It's been a busy few days with interviews. I'm hopefully getting close and if so, then it'll be nice to get things back to normal. I'm going to work hard to find something interesting to do today to keep my mind off this stuff... I sure hope it works.
Today I got hit by a Bidwatcher bug and decided to fix it. This program is really a mess of code. The UI isn't really great - it's OK, but it's not Wow! and the internals are a mess - pure and simple. What might be nice would be to gut the entire program and start with GTK/GNOME and build another that has all the functionality but without the limitations. TOday's bug was realted to the parser of the HTML received from eBay. Now this should be handled in as clean and generic a manner as possible. Anything less risks real problems when eBay changes the format even a little. Well, they did, and it broke the code.
So I fixed it and while it's working it's still a patchwork of a bunch of iffy at best code. The parser needs to be rebuilt, the strings need to be dynamically allocated and based on something better than char *, the object model is iffy... let's face it... This code has simply evolved and there's a ton of parts that are dead code - not used by anything, but still there. Oh well... there's a lot that Bidwatcher could benefit from.
Maybe if I have the time I'll start with the object model and then work in. It's just not obvious what to base the code on... data structures and good C code, or try object orientation and do a much better job. If the latter, then what to do about a really good string class... maybe GTK--? In any case, that's what I did today and it's working again.
Interesting thing happened this morning - I finally figured out what was happening and why the runs were taking so long. If the user put in single quotes in the quip, then the command line gets messed up and the default time-out is set to infinity. So... if there were quotes in the string, then most likely we'd have a problem. This explains everything... the logs, the frozen processes, everything. Nice.
So I fixed it in two ways - make the default timeout 20 sec. in the code and make the calling of the code work properly when the quip has special characters like quotes and slashes. The first was easy. The second wasn't too bad after I dug a bit into the PHP manual. There's almost a single-function answer to this problem, but it was a very easy two function call solution. Nice.
So now we have the quip working properly for the unusual data of quotes, etc. This is really nice as it'll probably save me a ton of grief on locked processes. Good to know that the possibility of being hacked has also returned to near zero.
Oh... I checked this morning and sure, enough, we're on Google and I can search (and find) my pages. Very nice. I've done several searches in Google and sure enough, I'm there in a big way. I love this game!
Well... this morning has been very interesting. I saw again that I had several copies of the quip solver going, and I was amazed to see that I had about 15 copies going full-steam. Several of them looked as though they had been run from the command-line. This seemed hard to believe because they all were run by nobody and yet they were run with options that my PHP code would never have generated. Very odd...
So I got to thinking - could it be that someone came in as nobody and started running it from the command line. Or someone could have built a new PHP script, placed it in my system and run it. So I checked last to see who's logged in - clear... then I searched for all possible PHP files - none that shouldn't be there. Very odd...
Then I noticed that the CPU usage didn't add up. In Solaris 7 there's an applet that shows CPU utilization and disk usage and it showed at least 20% usage when a ps -ef | sort -r -k 4 didn't show anything nearly that high. So I got to wondering if there might be a connection - in any case, things were looking very, very odd on sparky. So I did a reboot.
It took a long time to shutdown - there was clearly a serious problem with the machine. I then got him back up and saw that GAIM posted a new version that supposedly fixed a bug in the code that caused problems with X servers. This could certainly be part of the issue. So I got the latest GAIM and built it. The CPU utilization seems more under control and I can't imagine why the other quip-related stuff is happening. Maybe I'll put in a trap in the code to long who, what, when and then have a look at that. I can't imagine who might be doing this, but I was really surprised that I had all these quip solvers running when I know that they naturally timeout after 15 sec. Something must have been doing something kinda odd.
I also think I got crawled this morning. I looked at my web access log and there was someone trying to access robots.txt which is a dead give-away. Interesting... I've been hit by a search engine. It'll be fun to look in a day or two to see if it's been updated to Google or some other search engine.
Well... I found a little problem in the 0.11.0pre13 version of GAIM and posted a question regarding it to the SourceForge board. Eric responded quickly saying that it was indeed his fault. No biggie - it's just good to know that it isn't me.
Also, I've added a logging feature to the quip solver to hopefully fill in a few gaps in who's doing what when with these quip runs. Now I don't mind that someone wants to run the solver, that's why I put it up there, but I do want to make sure that someone is not trying to hack into my system and do serious ill.
