Recent Posts

Topic: Bounties / USB Stack progress.

As it seems that I am no longer welcome in the initial thead on the topic of my rewrite of the USB Stack, I shall post my progress updates here in. I will make sure to post only progress updates, and no side topics, so as to insure the integrety of this thread.

I feel that it is important to keep every one somewhat up to speed as this will likely be the last time that I will be contributing any code to ROOL directly.

Currently (on the 24th of May in the year 2013 Anno domini) I am printing out the source of the USB Stack so that I may mail it to Steve Revill along with a soft copy on a disk so that my first submission may be included in the ROOL source tree. It has taken a few days to print out.

While printing out an hard copy I have been working on oprimizing what is already done, as well as reviewing the legibility of my comments in the source. I am spendng a little time ironing out the remaining issues.

The new USB stack is quite different from what I had originaly pictured.

I hope that this will be as helpful as it is intended to be.

Topic: Aldershot / The choice of programming language?

At the inderect request of COLIN , This will by nesecity be my last post to this thread. Colin accused me of Trolling, as such I am taking the appropriate action by not continuing in this thread any further.

I do hope that this sparks a produtive conversation as it is intended to. I believe that it is question the way that we program that pushes us to find the best way to create our works.

At the inderect request of COLIN , This will by nesecity be my last post to this thread. Colin accused me of Trolling, as such I am taking the appropriate action by not continuing in this thread any further.

Topic: Bounties / Would like to accept half of a bounty.

Peter Howkins:
You know it is frusterating to have to reprint so much. I am attempting to make a submission and did not think that there would be any need to make any modifications. I donot keep count of the number of files very well, though it looks like more than I was thinking.

It would not be so bad if this were just some toy project as I usualy do not print out those until 100% complete. Though I promised an early submision and as such have archived a copy of the source into a single file in order to print it out for submission, as this is a paid project so I will make all submissions in hard copy as is appropriate for any paid project.

Here I have already used 4 reams of paper, and I am sure that a notable percentage of that is in the first page of files (hich contain the license).

Further at the inderect request of Peter Howkins , This will by nesecity be my last post to this thread. Peter Howkins accused me of Trolling, as such I am taking the appropriate action by not continuing in this thread any further.

THUS:
I MUSTREQUEST THAT WHOM SO EVER MODERATES THIS FORUM LOCK THIS THREAD, AS THERE HAS BEEN MADE AN ACUSATION OF A TROLL

THE VALIDITY OF THE ACUSATION MAY NOT BE PRESENT, THOUGH THIS DOES NOT MATTER AS THE APROPRIATE RESPNCE TO THIS ACCUSATION IS TO LOCK THE THREAD, SO I RESPECT THE FEELINGS OF THOSE MAKING THE ACCUSATION (BY REQUESTING THAT THIS THREAD BE LOCKED).

Topic: Bounties / Would like to accept half of a bounty.

To quote Monty Python.

And since when can I ciew a Youtube video on RISC OS?

Having just reread this whole of this 150+ post thread. Well played DavidS, this is perhaps the finest troll I have ever read.I do feel a little sorry for those that were expecting a USB stack though.

I am sorry that you feel that way Peter Howkins.

I do understand that my thought pattern is scattered. My one and only point of focus is the projects that I work on. Currently that is one operating systm for the RPi SBC and this USB Stack. I refuse to give up on any one project until I have succeded 100%, though if this is the apriciation that I will get for doing so for components of RISC OS then this will be the last component of RISC OS Open that I spend any time to replace.

If you wish to judge my intention by my posts first you would have to know a bit about me

1. I am autistic, thus coherent thought is not a strong point for me, My thought pattern wonders more than any thing you may imagine.
2. Do to my autism I am also quite focusd on the area of my point, in m case this is Programming, and electronics.
3. Spelling and gramer are probably my worste subjects.
4. My first and prefered languae is ASL (Amarican Standard Sign Language), spoken language (and by extension the written construct there of) is quite awkward for me.

And I am sure that I missed a few important points, for one is always the least complete observer of ones self.

I guess I have to do something that I had always been apposed to before and edit the license for the code that I have been printing out for the last 4 days, so that I can send it in for inclusion in RISC OS, I will have to figure out how to explisitly omit permission for 1 person. That makes me sad (and not just because I will have to rprint the first page of every source file [I actualy wanted to provide something that would not be restricted in any way]).

ERIC RUCKER

Bliee what you wish. I noticed that you omited the fact that I stated that while the OS structure was complete in each of the abovecases, with the exception of QDOS the operating systems were not 100% API complete as they had only to run a small number of apps and thus only had to support enough of the API to do so. And I had spent a greatdeal of my time oig those projects to extreme disproportion to what most would. I do not time my self, though when working on a project it is common for me to wake up, spend the entire day working on the target project and forget to eat, only going to bed because I can not force myself to stay awake.

