hello

aaaaa

aaaaa

aaaaa

Ok i see what you mean by that now. The main problem i see with doing this, is that wimp application can also call other system functions, and shared code, that are a priory not protected for sharing, and doing this without using the wimp event system.

Of course, the switching system must verify that you’re coming from an application code… If you’re inside a SWI call you should not stop it.

I saw the thing that the guy who wrote wimp2 wrote, he say that for example, he still needs sync for filesystem access, and for certain things, but wouldn’t that be a bit dangerous to just enable polling like that without protection ?

No, if you use a white list of applications that are working.

Nota: I suggest not to rely in Wimp2 source, since it’s GPL. So no way to integrate it inside the Wimp source code. Timer could be change for an HALTimer too.

Specially that with that kind of things, some bugs can obvious ok, some can be really harder to see, or happen only in very specific circumstances, and can be very hard to track, specially if it has to do with filesystem or other important system function, it can lead to serious problem.

If it does not work, you blacklist the application. I don’t see Wimp2 as a ‘PMT for all’ but as an accelerator for some CMT applications + a framework to make PMT compliant applications.

To me the problem is not that much that some application can’t run, but to be able to determine with certainty which application can be safely run without doing any trouble at all, and what can be done in asynchronous event handling.

No problem with PMT compliant applications. For the others, tests are needed, and a whitelist of applications should be written. Then you just need to launch the compliant applications with a Filer_Run2 command. That’s the way Wimp2 worked, and that’s OK for me. ‘Just’ need to test the 2-3K 32 bits existing applications :)

Then CLib and other libs (and BBC Basic, of course) could be tweaked to make only PMT compatible applications.

For modules and system code, PMT is not really needed, as the code is optimised. It’s not always the same with application code (if it was, PMT would not be necessary at all)…

At first, Wimp2 only give you PMT for specific ‘PMT compliant’ applications. The possibility to run a CMT application under the PMT system was an option, coming from another tool.

So two parts:
1/ a PMT system, that – of course – only apply to application code. Easier to re-make, as we now have timers in HAL.
2/ a tool to make a CMT application running as a PMT one. With a whitelist of applications. Lot of tricks can be applied here to adapt some existing code to PMT mode. A kind of Armalyzer to search for non PMT compliant code could be done.

You suggested to implement CMT as a PMT task. Wimp2 embed a PMT system inside the CMT one. Better way to do it, IMHO, as it’s a lot simpler. With such a tool, a lot of software could use threading (for example Netsurf). Of course, it will need the use of some new libs.

aaaaa

Would there be an easy way for a task switcher to know if a task is into an swi routine ? Are all swi routine code stored in a specific memory area ?

I don’t know…

The thing is, what criteria do you use to put application in white or black list ? How can you know for sure the application is not doing anything unsafe ?

Tests (tools can be cool too). As much as possible.

If you don’t have a way to figure out those unsafe condition in reliable and systematic manner, it’s not a very good idea.

Why not? If 100 people test an application without problem for months (let’s say Edit), you can say it’s safe. Of course, no way to be sure. The same with Wine under Linux for Windows emulation.

With a buggy application, you can choose to patch the PMT system (there are a lot of tweaks in Wimp2… but GPL), to modify the application or to blacklist it. Not a big deal.

And bugs in that can really lead to serious bugs in the whole system.

Bugs always do this on RISC OS :)

There are two choices: something not perfect, but soon, and something perfect, but not available, never. Let’s face it: we speak of CMT for almost 20 years, without results. IMHO, Wimp2 was the only step in the right direction.

The same with SMP. AMP would be much simpler, but no one wants to consider it. Result: we still use only one core.

aaaaa

The problem is that they are bug that can go undetected, no warning, no check, the application cannot detect the condition, and then it can corrupt a file, or screw up something very badly.

One more time, that’s always the case, even if here, making a CMT application working in PMT mode is source of new bugs.

Let’s face it: there is NO way to convert CMT code to PMT. Some will work, some won’t. Tests made by users can be a good way to solve this problem (it is for Wine, so why not for us?).

It can be better also not to rush into having something half working that will never be able to be fixed ever because it introduce a bad design too early.

IMHO, it’s better to have something half working today, and a path to true PMT applications for tomorrow, than the current state: nothing. Do we have an option for a better solution, doable in the next six months?

1/2 > 0

Yes like 0 = 0 is more like equation that you get in 15 years,congrat, impressive results on all level.

I’d rather get in touich in justin and try to see if something can be worked out with him rather than trying to deal with your lot lol

At least he seems smart, honnest and competent.

“You think Riscos is not perfect, and want to change anything about our good old way, you must be insane, your need help …” , i think i already heard that kind of things in cult ..

The whole future is more based on the promises of marvel to come coded by some miraculous genius, other than this well ..

not interested thanks.

Replying to your own posts seems a little odd.

Wibble.

Well if you think you don’t look odd well, i guess need to keep a good commercial facade on the internet, or at least trying to convince yourself you are doing a great job at it.

I’m not agreeing with the critical mass of users, or not singing in hearth at the great promise of the future of riscos, it must look a bit odd to you

Each one sees the oddity of the other, some are just more tolerant to it than others ..

Some move on , some stay on 10 year old conflict and stay 15 year old in the past. Always been the best way to get result right.

One can see easily who had the idea and done them or tried, and who has not :)