Should we really crack IBrowse 2.4???

[chris]

Interesting, but I always heard you needed a really fast/lots of RAM Amiga to run [NetSurf]?

OK..  So, what's your definition of "low spec?"

From what I can see on Aminet:
[blah, blah, snipped]

Netsurf is nice, but needs a GFX card, very much RAM and a fast CPU.

*sigh*

This is what happens when somebody ports the Framebuffer frontend (designed for debugging or no-GUI systems) to something it isn't really designed or appropriate for.  Forget about "NetSurf-68k" and only pay attention to what utri007 has written.

I have been trying to distance the OS4 version from the 68k version because the two are barely comparable, and FWIW the NetSurf core developers aren't keen on the 68k version either.

There are currently three different versions of NetSurf available for the Amiga (four if you include the old MorphOS version).  The 68k one is the worst of the lot, and has the highest requirements.  (The third one is the gtk version under AmiCygnix, which I also try to distinguish the OS4 native-GUI version from, but for different reasons)

[chris]

When I last spoke to Chris about porting to 68k he said that Netsurf renders internally in 16bit, converting that to 8bit for AGA would kill performance.

Actually that's old news now.  In 3.0 I have added 8-bit plotters, and the performance hit from dithering bitmaps down doesn't notice, probably because it is offset by the speed gained by not anti-aliasing text.  It's done via DataTypes so could surely be optimised should somebody have the inclination to do so.

OS4 users can try it out on the newly-available development builds by telling it to run on an 8-bit screen: http://ci.netsurf-browser.org/builds/

[chris]

We'll if it's rendering to 8bit that will make an AGA port more practical.   I don't want to commit to doing a port as I've already got 3 big AGA projects occupying my limited 'hobby coding time'

If you can give me a link to the source code etc I'll take a quick look at it to see what I think the chances of an AGA port are.

http://git.netsurf-browser.org or, more specifically, http://git.netsurf-browser.org/?p=netsurf.git;a=tree
ALL the Amiga specific code is in the "amiga" directory, it shouldn't be necessary to touch anything else.

lets assume its just additional work and maintenance involved he doesnt want to be doomed to. alright.

You assume correctly.

im not a coder, but ive looked at the source once and think that it is not trivial to make it run on 68k, since there is a lot of os4 functional dependency afair. however if youd succeed id push your contribution upstreem. perhaps it would hold some time without breaking and without the need of big adjustments.

Most of the OS4 functions used could be wrapped into compatibility functions quite easily.  The most troublesome bit is probably getting the window scrollbars working (AIUI the older window.class won't do scrollbars for you),  Personally I'd leave that until I have a working web browser that won't scroll

And, yes, if the port is done properly it will be pushed upstream and I won't seek to break it.  It would be beneficial to have somebody else working on the Amiga side of things beyond the initial port anyway, as currently it's just me.

Chris : Could you consider to take a part of job? That way it wouldn't require so much time? And if you could do it co-work with experienced coder like novacoder, it could materialice quite fast?

I already do bits where it benefits OS4 too (such as the recent 8-bit plotter work).  The majority of backporting to OS3 is going to be mundane things like changing functions to older equivalents, which doesn't need two people, but obviously I'll help out where I can.

It's a bit more of a pain these days though due to the code being in a git repository, I end up cross-compiling everything and copying it across for testing.

Did you say that you don't have 68k amiga any more, novacoder has, so problem solved.

No, I never said that.

My guess is that Chris is pretty busy, and might welcome some assistance in porting it to 68k, as long as that port is in line with what he is doing and doesn't cut corners just to get it ported..

That's pretty much it, yes.  If some of the bugs get fixed or other areas get improved along the way that's a bonus.

Basically if the OS3 side involves compatibility versions of OS4 functions where possible, and #ifndef __amigaos4__ in all other cases, I'll be quite happy.  If it involves replacing existing code with OS3 compatible versions (eg. ending up using functions when built for OS4 that are no longer recommended to use), then I'll be much less happy.

When it works (or mostly works) I'll get it into git and try not to break it. Ideally whoever takes this on will stick around to look after it and help out generally, in which case they'll get their own access to the repo.

btw, this is the updated OS3 todo list: http://wiki.netsurf-browser.org/Todo/AmigaOS_frontend#OS3_Support

[chris]

Ok I've download the NetSurf sources and started to compile a few classes to see how it hangs together on OS3.

The Amiga code that I've seen so far looks nicely put together, good job Chris.   What do you use to compile it Chris, I use AmiDevCpp myself.

Thanks.  At the moment I'm cross-compiling under Cygwin using the setup on zerohero's website.  It also compiles quite happily natively under OS4.

[chris]

Try contacting Zapek (David Gerber). Most likely the answer is no (and most likely there are other factors involved) but you can try. You dont get a 2nd chance -- be careful what you write to him.

"Be careful" is an understatement - IIRC he is (or was) very anti-OS3.5 and 3.9.  I hate to think his thoughts on OS4!  Things used to get quite, um, heated on the Microdot-II mailing list if you dared mention that something didn't work on OS3.5, with the fault being attributed quickly to the "pseudo-OS" (or whatever he called it) and no chance of it being fixed in MD2 unless you could prove it didn't work on OS3.1 either.

Actually I still use Microdot-II. It has an annoying bug I discovered recently, which I'd like to get fixed.  Maybe if Zapek is amenable to open-sourcing Voyager I'll have a go at getting the source for MD2.

[chris]

encouraging to hear. he might actually have had some point..

Regardless of personal opinions, if your software doesn't work correctly on a newer official version of the OS which it was designed for, it's highly likely that the software has a problem which simply was not evident before.  Dismissing it out-of-hand just because a bug does not manifest itself on an older version or derivative is (a) not good programming practice and (b) not going to encourage customers to buy your software in the future.

[chris]

@chris: or perhaps he just wanted to keep his support efforts down, not much else than telling people bugs have to be replicated with clean os4 install. but i was joking.

No, I think it was a genuine hatred of anything post-3.1 which wasn't MorphOS.

FWIW the bug still exists when I run MD2 on OS4.1, but it's an unimportant feature I wouldn't use anyway.

[chris]

@chris:
why, then we could have a good chance. this is not about os4, as os4 doesnt need voyager anyway. this is all about genuine amiga, even if it means up to 3.1 which is fine with me.

Good point.

[chris]

on all respect, despite what chris suggests its not just a walk by to port his netsurf back to 68k. if it was so easy he would best do that himself in an instant, gifted as he is and already familiar with the code and implementation details.

I do bits here and there to help with the back-porting (such as the recently-added 8-bit plotters), but I'm not interested enough to spend time doing all the boring bits - I'd rather progress the browser for OS4, where I actually use it.

If anybody wants to step up and give it a try or help NovaCoder out (who was having problems with cross-compilation btw, not the code itself) then that would obviously be appreciated.

[chris]

of course it will not give us planar potters magically, but i fear the same is true for netsurf.

Except I've already written those. They aren't planar per se, just built for any palette-mapped screen mode.