Bounty proposal: Compile RISC OS with free tools

Currently, you can have an almost-free RISC OS system. A RISC OS 5 ROM image is free, and you can run it on RPCEmu, which is free. The RISC OS source code is open. You can build the whole thing on RPCEmu. But for a few files: the C compiler, the objasm assembler, CMHG and a few others, which cost money. This is a major stumbling block to getting new people interested in developing.

Open source versions of these exist. GCC is quite capable these days, asasm (part of GCC) almost a complete replacement for objasm, and cmunge replaces CMHG. It’s possible to build the IOMD and Iyonix versions of RISC OS using asasm:
http://www.riscos.info/index.php/RISC_OS_Open_Support

So I’d like to suggest a bounty: adjust the RISC OS source, or adjust the tools, so that RISC OS can be built with the free tools (without compromising building on Acorn’s tools). Some of this is trivial (fixing filenames), some of it is more complex (getting so far as building SharedCLibrary). But I think the reward in being able to say that developing on RISC OS Open is truly free would be considerable.

Wouldn’t it make more sense for everyone to be using the same build tools, to avoid introducing incompatibilities or other issues? Perhaps a bounty to provide C tools CDs to interested parties (with a pedigree) might make more sense in terms of lowering the barrier of entry?

To be honest, I don’t really see why RISC OS has to be so “free”. The C tools are one way for ROOL to fund this site, and their show attendance etc. I think it would be best to support their efforts in this way. Do we really need to devalue RISC OS and its software further? As you say, certain parts can already be worked on with GCC, and the autobuilder will generate images from modified code in the repository…

The motivation is that shortly people will be running RISC OS on Raspberry Pis, costing £25. Tell them they need to spend £40 on a compiler to do development, and they’ll just laugh, they won’t pay up. The core RISC OS market (those who go to shows etc) have enough investment in the platform and the community that they not have a problem with that, but this is a small group and one that is static or dwindling. To get new blood means new developers, and even small barriers are going to mean they get diverted onto other things. It’s not even just about money, it’s that you have to wait a week to get the tools (with no criticism aimed at ROOL here, who are all busy people). If you decide to start hacking on Sunday afternoon, you discover the tools need to be ordered, and the moment has passed. You can’t fix that bug that was annoying you… maybe it needs to wait for the next wet Sunday afternoon in three weeks time, but then you instead get distracted with that phone app you just found…

That’s also neglecting the issue that for some people £40 is a /lot/ of money. Even in the UK, let alone other countries that might have RPis. Perhaps not people with full-time professional jobs but students and the retired, for example.

This is not to say that it should completely break the business model. As yourselves already realise, there’s a market for providing polished ready-to-go solutions even if all the components are available for a much lower cost by themselves. Things like the RPCEmu USB sticks are a good example… a packaged solution that just works, saves lots of hassle and, for those who have sufficient interest to turn up to shows and sufficient disposable cash, worth the time saving. Likewise, you sell your commercial software to people with that investment in the platform… if that number doubles, even if the number of ‘freetard’ users goes up tenfold, you still win. See the market for phone apps… if you make them $2, you make them easy to buy, and there’s enough users out there, you make bigger revenue than if they were $100.

My argument is that ROOL would lose the small amount of cash from developer CDs and gain more actual developers with time to do development (ie not the professional software engineers who could afford the price of a CD but have no time). Given that the price of a CD pays for less than one hour at commercial rates, this seems to me like a good exchange rate. In this instance, the revenue stream would still be there by a) selling packaged products to users and developers and b) the increased user base brought on by having more developers.

I agree using the same build tools is good. I would suggest that GCC is getting to the point where it’s as good as Norcroft, so I can’t see a huge advantage for using Norcroft (other than it works on A5000s, but that’s hardly a great feature these days).

Wouldn’t it make more sense for everyone to be using the same build tools, to avoid introducing incompatibilities or other issues?

This is something I would agree with. The Norcroft/Acorn/ROOL suite is very capable – so I would suggest the “free” alternatives make themselves compatible, not the other way around. Certainly, the code should not contain a pile of tweaks and modifications for other build tools…

might make more sense in terms of lowering the barrier of entry?

