Netsurf for 68k amigas, css capable web brower

[wawrzon]

isnt that the stuff after line 320 is optional? what if you really set endif on 320?

[wawrzon]

let it out for first, i would say

[wawrzon]

@seyia: then again probably would be better to take aros or at least mos owb source. os4 owb depends eventually on system functions not present in 3.x and on new reaction. both mui 3.8 and zune is available for 68k.

[wawrzon]

GOGOGO!!!

[wawrzon]

is not that easy 4 sure. lets be patient.

[wawrzon]

__artur:
why, i see you have made some progress with itix backup. just do-not-surrender.

[wawrzon]

@chris: great to have all fractions helping here.

@artur:
an actual discussion on amiga-news.de regarding a restricted os3.x unicode rendering implementation can be found here, if that is the problem, that is:
http://www.amiga-news.de/forum/thread.php?id=31281&BoardID=7
(unfortunatelly in german, but skilled coders involved)

[wawrzon]

very well! display is trashed unusable on most color depths, it works with partly wrong colors on32bit though (as described). complicated pages do not render quite correctly but neither completely wrong, so its going in the right direction (amiga.org renders in about 28s) there is always an exeption on exit pointing to ixemul.library (61.1). if i comment smart crash out in my ss i might maybe catch more details about it, but supposedly you know yourself. no hits while execution. seems quite stable despite still slow, but doesnt block the system.

ps:many thanks to you and bernd!!

ps: on 16bit it seems to crash the system. on16bit_pc it throws a lot of hits. i recorded hits that occure on quit. (the offsets seem always to be the same) as follows:

Code: [Select]


USP : 0920792C SR: 0004  (U0)(F)(-)  TCB: 09167470
Data: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
Addr: 08000ADA 08BB1520 00000000 00000000 00000000 00000000 3A67616D 080023B4
----> 0853221C - "Work:C/Assign"  Hunk 0000 Offset 00000214
Stck: 65747375 72663200 6E736662 2D73646C 00000000 00028600 0241523F 70FF4E75
Stck: 4AFC0920 794C0920 7966803D 09000920 79660920 79750920 79B06978 656D756C
Stck: 2E6C6962 72617279 00697865 6D756C20 36312E31 205B6E6F 74726170 2C203638
Stck: 3032302C 20363838 38312C20 616D6967 616F735D 20283139 2E332E32 30303929
Stck: 0D0A0000 000002E4 092079C0 09208468 09208494 092084AA 092084CA 092084EA
Stck: 092084E6 09224CA8 09224CD4 09208EFC 092206C4 09224D64 09224D88 09224D98
Stck: 09224DB0 0921F584 0921F5D0 09224DC8 0921F64C 09224E18 0922081E 09220814
Stck: 09220858 09224E4E 09220836 09220828 09220844 0920902C 09209328 092095C0
----> 09816474 - "Leer1:games/Netsurf2/nsfb-sdl"  Hunk 0000 Offset 000002B4
----> 092079C0 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00000078
----> 09208468 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00000B20
----> 09208494 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00000B4C
----> 092084AA - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00000B62
----> 092084CA - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00000B82
----> 092084EA - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00000BA2
----> 092084E6 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00000B9E
----> 09224CA8 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 0001D360
----> 09224CD4 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 0001D38C
----> 09208EFC - "Work:Libs/ixemul.library"  Hunk 0000 Offset 000015B4
----> 092206C4 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00018D7C
----> 09224D64 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 0001D41C
----> 09224D88 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 0001D440
----> 09224D98 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 0001D450
----> 09224DB0 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 0001D468
----> 0921F584 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00017C3C
----> 0921F5D0 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00017C88
----> 09224DC8 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 0001D480
----> 0921F64C - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00017D04
----> 09224E18 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 0001D4D0
----> 0922081E - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00018ED6
----> 09220814 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00018ECC
----> 09220858 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00018F10
----> 09224E4E - "Work:Libs/ixemul.library"  Hunk 0000 Offset 0001D506
----> 09220836 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00018EEE
----> 09220828 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00018EE0
----> 09220844 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00018EFC
----> 0920902C - "Work:Libs/ixemul.library"  Hunk 0000 Offset 000016E4
----> 09209328 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 000019E0
----> 092095C0 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00001C78
PC-8: 2F40000C 2C780004 226F000C 4EAEFE86 70142F40 0010202F 00102C5F 4FEF001C
PC *: 4E75514F 42A74879 09110A56 48790911 089C4879 09110898 4AAF001C 670A2F7C
09816454 :  2f40 000c                  move.l d0,$c(a7)
09816458 :  2c78 0004                  movea.l $4.w,a6
0981645c :  226f 000c                  movea.l $c(a7),a1
09816460 :  4eae fe86                  jsr -$17a(a6)
09816464 :  7014                       moveq.l #$14,d0
09816466 :  2f40 0010                  move.l d0,$10(a7)
0981646a :  202f 0010                  move.l $10(a7),d0
0981646e :  2c5f                       movea.l (a7)+,a6
09816470 :  4fef 001c                  lea.l $1c(a7),a7
09816474 : *4e75                       rts
09816476 :  514f                       subq.w #$8,a7
09816478 :  42a7                       clr.l -(a7)
0981647a :  4879 0911 0a56             pea.l $9110a56
09816480 :  4879 0911 089c             pea.l $911089c
09816486 :  4879 0911 0898             pea.l $9110898
0981648c :  4aaf 001c                  tst.l $1c(a7)
09816490 :  670a                       beq.s $981649c
09816492 :  2f7c 0911 0a5e 0014        move.l #$9110a5e,$14(a7)
Name: "nsfb-sdl"  Hunk 0000 Offset 000002B4

