madshi.net

We've got a complex application here that uses madExcept and Delphi On Rails (http://code.google.com/p/delphionrails/). The application acts as a webserver, serving HTML/JS content to clients and starting a websocket connection for data exchange. It shows reproducable connection errors (webserver not responding to clients at all) that occur only when madExceptPatch is applied to the application (after command line compilation) and we can reproduce the same behaviour while debugging from the IDE when enabling/disabling madExcept.

As we had to insert {$R-}/{$R+} statements to the Delphi On Rails code to fix some IDE exceptions, I suspect there's something going wrong there (I created an issue on their site, too - see http://code.google.com/p/delphionrails/ ... etail?id=7). Anyway, it's strange that not applying madExceptPatch/disabling madExcept solves the connection issues.

Disabling the freeze check in madExcept settings didn't change the behaviour and we don't send bug report e-mails, only create bugreport.txt files - so madExcept shouldn't interfer with network traffic.

So, my question would be: Do you have any idea if madExceptPatch is doing something that could explain this?

I don't think madExceptPatch is responsible, but I'm mostly guessing here.

Try calling PauseMadExcept(true) in your initialization. Does that "fix" the problem? If so, that would prove that it's not madExceptPatch, but the madExcept runtime that somehow catalyses this problem.

Strangely, calling PauseMadExcept(true) didn't change the behaviour.

But we now tried to rollback the Delphi On Rails updates to a prior state - after this, everything worked as it should, so we can now change the DOR code piece by piece and stop when errors occur again. There definitely was an error in DOR (very likely some invalid memory access) and the behaviour was only a very strange side effect of it.

Thanks for your help!

So madExcept appears to be innocent? If so, I'm glad to hear that.