Compiling tpserver-cpp under windows – Part 2
Well it’s finally done. I managed to get tpserver-cpp to compile under Windows. I just need to fix a few things and then I’ll upload “unsupported” binaries.
Well it’s finally done. I managed to get tpserver-cpp to compile under Windows. I just need to fix a few things and then I’ll upload “unsupported” binaries.
I had hoped that compiling tpserver-cpp under windows would be easy. Turns out I was very, very wrong. I am even going the easy route of using Cygwin to handle all the select/signal/pthread stuff which wouldn’t work easily under Microsoft products.
Compiling it was relatively easy once I had all the dependencies installed. Cygwin comes out of the box with guile (1.6 and 1.8) which was the only dependency I was concerned about. (Everything else is pretty standard.)
However, this is where I ran into a problem. The server uses shared libraries to contain important modules like game rule data and persistence. This doesn’t work to well under Windows. At first I should it should just be an easy step of convincing the autotools to build .dll versions of the modules. Turns out dll’s aren’t at all like .so files. They have to have “no unresolved symbols” which makes it very hard to do what tpserver-cpp does (IE modules use the core functions in the main server like the logger).
Next step was to find out what other applications did, I found this library made by libtool guys called ‘libltdl’ which lets you “fake” dlopen stuff. So I “ported” tpserver-cpp to use this instead of just a raw dlopen. (This should also make tpserver-cpp more portable to such weird operating systems as BeOS and HPUX.) Dunno if Lee will like it or not 🙂
Still not done yet but it’s looking much more hopeful.
It would be nice if you could say “swap out these applications at a last choice”.
The main reason for this is that I use the deskbar application quite a lot. I would really like the system not to swap it out so that it is always really responsive.
Searching google doesn’t turn up anything all that useful. I did recently find memlockd which would be quite useful for Servers.
Here are a bunch of things which I would love to see added to gnome to make it much easier for me to use.
This is needed because when the computer gets loaded the mouse interprets the clicks as click and drag. It is also a problem were people aren’t skilled with the mouse (IE hold and drag the item). The means panel ends up getting dragged about. I have the same frustration under older versions of windows.
Everything else seems to display useful information. The CPU applet displays the current CPU speed. The laptop temp applet displays the current temperature. Why doesn’t the network applets show the current configuration?
Recently the 512MB stick of ram in my laptop died. This means that I’m currently restricted to 256MB of ram. Gnome positively sucks with this much ram. This is annoying because XP runs quite nicely with this amount.
Evolution and Epiphany are the to biggest hogs but all the other little things really add up. Disabling the SPAM filtering in Evolution seems to have make a fairly big difference.
Not to mention my video card has to steal main memory (It’s a Intel i855, I have to restricted to 32mb).
To make matters even worse I have a very very slow hard drive (4200rpm, 1.5 Inch drive) this makes swapping even more painful then most people.
My hard drive seems to do a lot of work even when not swapping. Doing a “echo 1 > /proc/sys/vm/block_dump” shows that a lot of things are accessing the hard drive when they are idle or not doing anything.
The hard drive is one of the biggest consumers of battery life in the laptop. Reducing the accesses would make everything faster. I guess this is counter reducing memory usage however.
Firstly, every LCA (linux.conf.au) I have been to has been really brilliant. I’ve been to 6 other conferences and LCA is my definite favourite. I think that for the conference to be this consistently good, there must be because of a huge effort by the organisers and LA.
I’m interested to find out if any of the following suggestions have been considered and why they have possibly been discarded? I’m sure there are obvious reasons I have missed because I have never been part of either the LA committee or the LCA organisers.
A possible “People Choice” track. This would be a track were the talks are selected by the LA members directly. The biggest problem with this type of selection is that it can rapidly turn into a popularity contest. I think this would however be a good way for the LCA committee to check that they are actually providing what people want, by looking at what people are actually voting for.
Why are the submissions not publicly accessible? Having public submissions would have numerous benefits,
The proposals had to be submitted under an Open Source license anyway so that should not be a problem.
From people’s blogs I know that you only get about 5 minutes per submission. This doesn’t give you much time to write any type of feedback. However, I think a big part of any learning process is getting feedback on your work. A good start would be a set of check boxes which say something like:
On another feedback note, during the conference it would be nice if there was a way to provide feedback on the talks you attend. I seem to remember there being something similar to this in the past but my memory is failing me.