NetSurf 3.4 released!

[chris]

NetSurf is a small, fast CSS capable web browser for AmigaOS and other platforms.

NetSurf 3.4 features many optimisations to improve performance over previous releases. It also contains many bug fixes, including improvements to page layout. This is also the first release to contain the Duktape JavaScript engine. While our JavaScript bindings have seen a lot of development for this release, JavaScript remains disabled by default as the support is incomplete. We recommend all users upgrade to NetSurf 3.4.

*** Download (AmigaOS 4) ***

Additionally I have released a beta version of an AmigaOS 3.5+ build based on the AmigaOS 4 native frontend.  I need help to improve this.  Download (AmigaOS 3.5/9)

[chris]

How is the speed of this "beta version of an AmigaOS 3.5+ build" on an Amiga 4000 with CSPPC (MC68060@50mHz/PPC604e@200mHz; 128 mB ram) under OS 3.9/WarpOS 16.1?

No idea, try it.  It should run OK.

Is an PUP/WarpOS version available or planned?

Nope.

If my hardware is sufficient, how can I help to improve Netsurf?

Report bugs, fix code

Chris,  since it's you're now able to setup WinUAE with a CSPPC I would love to see you compile builds for WarpOS too. I'd be glad to test them for you.

That's not the problem, I need a compiler which can build for WarpOS and have absolutely no idea how to set one up (and I really, really, don't want to have to set up a third build environment to do this - ie. I'd like my gcc to build both 68k and WarpOS code if possible).

[chris]

Chris: Test build is still unavaillable.

Turns out Virgin Media have closed down webspace without telling me.

Try here: http://cy2.uk/netsurfos3

That could be an old or broken version, I didn't look.

[chris]

O.K. - I downloaded and extracted Netsurf 3.6 dev and copied the ENV-content to envarc:. Then I rebooted. I also have the guigfx.library_68060 and the render.library_68060 in Libs:.
But when I try to start Netsurf, nothing happens. Tried it via DOpus and by clicking the icon. Nothing.
But I made SnoopDos logs of both approaches

Run it with -v, it might give some clues. ("netsurf -v")
But try what utri007 suggests first!

[chris]

Ummm - which installer?
There was no installer included in the 3.6 dev Archive.
All there was was the instruction in the readme to copy all the content from the archive's ENV: Folder to Envarc: and the advice to edit a file (that I'm unable to locate in the achive) manually for the use with graphics cards.
No information on how to install NetSurf.

That's not my version.

[chris]

O.K. - it's a  bit confusing with two different forks of Netsurf for Amiga.
May I suggest to rename both?

Actually there are three, as there's a version built for AmiCygnix too!

You'll find Artur's version is called NetSurf68k.

The main factor which tells them apart is the frontend. Artur's is "Framebuffer-SDL", mine is "Amiga", the AmiCygnix version is "gtk".  Once you've seen them all it's easy to tell which is which. Before then, read the readme files.

Which is the version number of the "New Version" linked by utri007 in his posting #8?

3.6dev. It's equivalent to CI build 3602, I think.
Once you have it running, "about:testament" will tell you exactly which Git commit it is built from (and by whom, and when)!

[chris]

I'd go one step further. Put them all on one website, or on Aminet.

http://aminet.net/search?name=netsurf&path=comm/www

I count two versions for 68k, although admittedly it's not immediately obvious what the difference is between them.
My test builds aren't on Aminet as I try to keep that for release versions (the latest there is 3.5, next one will be 3.6).

Also: http://www.netsurf-browser.org/downloads/ contains the *official* builds (no 68k versions on there yet)

I do understand the confusion and try to be very clear in my news articles about which version I'm refering to.

[chris]

I see. But excuse my ignorance - could you please explain to me what "Framebuffer-SDL" means in this regard?

Yes, it's the frontend.  NetSurf is split into the "core" - which is the bit which does all the layouting, fetching, etc - and the frontends - which are the OS-specific bits, so the GUI, rendering, input handling etc.

