AROS68K development methodology and bounties

[MartinW]

Originally Posted by wXR

Hoho, just look at all these roadblocks being revealed as a mirage!

One of the biggest ones is the insistence on GitHub because it's more than just a version control system. There are plenty of options out there that also meet this requirement, although I'll admit not all of them are free for more than a small amount of people. Often times however, free accounts are available for open source teams / projects. The Atlasssian tools are a good example.

Even Microsoft's (shock horror!) Visual Studio Online, which actually I think they just rebranded again supports issue tracking Agile / Kanban planning boards, sprints, collaboration etc. I use that in my day job even though I develop for Linux / Mac. OK, I don't think that one would work because I don't think they have a free option for open source projects, but my point remains the same.

Anyway, will go back to lurking now.

[wXR]

@MartinW

GitHub -- effectively a developer social network -- is where most developers around the world hang out, and it integrates with Bountysource, which is what I will use to disseminate the bounties. It is my requirement that all issues which will have bounties attached be tracked there. It is not a requirement that all development takes place there. Far from it.

I'm asking comparatively very little; nothing of any significance has to change.

Meanwhile, we could discuss actual bounties? Here are some that were sent to me via private message:

1. Booting AROS68k/Wanderer on A1200, A600 and A4000 (perhaps A3000 and Vampire as a bonus) without any bootstrap mechanism (using AOS to load the AROS kickrom and the maprom the AROS kick is what already works today)

2. a certain score in AIBB on a similar selection of Amigas (individual scores for each model to account for the different hardwares) to ensure a good user experience (it seems that much of the common accusation of AROS being slower than molasses on a winter day comes from the graphics subsystem)

3. something for the looks. It's difficult do define a bounty for this, though. Ideally Wanderer should look both good and like a legitimate replacement for WB

4. Wanderer should become a feature complete replacement for WB 3.1 (not sure how much is missing)

5. perhaps define an upper margin for the memory footprint like "must do all the above with 8MB fastmem of which 4MB must be available after boot"

6. perhaps select a list of standard applications that must work on AROS68k (theoretically everything should work, no idea whether it dies)

People who are most informed about the project should speak up; a healthy and fruitful discussion about how each bounty should be defined and rewarded in proportion to the others is what we would ideally get out of this.

[Minuous]

How about:

7. support for the features that were introduced in OS3.5/3.9

[Samurai_Crow]

Originally Posted by Minuous

How about:

7. support for the features that were introduced in OS3.5/3.9

Those were the first versions to have an end user license agreement that forbids reverse engineering. Fortunately EULAs were struck down in Europe. It is just Americans like me that suffer.

[clebin]

@wXR

3. If you swich off themes, you get something that looks just like 2.x/3.x. It's also easy to install Mason icons or something like that. Generally it's up to the distro to decide the look & feel. I agree that a better default theme/icon set would help if only because it appears so much in screenshots and gives people a false impression.

4. Wanderer does a lot of things that Workbench doesn't - it can behave in a traditional 'spatial' Workbench-like way or more like a file browser. It's not so much filling in features but polishing what's there AFAICT. I'm not sure about the status of Wanderer's ARexx (Regina) port vs Workbench though. Kalamatee had/has a branch of Wanderer with new features including plugins, so that you can have a pane with a file-tree for example. Not sure on the status of it but Kalamatee has a lot on his plate already!

There is a lighter weight ROMable solution called Workbook but it lacks even basic functionality. There is already a bounty for improving it. Rather than an endless compromise between features and speed, I think Workbook could become a solution for low-end 68k with Wanderer aimed at higher-end 68k and modern hardware.

7. What features from 3.5/3.9 does AROS lack, besides Reaction? I'd prefer to see Zune reach parity with MUI4 first although I've nothing against Reaction.

Another thing to look at is to find the paper-cuts initiative that happened some years ago. The aim of that project was identify and fix a lot of minor bugs with the aim to improving the experience for end-users. Plenty of good things came out of it but not all of it was completed. For example, there are a few bugs in the way AROS handles custom screens which I flagged up (just to get my pet peeve in there!). There'll be a series of issues like that which will help with the fit & finish.

[wawa]

Originally Posted by wXR

1. Booting AROS68k/Wanderer on A1200, A600 and A4000 (perhaps A3000 and Vampire as a bonus) without any bootstrap mechanism (using AOS to load the AROS kickrom and the maprom the AROS kick is what already works today)

so how do you want to boot it if not softkicking it from an amiga rom? you imagine a reboot cycle is necessary to install replacement, right? 3.9 does the same.

btw, anyone who has a 1mb rom adapter/switcher can flash their own roms alright i guess. i think toni did something like that with an aca card. still it isnt wost convenient solution today, when improvements to the whole systemare being provided daily.