?
Please allow me to put it like this – once my Beagle is sorted and working reliably, I intend to buy the latest set of tools. I should point out that my wage is pretty much minimum (it’s crap pay, but it is a zero-stress job). To me £40 is a lot, but it is something I am willing to provide in terms of getting a decent compiler and helping to support RISC OS.

Or, to put it another way – who in their right mind gets a bargain ARM board such as the RPi and expects to build the operating system from scratch that day? My God – there are enough free tools around (inc. BBC BASIC) so if you want to learn to program RISC OS you can get on with it. You don’t need to spend £40 on a dev suite until you have some familiarity with the system and want to look a little deeper. You can write OS modules in BASIC, so you can do a lot from the outset. And yes, for free.

I would recommend the dev suite, certainly. But what I would never recommend is to try to dive in and build the entire OS without a good level of familiarity of the compiler setup and how RISC OS itslef works.

Seriously, how many entry-level people do you think would do such a thing? I’d like to, but more to see how it all works than anything else. Anything I have to contribute at this point can be loaded at runtime…

To be honest, I don’t really see why RISC OS has to be so “free”.

Delusions of Linux. Linux (and some others) is free, so everything must be free.

The C tools are one way for ROOL to fund this site, and their show attendance etc.

Exactly. You get the damned OS and a full start-up environment for free – my ARMwiki has a guide for setting up an emulator and installing RISC OS. Takes about five minutes.

What next? TechWriter and OvationPro for free ’cos the other guys have LibreOffice?

We have GCC. If that isn’t good enough, tough. I can’t get OakDraw to compile using OpenWatcom as it is missing some stuff Microsoft’s C provides. I should be using the correct compiler, not whinging.

Do we really need to devalue RISC OS and its software further?

To be honest, I’m surprised it is only £40. D’you know how expensive the version 4 was? D’you know how expensive proper dev suites are? The official ARM suite?

What does one mean by “truly free” anyway? I know I can build my own custom Ubuntu, but such a thing has never occurred to me as it’s way beyond my skill set, building an OS of that size/proportion.
When you start digging further, you’ll find that making the tools free isn’t free enough. As soon as ROOL offered tools for free, people would want the licence dumped and turned into some form of (the non-free) GPL. What a way to kill RISC OS…

tl;dr – you can code for RISC OS for free, you just can’t build the OS, but how many people are going to want to do that from the first day?

On the contrary, making development free and accessible adds to the value of RISC OS. It adds development time which people donate because of their enthusiasm, which is otherwise lost. There are more developers of this kind than there will ever be commercially available to the RISC OS market.

As an end-user I subscribed to both the Select and Unix Porting Project schemes. I’m convinced that my money was far better spent on the latter rather than the former because the work of the UPP can be downloaded free and the source code available for people to continue its development. Select development, on the other hand, travelled down a cul-de-sac, and the hundreds of pounds I spent has produced something closed and no longer available to anyone in a format useful for the latest hardware. And as my recent correspondance with RISC OS Ltd reminded me, will never be.

I’m amazed that people are still defending a business model which has failed. There are too few RISC OS users to fund the commercial development of software that will allow RISC OS users to do the things they want to do. The old model ensures a shrinking number of users (as users move to other platforms and the demographic ages).

It is the new open model which has allowed RISC OS to move to new hardware, to attract new users and to reduce barriers to access which are just unacceptable now that other excellent operating systems are available to run on cheap hardware with freely available commercial-quality software.

RISC OS Open’s bounty scheme shows people’s continuing willingness to inject money into RISC OS to speed-up development even if, and as I have argued above, because, it will all be made available free at the point of use and the source code available for all to edit and view.

I hope RISC OS’ commercial developers – of which there are almost none – embrace new business models or else they will soon disappear like RISC OS Ltd.

On the contrary, making development free and accessible adds to the value of RISC OS

Strongly agree. There’s no such thing as truly commercial development on RISC OS

(post deleted)

Select development, on the other hand, travelled down a cul-de-sac, and the hundreds of pounds I spent has produced something closed and no longer available to anyone in a format useful for the latest hardware. And as my recent correspondance with RISC OS Ltd reminded me, will never be.

