@Birbo

Quote:

So let's stop the discussion.

Good luck with that!

Dave Lister still posts to Amigaworld in an alternate timeline...

_________________
"Unix is supposed to fix that." -- Jay Miner

@TRIPOS

They might be close to having it working, but it's a massive waste of effort for zero utility.

_________________

@Hypex

Shell-Seg moved to kickstart, it didn’t vanish. Shell-Seg v45 (OS 3.9) and v46 (OS 3.1.4) has been working with OS 3.1 kickstart, but v47 of OS 3.2 does not. Instead it causes a 8700 000F at task 0020E070. Question is whether “rot” is intentional or not - OS 3.2 is littered with undesired “surprises”, even more so than 3.1.4. Sigh.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@Mobileconnect

Quote:

Mobileconnect wrote: @TRIPOS They might be close to having it working, but it's a massive waste of effort for zero utility.

Kernel support is not a biggie actually. It’s just a kernel. And bringing it into user space is not a biggie either, provided that you are willing to sacrifice binary- and scource compatibility. Otherwise it simply can not be done. But say no to the compatibility and it can be done quite fast. And 64-bit and SMP could have huge utility in a true Next Generation context. But if you are mostly interested in running 68k software from the 90’s forever, then it will of course have no utility whatsoever. Depends on what you are looking for. Either this or that. Can’t have it both ways. Either way, PPC is dead with zero future utility.

@TRIPOS

Yes and the scope of making system play nice with SMP, ties into so many things, Messages / MsgPorts are used in intuition, dos, every small tiny or big part. many programs that did not need to.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@TRIPOS

No 64 bit SMP system will run legacy 68K/PPC 32 bit software without a sandbox anyway, so your argument is moot.

_________________

@Mobileconnect

Quote:

Mobileconnect wrote: @TRIPOS No 64 bit SMP system will run legacy 68K/PPC 32 bit software without a sandbox

That is exactly what I am saying…

@NutsAboutAmiga

Quote:

Some programs might crash already when it opens on the wrong screen, it easier if you can catch it the act.

Then that looks like a bad program to use. The original modes aren't there anymore. And there is no planar so there isn't much choice of what mode to convert it too.

@Mobileconnect

I would tend to agree. In the least it could just suspend all 68K apps so only the emulation layer is affected. And the system can go on as usual. Given 68K is emulated at the end of the day regardless that can be controlled. I also think a shadow ExecBase should be in place for 68K. But, they still use $4 on OS4, so a trap handler would need to catch it and see if 68K or PPC touched it which complicated it. I would have depreciated $4 and passed it on program start. I think they do pass the Exec interface so native shouldn't need it.

But, OS4 doesn't know the difference between a task and an interrupt. I was debugging a bad crash that took the system down. The crash log was almost meaningless. The system blamed what ever task was running but that was wrong because the task code wasn't technically unning. An interupt was running. But because the OS makes no difference between the two and doesn't seem to record what context or state it is in it looks up ThisTask and blames it. The crash info and register dump is all corrupted. Surprised they didn't fix that. No wonder device drivers are hard to debug on AmigaOS if the system doesn't store what state it's in!

Getting back to Forbid(), it's hard to avoid it on native code, because they dragged it over to OS4. I don't see any changes to avoid it as the system was never updated to deal with operations that needed it. So even if they avoided the whole problem by sandboxing 68k stuff to its own eco system the problem still remains. It's hard to know what the locking is needed for. In system routines needing it they could try converting to a Cause() so the routine runs in an interrupt. The use cases are varied and not listed in the Forbid() autodoc.

@number6

Quote:

Modepro will inform you of the name of the screen it is looking for. You simply create it and define it as you wish. Done.

I suppose it does as it works live unlike Screens that works after the fact. However ModePro can get annoying if you let it query every unknown screen as I used to in 68K with RTG. But, on OS4, even if it's just on the 68k side, I still wouldn't like to run a 68k OS patch in order to allow another 68k program to work.

@TRIPOS

Quote:

But there are signs that AeonKit is working towards SMP in ExecSG, which is part of their ?System 54? effort.