[/QUOTE]
2. a certain score in AIBB on a similar selection of Amigas (individual scores for each model to account for the different hardwares) to ensure a good user experience (it seems that much of the common accusation of AROS being slower than molasses on a winter day comes from the graphics subsystem)
[QUOTE]

tests like aibb score eactly the same on aros and amigaos for the most part as far as i remember. its not the simple drawing functions that lack, the penalties are probably rather higher level, calling these functions.

3. something for the looks. It's difficult do define a bounty for this, though. Ideally Wanderer should look both good and like a legitimate replacement for WB

i have some planar iff skin and a few iff icons id like to contribute to aros at some point. the problem is aros converts these icons to full color internally and maps them back even on planar, which partly defeats the advantage of using them, especially the dedicated transparent pen.

4. Wanderer should become a feature complete replacement for WB 3.1 (not sure how much is missing)

wanderer prefs are not being read and saved properly on big endian. i know where the problem is, but cant or couldnt fix it up till now due to being c illiterate. besides there is graphics glitch i detailed lister mode, when decoration is turned off. i think a number of things could be optimized, in particular when loading icons in the lister (window). few years ago kalamatee mentioned he had a private upgrade to wanderer at hand, im curious what it is.

5. perhaps define an upper margin for the memory footprint like "must do all the above with 8MB fastmem of which 4MB must be available after boot"

aros needs 1mb to boot without the s-s and 6,5 mb to boot to wanderer. paerhaps this usage can yet be minimized disabling features, for instance not loading the whole set of usb classes while early startup, especially on a platform that has no usb hardware at all.

6. perhaps select a list of standard applications that must work on AROS68k (theoretically everything should work, no idea whether it dies)

mostb standard amiga applications i can think of work with aros. games that exclusively rely on some chipset features not correctly supported on aros may not (ask toni). im open for proposals, about what doesnt work and needs to be made working.

one example that comes to my mind is hd-rec. it has a graphical glitch that prevents proper usage especiylly when recording. i have dsicussed it with toni. there is workaround, but no final fix.

[wawa]

Originally Posted by Minuous

How about:

7. support for the features that were introduced in OS3.5/3.9

could you be more specific? i also think this needs to be divided in smaller distinguished pieces, that may be approached one after the other.

from what i gather reaction is your main concern? could you list the set of classes necessary? i suppose your expectation exceeds what is available via class act. one might try to approach the autors wiat a proposal for open sourcing and work up from there. but if the whole set neesd to be replaces from scratch i would think it is rather a low priority for now.

[Minuous]

@SamuraiCrow:

>Those were the first versions to have an end user license agreement that forbids reverse engineering.

Are you sure? I seem to recall a similar clause in earlier versions but could not find a copy of the 3.1 EULA, is there a scan available somewhere? Anyway EULAs are generally unenforceable not just in Europe but even in the USA, last time I checked. Plus of course adding support for such functionality would not necessarily require reverse engineering. The official line as I understand it is that every existing part of AROS has been done without reverse engineering.

@wawa:

There are brief feature lists at http://os.amigaworld.de/index.php?lang=en&page=7 and http://os.amigaworld.de/index.php?lang=en&page=14 but for the full changelogs you will need the OS3.9 NDK. And if you use the Developer Companion from Aminet or my site it provides an easy interface to the changelogs, just click the "Release Notes" gadget. There are hundreds of improvements and it would not be feasible for me to list them all here.

ReAction is a major and important component, yes, but there are many other areas of the OS that were upgraded. Obviously not everything could be done at once. ClassAct is a very early and buggy implementation of ReAction; there was a reason these classes were upgraded by H&P rather than just incorporated as-was. They would not need to all be replaced from scratch, there is an OpenReAction project that could be used as a starting point.

[wawa]

@minous

http://os.amigaworld.de/index.php?lang=en&page=7

going one by one:

Workbench

i think wanderer has async functionality

ACTION - AVI & Quicktime-Player

aros has several players, ffmpeg port amomg others, but also old amp2, which i tried and partly succeded to compile for 68, however ran into linking problems, since it was meant for warpos initially. i can share work with anyone who wants to, but i doubt this application has much value apart of historical. anyway playing such media on 68k may be a bit problematic anyway except with vampire, and they have a player already, ahich will certainly work with aros.

MPEGA.Library - Software MP3-Decoder

aros has mpega.library


PlayCD - New CD-Player

aros has cd-player (in contribs, one by salass i think)

CacheCDFS - CD ROM Support (*3.5)

aros has cd-rom support

AWEB 3.4 Special Edition - Internet-Browser

aros has owb, webkit based css browser. odyssey to come, perhaps. 68k asl netsurf works, netsurf itself might be included into contributions or ports, along with itix mui gui, if anyone wanted to put some work into this.