Certainly, it is a great shame that two parallel versions of RISC OS existed, went in different directions, and are now unlikely to meet in the middle. Essentially, all of the work done by the people working for ROL will need to be re-implemented for our flavour of RISC OS – if anybody has the time and/or inclination.

I’m amazed that people are still defending a business model which has failed.

Failed?
I would imagine much of the work around here is volunteer work anyway. The bounties, nice as they are, are not a coder’s wage – which may explain why things are outstanding and some useful ideas (keyboard shortcuts, better scroll wheel support) have lapsed.
If you want to consider a failed model, consider that an open source project is only as good as the people working on it. Right now we have people doing amazing work to bring new platforms into the fold. There is much work still to be done, and no, we aren’t paying, the best we have is a bounty scheme. However as I noticed with the (long lapsed) bounties for my PVR – it’s one thing to offer an incentive, it’s a different thing to get people to bite.

Meanwhile, this site exists. And as long as it exists, there will be hosting charges, bandwidth charges, and all the administration stuff to be getting on with. This is purely for the website – in terms of ROOL itself, there are many more costs (shows, company taxes, blah blah). This is currently paid for in two ways – one, a bounty. And two, sales of goodies. I just hope the two are able to help cover the majority of costs involved.

to attract new users and to reduce barriers to access which are just unacceptable now that other excellent operating systems are available to run on cheap hardware with freely available commercial-quality software.

So, in other words, you think making everything free will turn things around?
Sorry, I don’t buy that.

Maybe we are the last of a dying breed and in the future software will be a free development of interested parties. However the danger here is that interested parties will develop only so long as they are interested. What is the incentive to dedicate your time, and how does that pay the bills?

The OS sources are available at no cost, the OS itself can be downloaded and installed at no cost. If you want the official dev suite, you pay for it (looking at the going prices of commercial compilers, £40 is quite reasonable). If you want to use RISC OS in a commercial product, you licence it. [the alternative, of course, is you piss off Stallman who will attempt to pollute a software licence with notes about the hackability level of hardware (but by now I’m sure you understand that I believe that Stallman has a very corrupt definition of “free”)).

I hope RISC OS’ commercial developers – of which there are almost none – embrace new business models or else they will soon disappear like RISC OS Ltd.

If I read you correctly, there’s a fundamental fail here: a business that does not make money is, by definition, not a business…

Wait, are we really arguing over a £40 dev suite? It has come to this? :-/

Rick:

Please read back what I wrote. Each point I made was not reflected in your writing, so I’ll briefly clarify here.

1) How do you maximise the potential of voluntary development? Do you charge developers who wish to contribute their time for free? Illogical.
2) Are you better off charging high prices for commercial products to a shrinking number of people or growing the number of users and charging lower prices?
3) If you’re a geek with time on your hands, you like to explore software, pick things up, play around with them, learn, hack a little: would you ever do that with RISC OS if you had to fork out £40 for development tools when you aren’t even sure you’re that interested in RISC OS yet? Answer: no, because there’s loads of other projects out there with no barriers to access. Thus we lose out on potential new developers.
4) Bounty system demonstrates, albeit modestly, that people are still willing to give money to developers even if results will be free to download. Particularly in the RISC OS market, users have thrown money at almost anything so we should maximise this potential. Perhaps RISC OS Open could gain a more stable income from a subscription scheme? None of my ideas prevent RISC OS companies from collecting money to fund commercial development, but it does mean they have to collect money in a different way: i.e. not selling products for fixed prices to a shrinking market place.

might make more sense in terms of lowering the barrier of entry?

Seriously, how many entry-level people do you think would do such a thing? I’d like to, but more to see how it all works than anything else.