@djbase: you sure have tcp running?

[wawrzon]

@bernd:yes there are sites that load extremly long or do not load at all, like bild.de you pointed me once to. i wondered if this is maybe because netsurf doesnt get enough memory allocated or it requests further objects too lame, but this sounds bs.

to load netsurf_web_site my 060/50 setup needs 4.9 s, this is on a clean first attempt, but if you have surfed around and then tried the link first the actual time can vary around 5-6sec. same for reload (around 5sek)

wikipedia loads in 18,x s at best, but sometimes it takes up to 30s. idlea68k shows constant full cpu load while processing page.

reuters page loads basically some 14s (it only shows 4.9s shortly after) an then it loads further content. i have some hard drive activity like something is being cashed. i can scroll a little but it takes ages and netsurf seem to update the layout all the time. but maybe its due to my impatience. its a huge page too.
edit: well no, i have it open for some time now, it is rendered but being updated again and again

btw:i dont understand why you care for mos so much out of the sudden:P there is mos native port worked on by itix afaik, mos is surely better off with that. not that you sholdnt join the forces though.

[wawrzon]

so better for tests on a classic not use wipeout or other memtracker

i disabled all debug in background now but it doesnt make much difference if any at all.

yes and so i like to get 1 sourcetree.itix is very friendly, he help Artur and have some code in his source add that make it easier to work on OS3.x and MOS1.4.He have written in netsurf source, there seem working together possible.

this attitude is much appreciated.

but because we was not sure what problem is we want see netsurf running to see libcurl libxml libhubbub and the many other libs run well.and now can see the browser work and the way to a full browsers is GUI code only.

the sdl version you see have only very few elements.

there are still bugs to iron out. colors. text input. rendering (check http://www.bahn.de). not loading pages. i do not know how much of this is general fault of current netsurf distribution. replacing sdl with native functions might bring some speedup but i dont take it for granted. gui would be nice too, of course.

[wawrzon]

what happen on this page with a netsurf OS4 or MOS Version ?
that the colors are wrong is because endian problems.

that is clear. it shouldt be too hard to fix it, right?;-D

its really bad that OWB is program in C++.i thinks thats main reason that it is so slow and compile time is so extrem long.

and i need also much longer to find Bugs in a C++ source than in a fast c source.because when singlestep in debugger c++ source need lots more instruction for same functions

i see. bad thing: it is seems c has been abandoned already. most available sources are >c++ for what ive seen.

