@Hypex

I haven't read deeply into this but from the link posted on the first page it seems that it even if the compilers supported that functionality it could still not work because it's the code the JIT compiler produces that is endian dependent.

_________________

José

@Jose

You are another guy who made the false assumption we have to rewrite and find all endianess issues anytime it will be released a new Javascript version...

No consortium rewrites whole engines from scratch at any new release.

Once we fix Javascript endianess issues the most is done...

Then we just need to monitoring any new releases of Javascript engine and check just the miinimal modifications they could add and that need our attention.

_________________
"When the Amiga came out, everyone [at Apple] was scared as hell." (J.L. Gassée, former CEO of Apple France and chief of devs of Mac II-fx, interviewed by Amazing Computing, Nov 1996).

@Raffaele Quote:

Once we fix Javascript endianess issues the most is done... Then we just need to monitoring any new releases of Javascript engine and check just the miinimal modifications they could add and that need our attention.

But isn't that the root of the problem? Otherwise surely we would still have an up-to-date JavaScript engine?

i.e. It's asking quite a lot for someone to check everyone one of the (presumably insanely frequent) updates to the JavaScript part of WebKit, especially without any further remuneration...

Last edited by ChrisH on 28-Feb-2016 at 10:04 AM.
Last edited by ChrisH on 28-Feb-2016 at 10:02 AM.

_________________
Author of the PortablE programming language.
It is pitch black. You are likely to be eaten by a grue...

@ChrisH

The whole thing is blown out of proportion. It's not like Odyssey merges WebKit after every single commit they make. Also nobody has actually asked the WebKit team whether they'd accept the big endian patches upstream. It's almost as if people wanted to keeping things as they are, so they can keep moaning about how outdated Odyssey is.

Last edited by BSzili on 28-Feb-2016 at 10:24 AM.

_________________
This is just like television, only you can see much further.

@BSzili

As always, you spoke words of wisdom! You amaze me!

_________________
"When the Amiga came out, everyone [at Apple] was scared as hell." (J.L. Gassée, former CEO of Apple France and chief of devs of Mac II-fx, interviewed by Amazing Computing, Nov 1996).

@broadblues

What I was thinking is that in the OS4 case the GCC compiler is customised. For example the APICALL directive to specify an OS4 function call from an interface with compiler magic. So all we need is another directive to specify endianess. Yes not a big job at all!

I don't think macro magic could solve this cleanly or easily but it would be nice.

@Jose

If that was the case and JIT was causing an issue then we should rip the JIT engine out of TenFourFox which is PPC optimised.

@Hypex
Quote:

What I was thinking is that in the OS4 case the GCC compiler is customised. For example the APICALL directive to specify an OS4 function call from an interface with compiler magic. So all we need is another directive to specify endianess. Yes not a big job at all! I don't think macro magic could solve this cleanly or easily but it would be nice.

From what I've read, there might be some PPC instruction magic too. Here is a quote from the book "PowerPC Microprocessor Developer's Guide":

"Several features of PowerPC processors permit them to use either mode of memory access. A bit in a special register (in HID0 for the 601, in the MSR for others) selects the processor's memory access mode. It is possible to address memory opposite the current mode by using the byte-reversed load and store instructions (lhbrx, lwbrx, sthbrx and stwbrx). These instructions allow, for example, little endian storage access while running in big endian mode with no performance penalty."

If might be interresting to compile a simple program to assembly output, change the load/store instructions to work in little-endian mode and assemble it to an OS4 binary. I wonder if it would work.

_________________
X1000 with 2GB memory & OS4.1FE

@BSzili

Quote:

Also nobody has actually asked the WebKit team whether they'd accept the big endian patches upstream.

I have asked them. They don't care about PPC.

@ChrisH
Quote:

Now perhaps someone can tell me why this crazy idea wouldn't be possible?

Multitasking? The OS would need to know what mode to be in when it switches tasks. Maybe it would be possible if little-endian stuff was running in a seperate CPU core but I can't imagine switching modes would work unless it's somehow supported at the OS level. However, the special instructions mentioned in the quote I included in my previous post might be useful.

_________________
X1000 with 2GB memory & OS4.1FE

@Raffaele

Quote:

Endianess JavaScript Issues Bounty now exceeds 500 bucks!

According to this:

Quote:

... Deadwood Odyssey Web Browser project ... Donate to let developers being awarded for solve all endianess issues that afflict PPC Odyssey Web Browser Webkit Javascript Engine.

The bounty is about the Javascript Engine in the Odyssey Web Browser.

There I some annoying points:

1. I never saw an error that was connected with endian problems, so could someone please give me a link to someone knowing this stuff?

2. I never saw any comment from a developer (I knew working on Odyssey) about this bounty, could I get a comment from Deadwood or any one else working on Odyssey?

@ChrisH

I see that you have put some thought into this endian-ness problem; so have I, over the years (I was started my computer life with DEC PDP/11s, so I was brought up on little-endian CPUs).

This is not the place. Let's start a thread on Hyperion's forum - I suggest the "General AmigaOS" forum.

