Anyone Tried Compiling the latest WebKit?

It's great to see some collaboration with Odyssey Web Browser resulting in improvements.

Going forward, we will eventually need to update WebKit to a far more recent version. Websites using Bootstrap 4 tend to have layout issues, and some websites even end up being blank (e.g., if they use vue.js).

Has anyone tried compiling the latest WebKit? If so, how far did you get? Kas1e's Odyssey build instructions and SDK should help in getting started.

Hans

@Hans
There is progress on morphos side about, and how far they move with it: https://morph.zone/modules/newbb_plus/ ... hp?topic_id=12823&forum=9

In other words, you can't build latest webkit as it probabaly, and you need some work being done on. And i fear that will include some work to be done on SDK and some OS parts too. But that to be seen, once any skilled developer will start it.

Through morphos guys gone some other way , like building webkit as independent part, which they use later via any app they wish (that the right approach for sure), and probably that requery quite a lot of work.

All info in that thread i point on.

What is need it, it someone really skilled to start the work on new webkit and done everything "right". I.e. no cairo, no shmairo, pure amigaos surface and stuff. Quite a work for weeks if not months of skilled developer in full time.

@kas1e

Yes, I saw that thread on MorphZone, and read that they needed to make changes to their SDK and OS in order to make it possible. Jacadcaps mentions NetStack, exec, & GCC specifically, as items that needed updating.

That's why I'm wondering if someone's tried it on AmigaOS 4, and how far they got.

Hans

@all

Since nobody else seems to be doing it, I decided to try compiling WebKit myself. Well, just JavaScriptCore. I didn't get it to build, but the exercise was useful in figuring out what we need. Here's what I found:

1. We need our C/C++ stdlib headers to be brought up-to-date with the latest standard. Especially cmath. Newlib is missing multiple functions (log2(), asinh(), etc.).
2. Newlib also has gaps in its POSIX compliance. It's missing basics such as tzset()**
3. A whole bunch of dependencies need to be ported or brought up-to-date. JavaScriptCore relies on libicu v60.2+. WebKit also needs Harfbuzz, and a bunch of others
4. Source/WTF/ needs AmigaOS specific code. Some things can be deferred till later, because they're only needed for the Javascript JIT compiler. Other things, are needed now (e.g., we have no sys/mman.h, and a cryptographic source of randomness)
5. WebCore has a smattering of platform-dependent code elsewhere (i.e., they haven't managed to keep all platform-dependent code in WTF)

So, as with jacadcaps' MorphOS port, we need multiple people to work on updating various system components, the SDK, and libraries.

We really need a browser update, so I hope there are willing volunteers. I'm happy to pass on my AmigaOS *.cmake files to anyone who wants to work on WebKit. Just contact me via email.

Hans


** Clib2 probably has gaps too. At least, last time I tried switching from newlib to clib2 I encountered problems. Newlib is my preference, anyway.

@all

Also, having looked at the graphics requirements, and I think that the graphics.library (and MiniGL) simply isn't up to the task of hardware accelerating rendering. The amount of transfers back and forth between RAM and VRAM would probably kill any gains from partial hardware acceleration. So, machines without Warp3D Nova will be best off with software rendering via the Cairo backend.

For newer machines with Warp3D Nova capable cards, acceleration can be provided via Cairo-GL.

Hans

@Hans
I too more like newlib, but in our case the only real way are clib2. Working with Hyperion "together and fast" is unpossible, and things may never be released. So realistically only clib2 and fixing gaps and issues in it, together with our gcc.

Other stuff like 3d party libs are no problems. Problem is to find someone who will fast fixing bugs in gcc and filling gaps in clib2, because one person can't hanlde it all probably. And by fast i mean once reported - work start - fixed. Not "soon" , "two more weeks", "report and i will take a look later. Dunno if if we have anyone who can deal with it in acceptable timeframe and throw all other projects on back in favor of updting clib2 and gcc.

@Hans

log2(x) = log10(x) / log10(2)

@kas1e

Quote:

SDK is ExecSG team.

@LiveForIt

Proper funding might attract developers.

