>>>>Hi Herman,
>>>>
>>>>It's interesting that you've delved so deeply into this, and still the general case seems to be too difficult to handle with generic code. That strengthens the argument for making the compromise of adding some intelligence about the specific application that one is launching. For example you can easily get past a misleading splash screen if you know the window class name that you're waiting for in launching that particular application.
>>>>
>>>>Mike
>>>
>>>Hi Mike,
>>>Thanks for the comment :)
>>>
>>>Yes, I'm agree with you. Eventhough maybe we can solve most of the problem with generic code, But it's just going to be more dirty and dirtier. At some point maybe we could crash the apps or even the system, I'm not sure about it. So looking from the situation, I prefer to code in a clean way.
>>>
>>>If you check for the class name it will get confuse because many apps had the same class. For example VFP7, from the first time we launch VFP, it already has more than 2 hwnd with the same class (VFP77400000). Which one is the right one. I mean besides the parent. And also like you said it only works *if* we know what is the class name.
>>>Actually, if it is only the splash screen we could pass this. Just check for the application name (exe) using the hwnd. But it also works only *if* it is only one apps at one time. Well..I'm not sure :)
>>>
>>>Regards
>>
>>Herman,
>>
>>I think you understood me, but just to be clear, I used the window class name only as an example. For some applications, a known value of the main window class name would do the trick, for others, you'd need some other identifying characteristic, e.g. a pattern match on the window title. In general, I'm saying that with the help of such meta-information about specific applications or categories of applications to be launched, it may (in most cases) be possible to get a handle on the target without too much trickery. Another thing would be to know the preferred way of launching each app, when there are multiple alternatives, e.g. knowing when CreateObject/GetObject is available. If the target app is VFP-based, and if one has control over its design, this should be the best possible scenario.
>>
>>Mike
>
>
>Mike,
>
>If I understand you correctly. Do you mean that we already have all (most) of the information before hand ? And if the apps has the control, does it mean that we have a limitation ?
>Please correct me if I'm wrong

Herman,

Well, now I'm not sure exactly what you are asking, so I'll just try to clarify what I meant. Yes, I'm saying that we have some additional information about the application before we launch it, but not necessarily information that can be obtained in a generic, automated way. That why I called this a compromise, or if you wish, a limitation. We could be somewhat heuristic about it, using metadata-driven "intelligence" to make a reasonable guess as to the category of application to be launched. But in general, one would need to allow for a way of configuring the launcher for the known characteristics of any particular application, and for launching it in a particular way. Using a tool like Spy++, for example, one could determine how to detect a specific job and configure the launcher accordingly. So yes, I'm making some big assumptions about the nature of this facility and its inherent limitations. As it happens, I'm about to make a public beta release of such a product.

Mike