Internet / Network Access - TCP/IP Stack

aros has an amiga compatible network stack built in

AmigaMAIL - E-Mail Client (*3.5)

aweb was working on aros last time i tested (also 68k version on aros68k). it could be contributed, i guess.

RAWBInfo & several other tools

aros has most of that if not all, i guess. colorwheel gadget of 3.9 even comes from aros i think. id have to look in detail.

Automatic file type recognition

aros has this.

Improved Libraries & System Tools

bit vague. at least icon library exceeds the functionality of 3.9 one i think. you can scale icons to a bounding box. png support, the like.

Find & Clock

may be missing.

IomegaTools - tools for Zip and Jazz drives
New Shell & many other tools

iomega is obsolete
math libs are up to date
shell is improved, with history and such
fix fonts and font management is improved, ttf support sll the way
rexx is buggy

Unpacker Tool

XAD Library - Universal Decruncher System

xad and a choice of gui unpacker tools available

Improved Prefs

iprefs, ahi prefs, screenmode prefs, wb patterns backrounds all present. much more functionality.

Glow Icons - New Icon Style (*3.5)

can be used.

WarpOS Prefs

warpos doenst make sense in this context imho.

HDToolBox - Harddisk Toolkit (*3.5)

tool is present.

NSDPatch - Support for Harddisks >4GB (*3.5)

not needed, support for big volumes is built in.

HDwrench.library (*3.5)

as above.

Text-Editor-Gadget (*3.5)

is missing i think

[wawa]

altogether i dont see anything essential missing from 3.9 on aros in this respect. some minor parts can yet be delivered. i think incompatibility with some 3.9 libraries may stiil be an issue, which remains to be researched, with particular applications that demand those library versions.

[wawa]

http://os.amigaworld.de/index.php?lang=en&page=14

looked at this as well and cannot determine anything missing from aros, save for an own email client (there is a number other messaging, ftp and other network apps though)
also resource.library is missing. i dont know what that is for..

anything else, really? aros has (obviously) much more essential content and features than 3.9 for what i see.

[grond]

Originally Posted by wawa

tests like aibb score eactly the same on aros and amigaos for the most part as far as i remember. its not the simple drawing functions that lack, the penalties are probably rather higher level, calling these functions.

That's something that needs verification. AIBB supposedly shows the speed of the graphics system. The Vampire Team is working on an acceleretated P96 driver and AFAIK each overlay function that gets added to the driver causes some or several AIBB scores to go up. Thus, AIBB should be a good tool for quantifying GUI speed.

[wawa]

AmigaMAIL - E-Mail Client (*3.5)

aweb was working on aros last time i tested (also 68k version on aros68k). it could be contributed, i guess.

i meant of course yam. however aweb is working too, given class act classes are installed.

[wawa]

Originally Posted by grond

That's something that needs verification.

very well. anybody can verify things like that for themselves. first of all i might recall wrong, secondly things change. its been a while since i have done these benchmarks.

[idrougge]

Workbench 3.9 has extensive ARexx support.

[wXR]

I'm hearing here and in another thread ( http://eab.abime.net/showthread.php?t=87196&page=2 ) that we need to put some effort into the graphics subsystem.

Lets pick the first bounty soon. As I said in my original post, I'm not interested in bounties for mp3/cd players and stuff like this -- generally apps are out. This is about improving the core system on m68k.

[wawa]

Originally Posted by wXR

I'm not interested in bounties for mp3/cd players and stuff like this -- generally apps are out. This is about improving the core system on m68k.

good attitude. the apps are mostly there in this or other form. its the system we need to concentrate upon.

[Minuous]

@wXR, wawa:

I agree, apps like AMPlifier, AmigaMail etc. are not really what is important in this case. Of more concern are the V44 and V45 system components that are used by apps (libraries, classes, datatypes etc.) and things like the improved and bugfixed CLI commands. That feature list concentrates more on the new components than on the upgrades to existing components, the changelog in the NDA provides details on the upgrades to preexisting components.

[wawa]

Originally Posted by Minuous

(libraries, classes, datatypes etc.) and things like the improved and bugfixed CLI commands.

since these parts are being continously improved on aros (there is more effort than just plain 3.1 compatibility, the issues are, according to what i witness, about how to maintain the compatibility to different versions while doing it right) its a bit hard to tell in what state they are.

i think it might be good to keep some database of critical components, suspected not being compatible on aros with either major version of the os. maybe that way the best allround compatibility can be first considered and then eventually improved. care to contribute to testing? i mean testing existing componenets, not those yet missing, and delivering negative test results, ie. when something doesnt work 8with detailed description of the problem). ususally its what bugtracker is for..

[wXR]

@wawa

Great idea with the database. Actually that sounds like something our mate @gulliver is good at and likes to do. :-)