Perhaps more than you’d think. I did, for one. And the only thing vaguely resembling programming I’d done previously (outside of obligatory FORTRAN stuff while studying) was BBC Micro type-in listings. (If that puts me in a minority category of “curiously non-technical anoraks” then so be it.) However, I thinks it’s entirely possible that competent developers from elsewhere might experience something along the lines of:

1. Interesting OS, I’ll try it
2. …A bit retro but still holds my interest
3. I like the way it does x
4. Why does it do y like that? I wonder…
5. It could do z instead… the sources are there… I’m happy to spend a few hours tweaking a certain aspect to see how it works
6. …Sod it, I have no way of building the thing
   1. Either: Order the tools – try again in a few days
   2. Or: Good luck to them – I may return when I can build it for free
7. …
8. Developer interest diluted or lost

The ROOL financing thing is a valid point. For some reason I’m under the impression that the Acorn/Castle/ROOL tools complete comparable builds more quickly than GCC under RISC OS? (Perhaps that’s incorrect.) So that may be a reason for people to consider “upgrading” from a potentially free build environment. But there’s also possible cross-compilation with GCC on a faster system (I think, ultimately), so I guess any potential speed benefits of the Acorn tools may be less of a factor.

If Castle’s licensing model ends up earning them a useful amount, perhaps they’d be prepared to partly sponsor the upkeep of ROOL for a period.

1) How do you maximise the potential of voluntary development? Do you charge developers who wish to contribute their time for free? Illogical.

Funny, some of us paid for our development tools. We also probably paid for the hardware we are developing on and for.
Don’t get me wrong, I do understand your angle here. I think ROOL/Castle/etc will need to consider if they have more to gain/lose by making the development software free, or not.
This ties up a second question – if it is made free, where’s the incentive to continue work on it (or do you also expect sources?). The ARM is progressing steadily, I would hope the assembler would be capable of continuing to be useful in the days to come. Is this better as a free product, or a pay-supported one?

2) Are you better off charging high prices for commercial products to a shrinking number of people or growing the number of users and charging lower prices?

Something that makes sense to me is to charge a lower price to more people. The problem is, does a lower price readily equate to more people?

3) If you’re a geek with time on your hands, you like to explore software, pick things up, play around with them, learn, hack a little: would you ever do that with RISC OS if you had to fork out £40 for development tools when you aren’t even sure you’re that interested in RISC OS yet?
Answer: no,

Duh. You can write software under RISC OS without a single thing extra. !Edit is built in and BASIC, as well as being a language in its own right, contains a useful and capable assembler.
If you prefer C, well, there’s GCC.

I would recommend the dev suite, and certainly it is a requirement for building RISC OS itself… however to suggest that you first have to fork out for the tools before you know if you’re interested in RISC OS is just plain wrong. I coded loads before I got myself the dev suite (and initially I got it for the C compiler).

Thus we lose out on potential new developers.

I wonder if other, more intrinsic, limitations will be a bigger factor than whether or not a specific compiler is free.

4) Bounty system demonstrates, albeit modestly, that people are still willing to give money to developers even if results will be free to download.

I wonder if the “free to download” is even thought about, or is the bounty scheme more a way to encourage your pet bug to be stroked…

Perhaps RISC OS Open could gain a more stable income from a subscription scheme?

Yeah… ’cos that worked so well with the other company…

but it does mean they have to collect money in a different way: i.e. not selling products for fixed prices to a shrinking market place.

I understand what you are saying, honestly. I’m just worried that this might be a catalyst of the beginning of the end for some. I would imagine there are few (if any?) RISC OS developers that don’t also do something else to make their main income. If we are to say “slash your prices”, they might reply with something unprintable. Certainly, it makes economic sense to charge more people less, but we also need to factor in the hassle quotient, plus if the product that cost £100 last week is now selling for £25, will the users expect the same level of commitment? Perhaps if sales increase fourfold it might work, but is this going to happen? Can anybody guarantee it?

I do believe that things will eventually need to change, from the RISC OS community right up to Europe itself. But you understand, there is a certain resistance to change when the shiny nice future outcome is only one of the possibilities…

However, please, don’t give people the impression that you can’t code for RISC OS without this £40 product. That’s simply not true.

What about setting up some sort of compile service charging £x.xx per run or per day with the option to buy the Norcroft suite for the difference to £40, if finally hooked to it.

It is probably a silly idea for two reasons: It’s unlikely that the licence for the suite allows such a service. And it would be more administrative work handling accounts and stuff for even less money.

At £40 this is a storm in a teacup.

