As I indicated previously, we are not prepared to address the kinds of inflammatory issues you are presenting, such as how Tornado is written. When I say "we", I really mean "you". You implied WIND should "eat its own dog food" by developing its software products using its own tools. What in the world does an RTOS and associated tools have to do with the tools it takes to create a GUI like Tornado? Absolutely nothing. You are talking about things you don't know anything about, and it is not helpful.
You also said the unanswered questions on comp.os.vxworks concerned whether the developers at WIND use the Windows 95/NT implementation of Tornado to develop and build the Windows 95/NT version of Tornado. You implied that if they did then their interests were aligned with that of their users and all the kinks should be worked out eventually. This is not what concerned anyone on the comp.os.vxworks.
Most of the work on the GUI was done in WIND's skunkworks in France. Since the French are very abstract programmers, I would guess that the GUI was constructed using a high-level, multi-targeting GUI builder, probably in C++. One possibility is that a single version of the GUI resides in the configuration management system, with pragmas to differentiate between supported platforms. Tornado code in C++ for Windows 95, Windows NT or various flavors of Unix would be generated by the GUI builder by command. In this case only one platform need be used to create and maintain Tornado for all supported platforms. Problems encountered testing Tornado on any given platform, would be handled by adding special handlers (pragmas, etc.) in the GUI builder.
Another possibility, is that a C++ or C version of one platform was created early on and became the configuration managed as a master for such platforms, say for Unix. Once Tornado was working reasonably, the C++ or C source code could have been ported manually to the Windows platform, or vice versa. In this case, there would essentially be two primary versions that would need to be jointly developed.
All GUI programs have similar structure, irrespective of platform (event-driven with call-back routines), but Motif on Unix codes differently than Windows on 95/NT, and there are numerous "gotchas" that distinguish Windows 95, NT and Unix. There are even gotchas that distinguish one flavor of Unix from another, especially System V vs. other now non-standard Unix versions. For these reasons, it is always hotly debated how to handle the differences, abstractly as in the first approach, or concretely as in the second one. Again, given the source of most of the programming, I would guess the abstract approach was used, one way or another. But given that the development started around 1993, the limited functionality of GUI builders at that time may have obviated the practicality of an abstract approach.
But nobody who knows anything cares anything about what platform Tornado itself was built on. The only concern expressed on comp.os.vxworks was whether WIND developers used Tornado on all platforms, not what platform it was built on. Your "eat its own dog food" has no meaning in this context. Tornado can be used for many things, but one thing it cannot be used for is to create another system like Tornado.
But all this is obvious and elementary. What I don't understand is how an employee of Phar Lap would be confused over such trivia. If you are simply sabotaging, why try to do it with nonsense?
I suggest once more that we keep our analysis above programming mechanics. The RTOS space is dynamic and interesting, particularly the struggle for market share unfolding before ours eyes between INTS and WIND, and secondary players like MWAR and QNX. (Sorry, but no one ever suggested that Phar Lap is a secondary player.) This is where we need to focus our attention and energies.
Allen |