@LiveForIt

Let's get the stdlibs fixed instead of creating workarounds. The shortcomings are going to keep biting us, as more and more C++11 and newer code is written. I encountered some of the same issues while trying to build libicu.

And, AFAIK, Hyperion is still in charge of newlib and the AmigaOS 4 SDK.


@Kamelito

Yes, funding might help. However, getting it is a challenge. If I were confident in getting 50,000+ people using the Web Browser, then a Kickstarter campaign would be enough to get the ball rolling.

Hans

@kas1e

Yes, we need an active maintainer for adtools (GCC and the C++ stdlib), and clib2/newlib. Jacadcaps encountered things that needed MorphOS kernel changes, so help from the OS4 dev team may be required either way.

Hans

@Hans

If i understand correctly the guy behind the current development of newlib is salass00 ... perhaps we may start collecting some info to him, so to let him step by step start adding atleast some of that changes we may need later for the newest webkit ..
An update for OS4 is not exactly under the corner, it means theorically we may have some time to improve internals further

@Hans

Quote:

asinh() is already supported in newlib. log2() exists also but only as a macro in "math.h".

@samo79

Quote:

Please do so.

As for sys/mman.h, for example, I simply made override includes for those, and handcrafted my own solution for Jamiga. IIRC it was however only like two-three functions used, and they were simple to implement.
It’s not a long term solution, but perhaps something to make progress. I.e. crests dummy implementations in order to see how far one gets in the compile process.

@Hans and @BSzili

Quote:

I only know the information I read or talk about at amiwest, if some thing has changed, I do not know.

https://wiki.amigaos.net/w/images/6/62/Execsg-team.pdf

Software Development Kit (SDK) is on page 2.

You can blame Steven Solie for disinformation if it’s wrong.

@salas00
Quote:

Probabaly 50% of adtools reports are there already for:
https://github.com/sba1/adtools/issues

Half of them about all kind of errors "not a member of ‘std’" , and some failing cases of "std:xxx" tests. They fall the same and for clib2, and for newlib.

But no one know when newlib will be released, so that also can't help much and those who want to base some big project on, should probabaly consider clib2 instead anyway.

@LiveForIt
You really didn't get a point. Point is anything which is closed sources , be it called hyperion, execsg team, amiga20004040040 mean no real progress can be made, because its not open sourced, and not everyone can works/test it, and what is most important no one know when it will be released.

As for SDK, noone works on it right now. I never see any sign of live about it in last year or more. So, forget it as well.

@Hans
Seeing history of fixes for kernel, and that none of them was fixed when bugs was reported, but after year(s), it mean it still unpossible to do anything good by this way.

The only way to progress, its use all opensource components (clib2, adtools, 3d party libs), and when time come to something closed source, trying to workaround it as much as possible ,without needs to ask for fixes in kernel, or anywhere else which may never be released.

I do not know if there will be any sane person willing to spend their time on something, and in end meet a wall calling "soon to be fixed, just wait". Naah, deals can't be done like this. If one want browser, it mean everything should be open as maximum as possible, and someone in charge who will help should be adequaete enough to shift prioritizes. There should be just some person(s) willing to do things adequate, fast, with everyday works.

We already report a bug Jadacaps found, which need to be fixed : https://github.com/sba1/adtools/issues/82

Did you see anyone even asking anything about ?:) I am not. So how we can progress ?

@kas1e

C++ headers are provided with the compiler and not newlib/clib2.

Simply saying that something is missing from std:: namespace is not helpful for me as this is a C++ thing and not something that newlib (or clib2) can provide directly. More helpful would be to know what c library functions are missing that gcc needs in order to provide said feature...

@jaokim

Quote:
How about we get that into our stdlib headers? that way it's there for everyone to use, instead of every single developer needing to work around missing bits

@LiveForIt
I'm sure that Steven meant an SDK for ExecSG itself, and *not* the AmigaOS 4 SDK.

@kas1e
Yes, we know that Hyperion can be slow, and we know that we need active maintainers of adtools, etc, or we'll never get an updated browser, and more and more websites will become unusable to us. You're repeating the obvious.

I restarted this thread to call people's attention to what will be needed, with the hope that the goal of updating our web browser will motivate some action.

Hans