See above for some detail about me. You may believe what you want, though I would ask that you not intentionaly omit data in order to make something sound more extreme than it is.

Topic: Aldershot / The choice of programming language?

Classic stuff guys! As a matter of interest, do any of you ever sleep? I’m getting to read posts minutes after they are submitted and its mid morning in New Zealand so go knows what time it is over there!!!!

Topic: Bounties / Would like to accept half of a bounty.

(Apologies to James Rolfe for stealing that image…)

If you really believe that it would take so long I must assume that you have never written an operating system before? Yes it will take two to three years to properly clone RISC OS, I know this from experience: Cloning Amiga OS 3.1 for a commercial application (now dead) took about 9 months of my time (working alone), cloning Mac OS 7.1 for another commercial customer took 6 months (working alone), and cloning Sinclair QDOS took me 2 months (OK I admit QDOS is almost as simple as it gets).

Topic: Code review / USB IN endpoint buffersize

Re buffers. OK I can move it to usbmodule. The manual isn’t right for the Buffer module if the wordAligned flag is set (not documented in the copy of the manual I have) as in that case you lose 4 bytes – I vaguely remember the aligned buffer not being right somewhere – didn’t take too much notice as I was replacing the code. I’ll look into it. I think buffer size is only going to be critical in USB and only then for setting the buffersize the same as maxpacketsize. I may have fixed everyones buffers if they didn’t read the small print :-)

OK I admit it I found it annoying that the Buffer module lost 1/4 bytes :-) especially as it was easier to use it all.

Re ALTERNATE_FIX. I have a logitech webcam here and it has 3 interfaces 1 interface has 7 alternate interfaces (presumably the video interface) each alternate interface has endpoints 1 and 2 – one isochronous and one interrupt.

If I OPENIN "devices#endpoint1;interface0;alternate7:USB3 the current system will set endpoint=1 endpoint_type=bulk interface=0 alternate=0. There’s nothing you need to know about USB for the fix it just ensures that you get the setup you specify in the special field.

re bounty. Not bothered about the money. Just feel the NETBSD code is getting a bad press and if the RISC OS end is sorted then at least we can be more confident where the problem lies. Besides the USB sources are a good bedtime read – certainly send me to sleep.

Topic: Bounties / Would like to accept half of a bounty.

Having just reread this whole of this 150+ post thread.

Well played DavidS, this is perhaps the finest troll I have ever read.

I do feel a little sorry for those that were expecting a USB stack though.

Topic: Aldershot / The choice of programming language?

Wow. That reads like a mythology phonebook… bravo!

Topic: Community Support / iconmover

I’m not seeing what’s to fix? Drives have always appeared with the highest drive number to the left and lowest to the right (taking CDFS and ADFS as examples).

I know that. The Pandora is a special case. Two drives and it would be nice if the left icon would also be the left drive. You could also change the assignment of numbers.

Topic: Aldershot / The choice of programming language?

Well there is certainly labguage that I am apposed to. Never mind all the trolls; worry about all the bull-beggers, spirits, witches, urchins, elves, hags, fairies, satyrs, pans, fauns, sylvans, kit-with-the-candlesticks, tritons, centaurs, dwarfs, giants, imps, calcars, conjurors, nymphs, changelings, scritchowles, incubus, the spurne, the mare, the man in the oak, helwayne, the fire-drake, the puckle, Tom Thumb, hobgoblins, Tom-tumbler, Boneless and the rest, which are yet to make their digital appearances. Like Bishop Harsnet, we too live in an age of disinformation.

Topic: Code review / USB IN endpoint buffersize

This changes the Buffer module used by DeviceFS so that it uses all of the buffer size specified when registering or creating a buffer. The existing version creates a buffer 1 byte less that that requested.

I don’t think that’s the right place to make the change – it’s supposed to do that. See PRM4-88 which says “This call claims an area of memory from the RMA and registers it as a buffer. If you
register a buffer n bytes long, it can hold at most n – 1 bytes.”

So it’s the caller’s responsibility to add 1.

I’m not particularly a USB expert, so I’m not sure I can say whether the COLIN_ALTERNATE_FIX is sensible or not – do I remember right that someone had a USB analyser on their desk? Perhaps they can peer review that change for goodness.

I should add that in my view you deserve some of the USB bounty money. With the stated goal of “Update and debug USB stack” you’re taking a methodical approach to analysing the various sticking points and making incremental patches which solve them. This is exactly how I read “Update and debug USB stack” – the update bit being adopting the latest NetBSD sources and the debug bit being what you’re doing, rather than reimplementing the whole thing in Pascal, or whatever.

