conf or the one provided on the command line. I 
have two local config files and I change between them to test things. The 
sample.conf file doesn't do anything, it's all commented out.

> Uncomment stuff in config file, run tpserver -C nash.conf
>
> Now it is working.
>
> Note I followed typical user patters - didn't read doco, just expected
> the default to work.

I know of the open issues around the console. A prompt will be put in once I 
get it sorted.

I also intend to add a "wizard" or setup guide on the console when it detects 
no configuration (namely when a ruleset is not loaded). This is a big tasked 
and will not be completed soon.  I may add a "status" message on the console 
to give a clue about what the user has to do next.

> > > So some things I'd like to see/see improved:
> > >  * A quick overview of how to set things up
> > >   - It would be nice to have every step from the first apt-get/yum etc
> > > to entering data for the first turn.  Currently it stops at running the
> > > server.

I have updated it to help people a bit. Running a server is not for everyone.

<snip>
> > >  * Dependencies - Both distro and TP related ones
> > >   - The Python modules needed aren't always easy to figure out since
> > > normally you find yourself staring at a stack trace, and need to work
> > > out which python-X lib you need to install.
> >
> > I'm guessing this applies to the client?
> >
> > The following URL should list everything you need for the stable version
> > of the client,
> > http://www.thousandparsec.net/tp/gettingstarted.php#gamer
>
> Not quite... At that point I have a window open, and I need to figure
> out how to connect to a server.  I personally could not get it to
> connect to the main servers (the second test server didn't resolve at
> the time).  Hence I found my next step was to set up a server.

demo1.thousandparsec.net is almost always running.  If you have problems 
logging in, try a different username and password. demo1 has autoadd_players 
on.

> Also setting up an account 'New' wasn't what I expected.  I was thinking
> it meant "New Game", not "Create an account".
>
> > I've yet to write a "Getting Started" for developers.
>
> May be useful for AI comp...
>
> > The tpclient-pywx should be checking for all it's required dependencies
> > in the requirements.py. If you got a traceback instead of an informative
> > message, could you please post a bug report?
>
> Didn't see that script.  Just tried to run it directly, then tried the
> setup.
>
> > >   - The C++ ones are better, at least the configure ones are meaningful
> > >           -> However it would be nice if ALL the dependencies where 
> > > checked
> > > before it barfed - a nice "You are missing the following libs:
> > >   libfoo (>2.3)
> > >   libbar
> > >   ...
> > >           This is a general autoconf bitch however - it is tough to get 
> > > it all
> > > right.
> >
> > I don't think you can even do this with autoconf?
>
> You can.  Basically when you do your requirements if one fails you note
> it and continue as long as you can (avoiding expensive tests).  Then
> dump a list everything that failed as you exit.
>
> But as I said, it's a PITA, and autoconf doesn't make it easy.  I was
> impressed that inkscape does this in the build for the things it tests
> with pkg-config (gtkmm and friends)

I use pkg-config for a few things, as libraries slowly change over to them. I 
am still working on separating the libraries as some are only needed for one 
module, but are linked in server-wide.

<snip>
> > >  * Which server and client versions we should use... Python or CPP?
> > > Which is easiest to get going etc?  Which works.
> >
> > The Python server will very soon become the easiest to setup and install
> > (as it will be packaged with a setup wizard). However at the moment it
> > is in a state of flux as I added SQLite support a fix a few other
> > problems.
>
> This should also improve things ;-)  Unfrotunately the simple fact is a
> new user will try it now, and either get it to work or give up on it.
>
> > I agree that the C++ server needs better information about setup.
>
> Once set up it all seems good.

Cool. I hope the updated getting started page helps.

> > >  * Once a server is installed, tpserver-cpp should just work.
> > >   Having to set up a config file (by default in /etc/... too!) is
> > > annoying.  The supplied config file should be ready to go.
> >
> > What should the default config file do? There are currently two possible
> > rulesets that could be running.
>
> Then pick the most advanced complete one.  My understanding was that
> only minisec was actually fully implemented... I pulled that from the
> website somewhere.

It's true. I wouldn't run MTSec just yet. As I said above, I will provide two 
config files for the AI Competition.

> Even better would be if one server could run multiple games with
> different rule sets....

Apparently the python server can.  tpserver-cpp will not, just too complex.

> > >   Even better would be a quick start in the client to launch the server
> > > with the appropriate parameters.
> >
> > This is coming when tpserver-py is integrated into tpclient-pywx.
>
> Does that allow muliplayer though?  I was thinking of a more basic
> system("./tpserver-cpp --some --magic --args");

I think it would allow multiplayer, using tpserver-py.

> > > Positives:
> > >   There is a lot of documentation on the website, which is _Very_ nice.
> > > the main problem I think is organisation - most of the info is there,
> > > just not in an easily digestable format.
> >
> > Can you point out where
>
> You have already given two.  That forum post is gold - pure gold.  The
> getting started is the first half, the instructions on how to start the
> server and get the configuration is the other half.

Hopefully things are better now.

<snip>
> Anyway, hope there is some useful info in this post...
>
>       Regards,
>       nash

wrt: Darcs
> [2] As a semi-aside, getting the joy of using another RCS isn't a great
> thing, especially as the darcs home page is pretty poor for info - until
> you use google to search it... Not sure you guys can fix that however.

Blame me for that one. Darcs is nice, and as a scientist, I do understand it. 
I had been playing with it for some of my own projects. If we were to move 
RCS again, I would probably suggest Mercural (hg).

Later
Lee Begg
Main tpserver-cpp developer

Attachment: pgpU1Z32jBlby.pgp
Description: PGP signature

_______________________________________________
tp-devel mailing list
[email protected]
http://www.thousandparsec.net/tp/mailman.php/listinfo/tp-devel

Reply via email to