_________________
cheers
tony

Hyperion Support Forum: http://forum.hyperion-entertainment.biz/index.php

@wawa

Quote:

arturs sdl version is the customary amiga version, maintained since long and tolerated by the netsurf developers so far even if they want "native" implementation. this is do to say "quick and dirty" port. but requires less maintenance. chris has sone reaction version for os4 and only latels started to experiment wit porting his frontend to amiga. he has encouraged the people to help him do it but there is obviously none left, who would be interested or skilled enough for this. despite claims that "native" port will be faster and consuming less resources, because sdl is "slow and bloated", as the urban legend goes, last time i tried chris' port was muchg slower. this might have improved. but also plotters need to be maintained, which is not the issue with the sdl version, which is obviously portable and always as good as sdl library on a particular platform. all in all i think it would be worth to help chris. one of the reasons being the same binary would be suited to run on genuine amiga gfx and on rtg. but probably even better would be to have a mui/zune version of this frontend, suited for all platforms, as reaction is almost os4 exclusive and as well as depreciated on other platforms.

Thank you for the explanations, now I feel like I understand

Quote:

there is itix mui frontend for mos derived from rather old state of chris' work, i started to look into. alas it occured too much for me, differend codebases combined, a new exclusive build system, lacking programming experience, not to talk about, that it would be third version of a frontend, so i have given up.

I have been trying to compile Chris' version (as far as I understand...) but compiling and setting up the cross-compiler, the libraries, and so on, is such a pain... Still, hopefully, I will succeed and maybe will find a (legal?) way to share my installation in a VM so that other developers don't have to go through all these painful steps...

Cheers!

_________________
Tygre
Scientific Progress Goes Boing!

@tygre

Would a fully functional prebuild+source archive of the required compiler tree help?
Alongside version information and package listings from a working cloud hosted cross-compiling build machine help with this?

I had to deal with all this and still need to replicate it using scripting.

I'm currently looking into making a personal development server for use at home as a CI build bot machine.

@tygre

chris will certainly help you setting up build system, its not that difficult. it was me being dumb and trying to get there from all different ends at once. id try to help with it, just im too inexperienced and engaged with aros68k for now. under no circumstances id need to fork or workaround. what is important on this kind of project is to be able to contribute and push the patche upstream, so you have to engage wit the team to an extent. chris did, artur to an extent, but everybody else who condiders to join should consider this in first place i guess.

Last edited by wawa on 29-Feb-2016 at 01:27 AM.

@radzik

No problem they are not interested.

Once there will be an existing fix for Javascript then we could commit the changes to Webkit repository.

As far as I remember Webkit is still open source, or not? They can't refuse hosting working code for PPC platform...

Else we can create a fork...

Even those in Webkit team who are X86holics, must accept the fact that there is still people developing for PPC.

_________________
"When the Amiga came out, everyone [at Apple] was scared as hell." (J.L. Gassée, former CEO of Apple France and chief of devs of Mac II-fx, interviewed by Amazing Computing, Nov 1996).

@OneTimer1

All I know about Javascript endianess issue comes from various threads that explain Fabien Coeurjoly can't release MorphOS OWB 1.25 because he is facing a very high number of problems due to last version of Webkit Javascript.

Also there is Deadwood thread on Aros-Exec.org where he launches the idea of the bounty for resolving Javascript issues.

http://aros-exec.org/modules/newbb/viewtopic.php?post_id=97679#forumpost97679

_________________
"When the Amiga came out, everyone [at Apple] was scared as hell." (J.L. Gassée, former CEO of Apple France and chief of devs of Mac II-fx, interviewed by Amazing Computing, Nov 1996).

@tonyw

Quote:

This is not the place. Let's start a thread on Hyperion's forum - I suggest the "General AmigaOS" forum.

No thanks, I don't frequently OS4 specific websites, AND this topic is not limited to OS4.
Although, in the future it might be.

A quick question, if one of the OS' supported Power8 what would then be the best solution, using big endian code by default or trying to support code with either endian mode?

@radzik

That something we already knew. The question is whether they'd accept the big endian patches upstream.

_________________
This is just like television, only you can see much further.

@Raffaele

Quote:

No problem they are not interested. Once there will be an existing fix for Javascript then we could commit the changes to Webkit repository. As far as I remember Webkit is still open source, or not? They can't refuse hosting working code for PPC platform... Else we can create a fork... Even those in Webkit team who are X86holics, must accept the fact that there is still people developing for PPC.

well, it is a problem if you need to maintain the code with your changes spread all over the source. guess why there is usually simply ports of some version of certain application but no updates, firefox/timberwolf being a major example. because its pita to maintain it, especially when your port is outdates to begin with.

the situation would be easier, if one were just to separately maintain some frontend while there would be no necessity for major interventions in the core. unfortunatelly its not the case here, as it looks like.

and the webkit people owe us nothing. open source doesnt mean that they have to accept to host on their repository your porn ocllection or whatever you would throw at them. there is certainly project managers and maintainers and you need to enterer into an agreement with them in order to receive commit privilieges. so you will have to fork, as it happened with odyssey or timberwolf. if something else was possible it would likely have been done.

Last edited by wawa on 29-Feb-2016 at 10:28 AM.
Last edited by wawa on 29-Feb-2016 at 10:21 AM.