That's old news. It's been almost ten years, almost, since the X-Kernel was created to do that. Here's a post from 2013 also on MorphZone.

However, in regards to the UBoot mod, it certainly isn't needed for dual core support. My proof is Linux. Linux, unless the kernel is crippled, runs perfectly fine on two or four cores. So I would tend to agree that if they couldn't modify ExecSG instead they likely were doing something wrong. Or wanting to do X so Y could work.

The point he made about doing a MorphOS kernel from scratch is interesting to support multicore. But I thought the low level kernel was that Quark Xpress thing. Which I expected to already support that along with 64-bit able and being portable.

Quote:

Maybe AeonKit can recreate on their own what AROS and MorphOS did two decades ago? Maybe not?

They are! They already wrote 24 commands from scratch Why are they rewriting AROS again? They can write 24 commands from scratch someone already wrote twice before, but they can't write a web browser! Talk about mixing your priorties up! This is the problem with OS4 development, it's all about reinventing the wheel and not inventing anything new.

So now it's obvious why Enhancer has jumped from V2 to V54. V54 is what OS4.2 would have been. They have taken so long now AeonKit are now cloning OS4 and are going to pick a fight with OS4.1 FE. Hyperion will have to deal with both Cloanto and AmigaKit now. They have been taking over their turf while Hyperion was busy else where in court. MorphOS isn't the only competition now.

But will the users accept a clone of Amiga OS4? Again not based on the orginal code but doing an AROS style copy? Another clone of AmigaOS that still doesn't work on x86!

@Hypex

I posted the solution in the other thread. In short - don't delete system objects when they're 'freed'. Then the danger of them 'going away' disappears. Then it's safe to make Forbid a no-op.

Memory leaks? Not exactly, not if they belong to a recyclable object pool. Plus we generally have at least 10x the memory the system was designed to run with these days, and using an object pool reduces fragmentation.

_________________

Amiga Os 4 should stay as is it.
It is in Ben hands? Ok.
What Amiga commmunity need is good open source reimplementation of Amiga gui and graphics.
On top of unix of course to avoid all problems with hardware, drivers, new technologies.

@ppcamiga1

you have it, it's called ARIX. Give it a go :)

@Mobileconnect

Quote:

I posted the solution in the other thread. In short - don't delete system objects when they're 'freed'. Then the danger of them 'going away' disappears. Then it's safe to make Forbid a no-op.

Its kind memory system is it recycles blocks of memory,
let’s say you pool of objects.
Clock program starts.
you won’t send message to “clock”, but clock quits.
objects are freed,
and you start another program that use the blocks.
you then complete message, but object is used by same program anymore.
So, you now have really horrible bug, hard to find, and impossible to fix.

Last edited by NutsAboutAmiga on 24-Oct-2021 at 08:26 PM.
Last edited by NutsAboutAmiga on 24-Oct-2021 at 08:26 PM.
Last edited by NutsAboutAmiga on 24-Oct-2021 at 08:24 PM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@Hypex

>But will the users accept a clone of Amiga OS4?

Well Aros does not run on the Amigaone's which is the target of V2 / V54 and problably Aeon.

Another thing, notice that the Frieden Brothers are within the development team, so their work for Hyperion will halt (?). If V54 does give us new features and Hyperion's OS4 will get no major updates anymore, the choice becomes easier.

@NutsAboutAmiga

If you bothered to read what I wrote, you'd understand that the objects are not immediately reused, are marked 'dirty' for a period of time, and a garbage collector resets them when they're no longer in use.

_________________

@IridiumFX

ARIX is just AROS after little facelifting. It is not worth use.

Next Amiga NG should be just Amiga gui and graphics on top of unix.
Everything below Amiga gui and graphics should be cut off.
No not compatible exec, dos, devices etc.
Nobody need this.

@ppcamiga1

Please install AmiWM on your linux, be happy about that and go trolling elsewhere ;)

@michalsc

From developer pov AmiWM is not Amiga gui and graphics on top of unix.