There are a lot of people looking for a VC++ IDE for platforms other than Microsoft. I've gotten another one today for GTK/GNOME, and while it looks OK, I don't quite understand the operation well enough, and there's not a lot that it can do for me that I can't do in the tools I already have. Let's face it... gvim, make, and ddd are about all you need to edit, build and debug applications. Sure, when you have a GUI, it's nice to have a GUI designer, but that's what Glade is all about, and I've got that too. I guess I've to the point that fancy tools don't make a lot of difference to me - I've learned how to cope without them. And for that, I'm thankful. While I like a well done IDE, a bad one is a pain to slug through and doesn't help the development process one bit.
While I'll use a good IDE when I'm offered one, I find that Unix with gvim, make and ddd/gdb are all I need to do even the most serious development. Anything more is nice, if it works, but this IDE today was clearly not in the working camp. I'll keep it for now, and maybe play with it more. Who knows, maybe it's time for me to get involved with this Open Source project and fix some of the problems...
Last night I was down in my office and saw that someone was having at the quip solver. I also realized that there was a problem when the quip solver would take too long to return it's answers. So this morning I decided to fix it. What I did was allow for an optional time limit argument to the program and then check with the solver routine after each word. It's not precise, but it is very reasonable. I then put the default of 15 sec. into the PHP3 code and tested it.
Great! It works fine on the example because it takes only a second or so to run that quip, and on simpler quips the code runs and returns a ton of responses. In either case it doesn't hang the browser, and the user will always get a response. There's also the point that I won't have a run-away process anymore - which is good news for me.
It was fun... that's enough.
Interesting thing happened this morning - GAIM/Oscar started working again. I saw that the TOC connection had been paused by AOL and at that time I decided to try the Oscar connection again. It worked. What can you say? When it works, I'll use it, when it doesn't, I'll use TOC. Goofy stuff... AOL. Really goofy.
I was looking at a finite element modeling package today and realized that I didn't like the fact that sparky didn't properly view PNG files. So I spent most of the rest of the day digging into Plugger and it's associated configuration and helper apps. Plugger is really just a configurable plug-in stub. I still need the programs to view the PNG files, but Plugger allows me to put what that program is in it's configuration file and handle all the communication with Netscape for me. Nice, but not terribly wow because almost all apps can be put in directly on the Netscape Helpers page, but this is nice in that it handles it for me.
Anyway, I got xli because I had not xloadimage and I got several viewers, a few movie viewers, and even a MIDI player. All very nice, all work fine with Plugger and now I have not only working PNG viewing, but also Quicktime movies and MIDI files. Very nice.
Well... it seems as though AOL is doing some kind of blocking once again. So I got the latest GAIM 0.11.0pre12 and built it. No dice. The word from the GAIM folks is that only the CVS repository is going to be working. I guess as long as AOL can keep changing things the GAIM folks want to keep the changes in CVS and not generate a ton of releases. Too bad... it'd be nice if they could simply encapsulate the required code into a network resource. Come to think of it... they do. There's a PHP-based server that's dishing out the MD5 checksums but still they need to have code changes. Oh well... I'll stick with the releases because I'm a little conservative.
It's interesting to think of it... AOL is giving away their code. If I were using their code, I'd be OK and ready to go. But that I'm using someone else's means they want to shut me out. Very interesting... AOL will have to give in eventually. There's too much for them to loose.
Interesting thing happened this evening. I was down here catching up on some reading and I noticed that I had two cryptoquip solvers running full-tilt on the processors with very simplistic quips. I think it's clear that someone is running a quip but they aren't aware of the search space issues and so they're getting lots of possibilities. Interesting that someone is running this, too bad that they're not running it well so as to get a good feeling about the results.
I've added something to the page to say "try hard examples" to keep people from putting in bad data and getting too much results. We'll see how it goes.
This morning I found out something that I hadn't expected to see... that I was probably running both Sawfish and Enlightenment as window managers. I know think that my problems a few days ago with the theme selection were related to this. What I first noticed is that GNOME 1.4 wants to use Sawfish, and that I had Enlightenment as the default in GNOME. So I switched to Sawfish as the default and hopefully things will go much better now. There isn't a significant difference in the window managers - from what I see, but the integration with GNOME certainly tips the scale from my old favorite Enlightenment to Sawfish.
I also was reading a lot in the newsgroups and thought I might be able to catch some reasonably intelligent conversation about SGI and their future. But alas, when I went to a few intelligent-sounding rooms I was greated by the same banter that I've heard time and again in chat rooms. It's too bad, too. Because while I have time to do this, and interesting things to talk about, I can't find an interesting person to talk to. Oh well...
While I was messing around with xchat today I thought I might as well get the latest stable release - 1.6.4 as they are working on 1.7.0 (unstable) now. No biggie, but I think I'll try once more to get them to incorporate my perl fix so that I don't have to hassle with putting it in by hand each time I get the code.