OWB Origyn Web Browser - RISC OS

Amiga systems (MorphOS, AROS and OS4) have used OWB as base for browsers. OWB is webkit + sdl.
It is quite easy to remove SDL and replace it with something native. No need for big bloated Qt.
It has also been proved possible to upgrade the WebKit part.
I have tried to build OWB with gccsdk and link it static. I get segfault in the global constructors and I don’t know how to solve it.

Feel free to jump on the train!

http://web.archive.org/web/20100730213400/http://www.sand-labs.org/owb/downloads , the rather old originals.

Interesting. I was wondering what the ‘port an existing browser’ options looked like the other day. Is this browser alive, as in developed? There isn’t much activity on the project GitHub.

Amusingly enough, here is the opposite:

http://eab.abime.net/showthread.php?t=73403&page=23

The MorphOS version is active , supports HTML5 video and more.

BTW , I am trying a dynamic build now.. hopefully easier to debug!

I get a bit further with shared build. It creates a SDL window, but not its internal window.
There is a odd getBCIWindowManager that only seem to exist in header , no implementation from what I can see.
So crash!

Restarted on a bit newer sourcetree .. Anyone have a lot of shared libs (compiled)? I have problems with autobuilder.

Anyone have a lot of shared libs (compiled)?

Otter browser? There’s like a hundred megabytes of stuff that gets loaded by the shared object manager thingy…

If you find the Otter thread in Announcements forum, there should be a lot of pre-compiled libs in the “SharedLibs” zip that Chris links periodically. Note – you probably want to start at the end of the thread and work backwards, as the newer builds tended to be later.

Anyone have a lot of shared libs (compiled)? I have problems with autobuilder.

I would strongly recommend to sort out your autobuilder problems instead of working with precompiled libs. It is very unlikely that you’ll get something working without changing/fixing at least some of the libs.

Which problems do you have? Not that I’m an autobuilder expert, but I would hope that people like Chris Gransden or Theo Markettos could help to sort out problems. Maybe other people encounter the same problems, it is always valuable to provide solutions for such problems in a public forum.

I managed to build all support libraries except for libcurl ( I did one manually , but without ssl).

It still crashes.

Some progress. It seems like riscos doesn’t like static variables in main.cpp
This is problematic since they are used in the event loop. Even more odd, I need to allocate memory manually before giving them memory from webkit/view. Gah c++.

First thing rendered! Text fails with invalid face from ft_load_glyph.

Wow, excellent stuff.

Is Freetype part of your build? Or did someone else build that in the past? Do the font names need translation?

Is Freetype part of your build? Or did someone else build that in the past?

libfreetype6 is in the GCCSDK autobuilder, I guess the Qt needes it.

I have freetype (using autobuilder) and fontconfig. I don’t know why I get the error..
I have a workaround now , using embedded fonts. I feel it is a deadend with the version I managed to build and use. I will use the most “recent” instead. Non of them can be considered new, but seeing what has been done on the Amiga platforms it shouldn’t be impossible to rebase it.

The main advantage to build something out if this is speed. OWB starts in a second (or so) which is instant compared with Otter :)

AND

The goal is to have it more native , not using SDL.

Started with new codebase now…
void* mmapResult = mmap(NULL, BLOCK_SIZE + extra, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANON, -1, 0);

Fails with cannot allocate memory. The same mmap worked in the old version. BLOCK_SIZE is 64k and extra 0 ,

It is well-beyond my experience, but out of interest, where is the source for this browser? It looks a bit abandoned (though I assume you are suggesting updating its WebKit engine?)

It looks a bit abandoned

There is a more recent one for AROS and MorphOS, but since it uses MUI/Zune I decided to start on a version without gui.

One of OWB’s main “selling points” is that the browser engine is completely abstracted away from the UI, to the point where it just presents a series of data structures to the “client” functions which are then expected to do as they see fit. From that perspective, it seems an ideal candidate for a more native RISC OS front end, as that is exactly what it is designed for. As such, I’d be wary of going too far down a UI-kit-specific port (eg. MUI/Zune etc) since you’d likely want to throw away the UI portions anyway.

I think Michael’s original SDL-proof-of-concept is definitely the way to go, albeit with a plan to remove the SDL dependency the moment it starts rendering (so that you can move to a RISC OS native UI). Of course, you have to make it not blow up first, which seems to be the biggest challenge.

Hmmm memalign also fails with the same error. Works from standalone app.
How can I alloc memory in library?

Google…

The sad thing is that memory allocation is a mess.. so I can’t search. But java script is ok , I get suggestions if I start to type in a search.

That Swedish?

My two big questions, now that it gets as far as showing Google, are:

• What’s the memory footprint? Otter was stupid-large to the scale of being “for demo only” on a 256MB machine.

And:

• Is it quick at sorting out its fonts? Otter took a famously long time to deal with fonts. Not just the initial scan, but the moment a site hits fonts.google.com (I tried either Google Drive login, or Bing, I don’t remember now) expect the browser to stall forever (not hyperbole, I had a bath, came back, and it was still stuck!).

What’s the memory footprint? Otter was stupid-large to the scale of being “for demo only” on a 256MB machine.

About 70Megs (google loaded)

Is it quick at sorting out its fonts?

Right now I am using embedded fonts , since I cant find out how to use freetype.

About 70Megs (google loaded)

Oh, way better!

It’s still cringe worthy to somebody who squeezed a simple GIF decoder into 1K, bit I know browsers take inordinate amounts of memory.

Right now I am using embedded fonts

Well, okay, that’s one way of solving the problem. :-)

Just out of interest, how far does it get with this: https://www3.nhk.or.jp/news/easy/

NetSurf, having no scripting to speak of, doesn’t get far.
Otter (0.9.10-dev), amusingly, makes a mess and renders the text as many empty boxes. So much for the lengthy font scan…
QupZilla 1.8.9 as for Otter. Same back end? The rendering looks near identical. Shame, as the latter two manage to display the content, just not the text.

Just out of interest, how far does it get with this: https://www3.nhk.or.jp/news/easy/

That will not work with embedded fonts..

:-) I kind of guessed, but thanks for trying.