[wawrzon]

it isnt noticeably faster alas.
the sashimi output on exit looks like that now:

Code: [Select]


09-Jun-09   19:30:11
WORD READ from 73757266 (INST)                 PC: 73757266
USP : 0A885068 SR: 0004  (U0)(F)(-)  TCB: 0A7E4BA8
Data: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
Addr: 08000ADA 08BC3178 00000000 00000000 00000000 00000000 616D6573 080023B0
Stck: 325F3100 6E736662 2D73646C 00000000 0963BF38 00000000 00000003 00000001
Stck: 00000000 00000000 096A37B8 00000000 00000000 096A382C 00000B18 09609F2C
Stck: 00000000 00000000 00000000 00000000 00000000 00000A88 50C80000 00000000
Stck: 0A8850B4 00000000 00000000 00000000 00000000 00000000 00000000 00000000
Stck: 00000000 00000000 00000000 00000000 00000000 00000000 00000A88 510C0000
Stck: 00000000 0A8850F8 00000000 00000000 00000000 00000000 00000000 00000000
Stck: 00000000 00000000 00000000 00000000 00000000 005C0871 123441C0 0002FFFE
Stck: FFFE0000 00000B18 4A2EB0CA 00000000 4A2EB0CA 00000000 4A2EB0CA 0871005C
----> 096A37B8 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00017B60
----> 096A382C - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00017BD4
PC Address invalid
Name: "nsfb-sdl"  


09-Jun-09   19:30:11
Exception !!   00000002     TCB: 0A7E4BA8     CTX: 08178A90     SSP: 080023B0
USP : 0A885068 SR: 0004  (U0)(F)(-)  TCB: 0A7E4BA8
Data: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
Addr: 08000ADA 08BC3178 00000000 00000000 00000000 00000000 616D6573 080023B0
Stck: 325F3100 6E736662 2D73646C 00000000 0963BF38 00000000 00000003 00000001
Stck: 00000000 00000000 096A37B8 00000000 00000000 096A382C 00000B18 09609F2C
Stck: 00000000 00000000 00000000 00000000 00000000 00000A88 50C80000 00000000
Stck: 0A8850B4 00000000 00000000 00000000 00000000 00000000 00000000 00000000
Stck: 00000000 00000000 00000000 00000000 00000000 00000000 00000A88 510C0000
Stck: 00000000 0A8850F8 00000000 00000000 00000000 00000000 00000000 00000000
Stck: 00000000 00000000 00000000 00000000 00000000 005C0871 123441C0 0002FFFE
Stck: FFFE0000 00000B18 4A2EB0CA 00000000 4A2EB0CA 00000000 4A2EB0CA 0871005C
----> 096A37B8 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00017B60
----> 096A382C - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00017BD4
PC Address invalid
Name: "nsfb-sdl"  

so, it seems its been improved. it isnt deadly anyway. if you ask me the stability of the system is not affected at all.

[wawrzon]

@piru: we dont need to be that picky about the results atm. it might become important while the port will be optimized, but at the moment approximate results will do too. the most important is how netsurf does in praxis. and i have to say even with this simple sdl solution it is beyond my expectations. im testing the os4 post as well on the same system and have to say though more advanced os4 version seems a little slower on simple websites and less stable, which is but the fault of the system itself i suppose. os4 netsurf had stability issues from the start. on the other hand the current version renders correctly quite complex websites, which 68k port is not able to do yet.

[wawrzon]

i dont believe that a wrong display website depend on sdl port or not.

the http://www.bahn.de is same wrong on OS4 ?

i have already reported it to you yesterday by mail. the problem looks the same. it is not a sdl issue.

[wawrzon]

owb on os4 is in meantime using the clever amiga-way to render pages too. lol.
reuters is really quite an unusable benchmark if:
1.netsurf doesnt load it correctly anyway.
2.there is perdefinied missunderstanding as to when the page is loaded completely.
3...
go, find a better one. but anyways to me its clear already that netsurf is quite fast even though the engine seems to be still a little limited.