Framebuffer is a non-GUI interface, it basically just writes the browser direct to the screen.  All the fonts and graphical elements are built internally and the OS has little involvement.  Framebuffer has some other outputs built on top of it, so there's Linux Framebuffer (I assume this will run from a command-line without X, never tried it) and SDL (which uses the SDL libraries to create the display and handle input).

As you can see, the Framebuffer frontend isn't really designed for running on a multitasking windowed operating system.

Arrgh - to top it all both versions (Arthur's and yours) have the same version number! The SDL version I tried to run is also 3.6 dev...

Yes, it's NetSurf internal numbering
Anything between releases gets the next version+"Dev".
Release versions don't have the "Dev".
My versions also have a standard version string, so you can run "version" on them and see the build date (the version number there wil be 3.60006 I think)

Most people run the CI builds, which have the next version number plus their own numbering...

What does "CI build 3602" mean?

It's literally a sequential build number.  The auto-builder builds NetSurf after every commit, build number 3602 occured at around the time I manually built the current "3.6dev".

The version command there will show 3.3602 (which is why release versions are +.60000)

[chris]

[On Reaction and MUI]
Why not support both GUIs and let the user decide which one he likes more ("skins")?

You can't just "plug in" MUI instead of Reaction.  If you wanted to use both you're re-writing at least 50% of the code, and if you want to switch between them you have an additional layer of abstraction.  It could be done, but it's certainly not something I'll be doing.

It turned out that I tried to start Arthur's version with no success.
[...]
And to be honest, after learning that it doesn't have a GUI and will only run from a command-line I don't feel very inclined to test it.

That's not what I said

The "3.6 dev" version CANNOT "work out of the box", as SnoopDos clearly showed me that some files were not in the drawer where they should have been according to SnooDos, e.g. the file "Resource.map" is in "Resources" - not in "Resources/de/" and "Resources/en/", where the system is looking for it upon start.

I had to copy it over manually in order to get an "O.K." instead "Fail" in SnoopDos.

Please, no, don't do that.  Those fails are correct and normal.  If you copy files around like that NetSurf is going to start looking for other files in places where they aren't and there's no telling what will happen.

That resource structure should not be fiddled with, it is searched in a specific manner and missing files in certain locations are expected and dealt with.

"More fiddling"? "Experimental Stadium"?
I would only expect a "dev" version to be in "experimental stadium", e.g. "netsurf 3.6 dev".

All other, earlier versions should run stable, or did I get something completely wrong?

My build is somewhat experimental but 3.5 is pretty stable, and anything marked "dev" is as stable as you'd expect (ie. wavering between fine and entirely non-working).

[chris]

FPU?  It shouldn't need one, but I've not figured out how to stop GCC from insisting on building for it.  I need to force it to use -m68020-60 -msoft-float.

[chris]

I have removed -m68020-60 and it worked.

Something not quite right there, then.  I don't really want to build for plain 68000, which is what that will do.

(it has just made me realise the other libraries are probably 68000/nofpu and NetSurf is building 020-60/fpu rather than the other way round though)

[chris]

I will try building a plain 68000 binary when I get chance.  Pretty sure that's the issue here.

[chris]

Dandy, please try the version I've just uploaded: http://cy2.uk/netsurfos3

[chris]

Ok - will download and try it when I'm back in Cologne tomorrow night!
Is this build compiled for plain 68000?
As I use 68060...

Yes. Keep hold of the previous one in case that's not the problem.

Regarding the "Miracles we do at once, the impossible takes a little longer" in your signature - we have a similar saying here in Germany:
"The impossible we do at once, miracles take nine months..."
:biglaugh:

[chris]

Just tried it - it also crashes.
Will try it now from shell with -v option...

EDIT:
Starting from shell with -v also freezes my system.
Now I hope someone can enlighten me how to get my RAD running...

Do you get any output before it freezes?