NetSurf 3.2 68k released

[chris]

As a side note, does Netsurf load pictures using datatypes?

NetSurf uses internal decoders for common file types.  The OS4 frontend additionally registers DataTypes as a fallback, but I don't think the 68k version does, and, in any case, the internal decoders get priority.

[chris]

Hi!  Thanks for the responses!  I actually downloaded four different versions, but was having a hard time even telling which versions were for 68k, which for OS4, etc.  I'm gonna say this is due to insufficient documentation, but probably more due to the fact that I don't know anything about the software and have never used it.  Or as we call it in the IT field, "operator error".    Was trying "3.3 Dev" that said it was uploaded to Aminet 9/29/2014, so I didn't think it was too old.

My non-Framebuffer Alpha is still here.  It's likely very slow but I'd be interested to know if it works for you.  There's another thread... somewhere... about that.

I wish there was better documentation for that "Options" file.  I'll keep poking at it!  

Try here and here.

[chris]

I do not care about what is "native version" because it simply not works for me.

A bug report which just states "it does not work" is not very helpful.

[chris]

Oh sorry

it was not about downqualifying your work

it cannot work because you use Reaction but Reaction is closed. I have Classact but Classact and Reaction are not identical

Reaction is included in OS3.5/3.9.  Oh, I've just realised you're the AROS guy.  That explains it then, fine.

[chris]

I guess I just don't understand the point of developing an Amiga web browser that only works on emulated Amiga's?

Me neither, which is why I was targeting 030+AGA for my version.  Sadly I don't really have the motivation/knowledge to do much more with it, and nobody else is interested (I wouldn't have got as far as I did without Matthey's help though) so it has stalled.  In particular, there's an optimisation bug which really could do with being fixed.

[chris]

sad news that nova is leaving, but cant blame him. id probably be able to compile netsurf, with chris help but at some point ive lost temper. im not very patient person. and netsurf build system is too complicated. on aros you simple do configure and make

It is just "make" with a few parameters, the problem is you have to build quite a few things in the right order and cross-compiling complicates matters.

There are pre-built toolchains for x64 Debian here, which might simplify it a bit: http://ci.netsurf-browser.org/builds/toolchains/ (m68k-unknown-amigaos-18 is the latest one at the moment)

With that in situ the libs should build with "sudo make HOST=m68k-unknown-amigaos install" and NetSurf with "make TARGET=amigaos3"

[chris]

What is that, bytes?  Kbytes?

Bytes.  If you'd read the links I posted earlier you'd know that.

Long story short, I feel like this program is too complicated for a layman to understand.  *If* these options are the reason why it's crashing, then perhaps a GUI-based preferences program to help a user set them, prior to running the program, might be in order?  

This is the problem with using a frontend designed for non-GUI and debugging.  All the other frontends have proper prefs GUIs (including mine for OS4, IIRC the prefs window doesn't open on the OS3 build, but as I said before nobody is helping with that so it's unlikely to be fixed anytime soon)

[chris]

Edit:  I wonder if it would be possible to run the application from RAM:, but assign the cache location to a hard disk somewhere?

Probably not with the Framebuffer version, but you can set the disk cache size to 0 to disable it.

[chris]

Hi there - is there any version available that doesn't require a FPU?  I see in the changelog - 3.4dev does not require ixemul.library -- but the latest version on aminet does appear to require ixemul.library..   which requires a FPU .

Just trying to get Netsurf working on my humble ACA1221 (68020@28 mhz, 32MB ram, no FPU) A1200 .

Thanks!
John H

My version was working without an FPU, but at some point it decided it needed one.
I can't figure out what I need to do to gcc to get it to default to "-m68020-60 -msoft-float", that should fix it as I think one of the dependencies is building with the wrong options.

[chris]

Chris - any chance you'd have an older non-FPU version on hand?  I don't need the latest, just wanted to play around a bit..  Greatly appreciated ..

(Figuring it'll be a while before Vampire A1200 is out and makes web browsing.. a lot better!)

Maybe on my CF card, but I think I deleted it, and I don't have an easy way to check whether it's an FPU version or not!

[chris]

You are right, 2 or 3 minits is about a same. But with 040 it matter of useability, if loading time rises from 40 seconds to 60, it is huge leap of patience. 40 seconds is OK, minit start to be long time? With 060 difference ist not that big, just few seconds.

How are you even measuring this?  My build of NetSurf is built with soft-float as are all libraries, it's only some of the dependencies that might not be.  There's very little floating point maths in NetSurf anyway AFAIK.

Could it be possible to make it support tool types? So that FPU could be enabled/disabled that way.

No, that's not viable.

Though, one question is Chris still interested to develop Netsurf further?

I've done about as much as I can with the OS3 frontend, which is more than I really had interest to do in the first place.

When I update OS4 frontend the OS3 frontend automatically gets the changes (unless it's to do with new OS features), which was the point of deriving the OS3 build from my code and keeping it all together (rather than forking it).  Core changes are obviously going to apply to both versions too.

Anything further I'd appreciate somebody else working on.

[chris]

Test results are from that old build wich you made for someone who asked.

IIRC that was the one where I built GCC using a "no FPU" switch which wasn't documented anywhere to say it would work on 68k target, and nobody told me whether or not it worked without an FPU, but I got the impression it still needed one.  God knows what that was doing.  The current build is soft-float still, but without that iffy switch.

[chris]

Chris - any chance you'd have an older non-FPU version on hand?  I don't need the latest, just wanted to play around a bit..  Greatly appreciated ..

August 2015: http://homepage.ntlworld.com/cdyoung/tmp/netsurf_os3_old.lha
Feb 2015: http://homepage.ntlworld.com/cdyoung/tmp/netsurf_os3_feb2015 (executable only)

They may or may not work, and they will almost certainly be horrible.

[chris]

Could you you consider to get rid of friend bitmap and AGA version would need some love too. There is a problem with png pictures with invisible background, but just with AGA.

It uses a 1-bit mask.  I believe this works, but it's not ideal.
It's a shame guigfx doesn't support alpha blending.  render.library does, but it's not feasible to blend everything in raw ARGB bitmaps, and I think reading the render BitMap into a memory area, blending across and re-writing to the screen will be too slow (not to mention a lot of hassle).  I might try it at some point.

Those fixed only bigger thing to do is find memory leak / fix fragmention?? Am I right?

Yes.

[chris]

I forget, aga has that "red problem", like white and red are swapped.

That's not an AGA problem, it's an 8-bit mode problem.

I thought it might be a colour byte order problem but I tried all plausible byte orders and none of them were better.

If it's working in 32-bit modes then that would suggest the byte order is correct, and there's something wrong with guigfx's pen selection instead.

I'd be surprised if there was that obvious a bug in guigfx/render.library but you never know...