SCI Companion - Language?

[Nostalgic]

I was checking out SCI Companion and, in general, it looks good. (The help file appears to be constructed incorrectly, however. When I download the latest, I get a the equivalent of "404 errors" for each help page.)

Just out of curiosity, what is SCI Comp written in? (Please don't tell me Delphi !!)

It seems to have problems on various operating systems, but I've isolated it to situations where different Windows service packs are installed. This would suggest it's not written in .NET (C# or Managed C++) but it's unlikely to be C++ either, because native code isn't usually affected as much by service packs. I put it into system debug to see if I could at least trace the problem but every debugger I have says the file SCIComp.exe has invalid headers.

[troflip]

It's written in C++.

Re: the help file, it sounds like you're using XP SP2.  Right click on the help file in windows explorer, click on Properties, and choose "Unblock" in the General tab of the Properties dialog.  It should work then.  I'm not sure of a way to get around this.

If you find it has problems on certain OSs, I'd really be interested to know which and what the problem is.  Others have complained about problems this, but no one ever follows up and helps me figure out what the problem is.

[Nostalgic]

It's written in C++.

Okay, that makes sense. The system level debuggers will just dump the assembly then. For some reason when I looked at the interface, I had the idea it was .NET.

Cool ... so ...

Re: the help file, it sounds like you're using XP SP2.  Right click on the help file in windows explorer, click on Properties, and choose "Unblock" in the General tab of the Properties dialog.  It should work then.  I'm not sure of a way to get around this.

I'm not sure how to get around that either. That's interesting. That said, your solution does work just fine.

If you find it has problems on certain OSs, I'd really be interested to know which and what the problem is.  Others have complained about problems this, but no one ever follows up and helps me figure out what the problem is.

Okay, well, here's what I got so far:

SCI Companion works just fine on my WinXP machines. It's on Windows 2000 that the problem seems to come up. I'm seeing that this seems to be most consistent when I have Service Pack 4 installed. As far as what happens, I get this:

"The instruction at '0x77e16078' referenced memory at '0x00000000'. The memory could not be read. I debug that with Visual Studio and I get:

"Unhandled exception at 0x77e16078 in SCIComp.exe: 0xC0000005: Access violation reading location 0x00000000."

I then break the execution there. Since I odn't have symbols for the call stack, all I have is the disassembly. I have attached that to this post as SCIComp.dmp (zipped up). That's the stack trace dump up to the point where I hit the problem. (That's in Microsoft's binary format, of course.)

Hopefully this helps a little bit. Let me know if I can provide more information or provide what I have given you in a different way.

[troflip]

Thanks for the dmp file... it helped, but the fix isn't obvious yet.  It looks like a problem with the controls framework I'm using.  It's trying to ask for the number of items in a combobox, but it appears the combobox actually hasn't been created (hasn't received WM_CREATE) yet, and this causes windows code to crash.

[troflip]

I've fixed the problem, and uploaded a new version here.  It should work on win2k now.  Let me know if there are any problems.

[Nostalgic]

I've fixed the problem, and uploaded a new version here.  It should work on win2k now.  Let me know if there are any problems.

Thanks. I just downloaded it and I'll give it a try.

[Nostalgic]

Well ....

Sort of a good news/bad news thing.

The good news is that the problem as I originally described it doesn't happen. However, when I try to run SCIComp.exe on Windows 2000 now, literally nothing happens. It just doesn't start up. It does still start up on WinXP, however.

I checked Task Manager to see if SCIComp was a running process and it's not. There's no application event in the Event Viewer. So I can't offer much in the way of tracing what's going on, unfortunately.

[troflip]

I don't see any good news in that - it sounds worse than before!  Can you start the program under a debugger and get a .dmp like last time?  Maybe it is crashing, and for whatever reason Windows isn't "catching" it (I don't recall what win2k does when an application crashes)...

[Nostalgic]

Can you start the program under a debugger and get a .dmp like last time?  Maybe it is crashing, and for whatever reason Windows isn't "catching" it (I don't recall what win2k does when an application crashes)...

With just the executable, I can't start a debug process enough to catch anything. The built-in debugger isn't picking up anything during a straight execution of the .exe, which is interesting.

I'll keep trying. I think the debugger isn't picking up anything because it's literally stopping before it really starts, so the debugger can't even hook in. I tried a dissassmebly first just to see if that would give it enough to start with, but no dice.

Again, I'll keep trying.

[troflip]

Which built-in debugger?  ntsd? (is that even included with win2k?)
You could probably use visual studio to run the executable too.
But if the process is just exiting for some reason, then yeah, it won't break in to the debugger.

I just uploaded another version that has some logging capability.  You can specify a log file as command line parameter

Code: [Select]

scicomp.exe /log log.txt
If the log file ends up being empty, then yeah, the app isn't even really starting.

I guess I need to find a copy of win2k and try installing it on virtual PC or something.  Now I'm really curious.

[Nostalgic]

Okay, I tried the logging version of the executable and the log on Windows 2000 says:

"Initializing application.
Couldn't register window classes.
Application exiting..."

[troflip]

Awesome, thanks!  That narrowed down the problem immediately.  I uploaded another version with a fix.
There's probably more stuff that doesn't work on win2k, but only one way to find out.
thanks so much!

[troflip]

I guess I need to find a copy of win2k and try installing it on virtual PC or something.  Now I'm really curious.

I finally got winME working on Virtual PC (I don't have a win2k to install unfortunately).  Many features are broken on WinME.  I don't know if it's worth it to fix.
Maybe I should just declare it requires WinXP.

[Nostalgic]

Okay, I just tried it on Windows 2000 and, so far, it looks to be working fine. I haven't tried every single thing yet, but I'll go through it and respond back if anything comes up.

As far as the WinME issue, my guess is you'll find that on Win98 as well. I don't know what the issue is necessarily but maybe requiring an "NT-based" machine is the way to go (either Win2000, WinXP, Vista). I think SCI Studio -- even though it doesn't say it -- has some flakiness on anything not NT-based as well.

Thanks for looking into this.