To pluck the “wheat from the chaff” of my earlier post, I suggest that new people might be better served looking at the bundled apps to cut their teeth on (using GCC etc) before ploughing into the core modules. Whilst more programmers are appreciated, most of the recent progress has been from a small number of experts taking responsibility for high quality work, rather than broader dilution. Larger number of coders requires significant management/oversight, and I’m not sure we have the resources for that – especially with no corporate entities like Canonical/RedHat to fund salaries.

Writing code does not equal building the OS.
Most new users will only want to write apps and utilities, which as Rick says is eminently feasible with BASIC or GCC.
If they then find they need to help do things to the OS, £40 is no real barrier.
Its cheaper than a new GAME for XBox/PS3/others and will last longer….

Err, what about students who may rather spend their money on other things? They may still want to delve into the OS sources, and could have some spare time over the summer. Even this coming summer, in fact?

I don’t agree you have to be a ‘serious’ developer to want to build the OS, or that you have to have had plenty of experience with RISC OS before you start.

Example: somebody just posted on the ‘General’ board about running RISC OS on the RCA NC1020 (Hi iVan, apologies for taking you as an example :-). This is a machine I’ve never heard of, but it turns out it runs NC OS, and therefore should be capable of running RISC OS 5. Two problems… he’s probably the only person to ever try running RISC OS 5 on such a thing, and it may well have hardware different in some ways from an A7000+. Indeed, I don’t know how well the IOMD port itself supports the ARM7500FE anyway (it’s mostly been targeting Risc PCs), and I have no way of trying.

It appears the NC1020 was dumpster-fodder 10 years ago, so perhaps an NC1020 port isn’t of interest to many people. But if someone with coding skills on another OS is motivated because they want a port on their device, and we get better ARM7500FE support out of it, both sides win. And maybe they’ll become a more regular contributor too. But if they have to spend £40 to get something working on that machine they pulled out of a skip, they’ll just throw it back in the skip and walk away. For £40 they could buy a secondhand PC – the only interest is that the box they salvaged from the skip is in front of them now, and hacking code at 3am seems like a good idea at the time. If you put a barrier to that 3am coding session, it’ll simply never happen.

I think this stuff just has to be free; in fact, I just looked, I couldn’t even buy it if I wanted to! — it’s “temporarily unavailable!” (paypal + downloads anyone?).

Personally, I don’t think I’ve touched a BBC Micro or Archimedes for over 20 years, until I got my RPi the other day — I was thrilled when I found out that RISC OS is still going strong and that I might soon be able to put it on my RPi. Heck, I can plug it into the TV and have my kids learn to program on the grandchild of the BBC Micro that I learned to program on!

I totally buy into what the RPi foundation is doing; they’re right, the UK government has wasted 20+ years with their failed ICT policies teaching kids how to do Word, & Excel instead of real engineering. I think the RPi + RISC OS is the best chance of re-igniting the Acorn/BBC Micro/Arch “golden age” for the current crop of kids who’s minds are being rotted away by X-Box and Facebook. So you can get one of the Linux distributions for free, but RISC OS is much more accessible, it’s just there, instant on; you can type away right at the prompt. There’s no apt-get this, wget that, make/make install, finding an editor, installing gems etc. etc … with RISC OS, it’s all there (except I guess for the DDE!).

Yeah, it has to be free! I think of myself as a 11 year old when we got our BBC Model B, £399 I believe … it was a fortune at the time. But once I had it, I could write Basic, assembler etc., so that’s what I did. I never did buy the Pascal et. al. ROMs, they were too expensive for me as a kid, but I learned assembler instead and wrote ROMs for the BBC Master for fun, what a blast it was!

Ultimately, having an SD Image of RISC OS plus a decent set of apps, including a copy of the OS sources along with the tools to build it — that you can download for free and burn to a $9 2GB SD card for a $35 RPi — would be just fantastic! Build this and ROOL + RPi Foundation can go and lobby the DfE to provide a home-grown platform for Computer Science & Robotics Engineering in UK schools (… and maybe undo some of the damage that RM/Microsoft did starting in the 90’s).

One of the best things for me about getting a “new” Mac when OS X first came out was that the development tools were on the OS Install Disks, what a brilliant move!