Topic: Community Support / iconmover

SDFS:0 is the Pandoras left slot, SDFS:1 the right one but the right icon on the iconbar is SDFS:0 and the left SDFS:1

Or you could just fix SDFSFiler

I’m not seeing what’s to fix? Drives have always appeared with the highest drive number to the left and lowest to the right (taking CDFS and ADFS as examples).

Topic: Aldershot / The choice of programming language?

DavidS is trolling?

Well, we hope he is trolling? :-)

Topic: General / ARMiniX experience?

On native hardware the best “disk” perfomance is by using SDFS (or FAT32fs on another partition) and a suitably quick SD-Card.

The limitation is the size as the maximum size of SD-Card that the PandaBoard ES can handle is 32GB in total.

Various of us that have gone down the DIY route have posted up speeds in other threads in these forums.

Project posh and Project cool (my DIY RISC OS PandaBoards) just use 8GB cards and I store everything else on a NAS as the ethernet is also over USB so there’s no point in having a local large hard drive.

Topic: General / ARMiniX experience?

The answer I got was that it’s a USB-to-SATA bridge, so it won’t give speed advantages over using USB.

Topic: Community Support / iconmover

I say SDFS, I mean SDFSFiler, wherever that is.

Topic: Code review / USB IN endpoint buffersize

I’d like to submit these changes. These fixes fix a couple of problems with setting options in the USB special field string. There are 2 modified files

castle.RiscOS.Sources.HWSupport.Buffers.s.Buffers

This changes the Buffer module used by DeviceFS so that it uses all of the buffer size specified when registering or creating a buffer. The existing version creates a buffer 1 byte less that that requested. All I’ve done us use a separate variable for freespace size instead of using the difference between the insert and remove pointers.

mixed.RiscOS.Sources.HWSupport.USB.NetBSD.build.c.usbmodule

This fixes start_read so that it will fetch if freespace in buffer is >= an endpoint maxpacketsize. At the moment it won’t fetch if endpoint maxpacketsize = freespace in buffer.

The changes are tagged COLIN_BUFFER_SIZE_FIX in both files. I’ve kept the original next to the changes so you can see the change.

This change will mean that a special field entry for an endpoint of size will work as expected ie usb will see a maxpacketsize buffer.

Also in usbmodule tagged COLIN_ALTERNATE_FIX I’ve fixed the following.

At present if endpoint and any or both of interface and alternate is specified in the special field then interface and alternate are ignored in favour of the interface used by the found endpoint in the configuration descriptor and alternate is set to 0.

Additionally if the endpoint_type is not specified it is set to bulk which may not be correct for the endpoint and can be determined from the endpoint descriptor.

So I’ve changed things so that if just endpoint is specified endpoint_type, interface and alternate are set from the configuration descriptor.

The change also results in an error if a specified endpoint/interface/ alternate combination isn’t found. At the moment if endpoint exists the other options are ignored but it may be that the endpoint number is in error.

With these changes you can

OPENIN “devicefs#endpoint1;size8:USB1”

and get the correct endpoint_type and an 8 byte buffersize – behaviour
if you don’t specify the endpoint is as it was before ie its set to the
first bulk IN endpoint.

I’ve included a the changed files in USBBufferFix.zip. It includes softload of RISC OS for an Iyonix if you want to try it. For those who haven’t got the latest riscos and are playing with USB, all special field strings work correctly in this version eg nopad and noblock and order doesn’t matter. I’d be interested if anyone tried the changes.

I’ve only tested the changes on an Iyonix.

Topic: General / ARMiniX experience?

SATA: I for one will be interested in CJE’s response. At the moment my RISC OS platform of choice is RPCEmu0.8.10 running RO5.19, which which on my Win7/64 PC gives approx. 98MB/sec HD reads, and 71MB/sec writes under RISCOSmark 1.01 (1MB file in each case). OTOH the ARMiniX/PandaRO CPU performance would be much better – I’m getting 610k MIPS.

Topic: Community Support / iconmover

2. Or you could just fix SDFS.

It would be nice if I could. Unfortunately I’m too stupid.

Topic: Aldershot / The choice of programming language?

DavidS is trolling?

Topic: Bounties / Would like to accept half of a bounty.

Interestingly, although it is absent from http://www.netsurf-browser.org/downloads/, it is present at http://ci.netsurf-browser.org/builds/

Topic: Bounties / Would like to accept half of a bounty.

Because Windows is not yet a supported platform and the port is unfinished.

Topic: Bounties / Would like to accept half of a bounty.

Windows builds of NetSurf

Thanks. I wonder why it isn’t on the downloads page?

Topic: Aldershot / The choice of programming language?

Rick, you’ve been caught hook line and sinker.

In what way?