Winston – so you coded with BASIC on the Beeb, you didn’t get the OS code, and you didn’t buy any of the AcornSoft programming tools?
How is this different to getting RISC OS on the RPi and…not buying the “official” compiler? You can still code, you can still write modules and stuff…

Theo – while I would not wish to put anybody off of working on RISC OS, I do think that somebody who does not understand the system is a very unsuitable person to port it. Part of the joy in porting is making it work without breaking a load of stuff elsewhere…
Take me for example. I have datasheets and lots of badly written notes to aid a port to the DM320 (a pre-OMAP ARM926 chip); but if I do it at all, it is going to come after I feel that I understand how RISC OS brings itself up (and to experience this in action, I am considering a 1V8 compatible JTAG gadget to trace boot-up on the Beagle – should I mention this bit of kit will cost twice what the dev tools cost?).
I believe the person that “hacks” at porting an OS they don’t fully comprehend is just setting themselves up for a world of pain.

(dupe post deleted, thanks Android…)

Or, to put it another way… Would you want A.Random Newbie dicking around inside FileCore to “fix a few things”?

Rick – The point was really about accessibility, in those days, buying the tools cost money and required a trip to a store — I would’ve loved to have a C compiler! I can understand a vendor charging say $1299 for the tools (as Microsoft did/do) because it’s a business, but £40 seems self-defeating. It’s enough to put people off (I think even £1 would put people off simply because of the PayPal rigamarole), but at £40 that’s not enough [given the market share] to generate the kind of income that allows you to hire a couple of dozen engineers to further develop the product. Unfortunately, I agree with a previous poster — RISC OS is dead commercially. I think it’s got a whole new life as an open platform (incl the tools), but people will be put off when they find out they have to buy stuff (e.g. the DDE and the CMOS Widget).

A.Random.Newbie has to learn somehow! In college (CS Major), we wrote our own Linux device drivers (I’ve written a number of device drivers for various platforms (commercially) since), in High School, we used to write our own ROMs that did various things, including a packet analyzer for Econet (on the Beeb). I assume I’d need the DDE now to do that for RISC OS. The biggest issue back then was lack of documentation (no internet to download the MC68B54 ADLC datasheets from; had to write to Motorola and wait a month for that!).

RISC OS is a great OS to learn with!

I assume I’d need the DDE now to do that for RISC OS.

Back in the day, BASIC came with a 6502 assembler so you could assemble ROMs and such.

These days, BASIC comes with an ARM assembler so you could assemble modules and such.
DDE greases the wheels, but it is quite feasible to write and build a mini-OS in ARM code using nothing more than BASIC… ;-)

RISC OS is a great OS to learn with!

Indeed. Minix is kinda cool, but has far too much Unixy baggage. I still believe RISC OS could have a place in the embedded world where a small lightweight OS is required that does not have all the baggage inherent in a generic Linux distro, when all that fluff isn’t required…

dicking around inside FileCore

CVS write access is only available upon request, so that’s perhaps not too much of an immediate concern. People could break things in the ROM and upload a version incorporating their changes somewhere. But I think such a case is unlikely to be confused with a legitimate ROOL distribution.

Being able to build the OS and associated components using GCC would mean that e.g. some of the minor bugs (including enhancements) could be investigated easily. This includes the ROM apps and Configure plugin, etc.

Changes could be discussed here in the forum, with write access granted if deemed a low risk. I think most people can generally be trusted not to stray into areas they know little about. There would remain the small risk of, e.g. mistaken deletions, changing components which don’t need changing. But IMHO ill-advised messing around (or the more extreme case of mass vandalism) is unlikely to present itself.

The unfortunate side effect of potentially granting write access to more developers (and encouraging discussion of changes here in the forum) is that more time may be needed by RISC OS experts for policing and commenting. This could be the price to be paid here, but (as Andrew pointed out) RISC OS has no corporate funding to help in this area.

"A good first step towards cross-compilation and support for alternative toolchains such as GCC would be to rewrite the makefiles one by one to use these makefile fragments. This would probably become a fairly mechanical process once you’ve written one or two. Any volunteers?"